Available in English, Special Serie French and German

What can you expect from IRIS Special What is the source of the IRIS Special in terms of content? expertise? Published by the European IRIS Special is a series of publications from the European Every edition of IRIS Special is produced by the European Audiovisual Observatory that provides you comprehensive Audiovisual Observatory’s legal information department in Audiovisual Observatory factual information coupled with in-depth analysis. The themes cooperation with its partner organisations and an extensive chosen for IRIS Special are all topical issues in media law, which network of experts in media law. we explore for you from a legal perspective. IRIS Special’s The themes are either discussed at invitation-only workshops approach to its content is tri-dimensional, with overlap in or tackled by selected guest authors. Workshop participants some cases, depending on the theme. It offers: and authors are chosen in order to represent a wide range of 1. a detailed survey of relevant national legislation to facilitate professional, academic, national and cultural backgrounds. comparison of the legal position in different countries, for example IRIS Special: Broadcasters’ Obligations to Invest in IRIS Special - unique added value Cinematographic Production describes the rules applied by IRIS Special publications explore selected legal themes in a way that 34 European states; makes them accessible not just to lawyers. Every edition combines 2. identification and analysis of highly relevant issues, a high level of practical relevance with academic rigour. covering legal developments and trends as well as suggested While IRIS Special concentrates on issues and interactions solutions: for example IRIS Special, Audiovisual Media within Europe, it takes a broader geographical scope when the Services without Frontiers – Implementing the Rules offers theme so requires. a forward-looking analysis that will continue to be relevant long after the adoption of the EC Directive; Who needs to read IRIS Special? 3. an outline of the European or international legal If you hold a position of responsibility in the media industry, context influencing the national legislation, for example particularly cinema, broadcasting or new media, if you work for IRIS Special: To Have or Not to Have – Must-carry Rules a specialised law firm, for a regulatory authority or in media explains the European model and compares it with the and cultural policy making, or if media studies is your academic American approach. discipline, then the IRIS Special series is a must for you.

So far published in the series: n Editorial Responsibility n Tomorrow’s Delivery of Audiovisual Services SearchingSearching forfor ISBN 978-92-871-6476-6 - 50 pages - © 2008 – EUR 75 ISBN 978-92-871-5707-2 - 56 pages - © 2005 - EUR 35 http://www.obs.coe.int/editresp http://www.obs.coe.int/tv-future n Legal Aspects of Video on Demand n Political Debate and the Role of the Media Audiovisual ISBN 978-92-871- 6391-2 - 91 pages - © 2008 - EUR 89 Audiovisual ISBN 978-92-871-5675-4 - 125 pages - © 2004 - EUR 44 http://www.obs.coe.int/vod-legal http://www.obs.coe.int/political-debate n The Public Service Broadcasting Culture n Regulating Access to Digital Television ISBN 978-92-871-6188-8 - 184 pages - © 2007 - EUR 62,50 ContentContent http://www.obs.coe.int/psb ISBN 978-92-871-5401-9 - 127 Pages + 28 pages Glossary - © 2004 - EUR 44 n Audiovisual Media Services http://www.obs.coe.int/digital-tv-access without Frontiers n ISBN 978-92-871-6115-4 - 76 pages - © 2006 - EUR 58,50 Co-Regulation of the Media in Europe http://www.obs.coe.int/av-media-services ISBN 978-92-871-5141-4 - 140 pages - © 2003 - EUR 27 http://www.obs.coe.int/co-regulation n Broadcasters’ Obligations to Invest in Cinematographic Production n Jurisdiction over Broadcasters in Europe ISBN 978-92-871-5971-7 - 127 pages - © 2006 - EUR 62,50 ISBN 978-92-871-4854-4 - 64 pages - © 2002 - EUR 27 http://www.obs.coe.int/tv-invest-film http://www.obs.coe.int/jurisdiction

n To Have or Not to Have - Must-carry Rules n Television and Media Concentration ISBN 978-92-871-5897-0 - 52 pages - © 2005 - EUR 58,50 ISBN 978-92-871-4595-6 - 89 pages - © 2001 - EUR 27 http://www.obs.coe.int/must-carry http://www.obs.coe.int/media-concentration

European Audiovisual Observatory: tel. +33 (0) 3 88 14 44 00 - fax +33 (0) 3 88 14 44 19 - E-mail [email protected]

IRIS_Special_Searching for Audiovisual Content_Cover [6mm].indd 1 2/12/08 16:35:39 IRIS Special: Searching for Audiovisual Content

European Audiovisual Observatory, Strasbourg 2008 ISBN: 978-92-871-6559-6 EUR 89

Director of the Publication: Wolfgang Closs, Executive Director of the European Audiovisual Observatory E-mail: [email protected]

Editor and Coordinator: Dr. Susanne Nikoltchev, LL.M. (Florence/Italy, Ann Arbor/MI) Head of the Department for Legal Information E-mail: [email protected]

Contributing Partner Institution:

Institute for Information Law (IViR) Rokin 84 NL-1012 KX Amsterdam Tel.: +31 (0) 20 525 34 06 Fax: +31 (0) 20 525 30 33 E-mail: [email protected] URL: http://www.ivir.nl

Editorial Assistant: Michelle Ganter

Marketing: Markus Booms E-mail: [email protected]

Translation/Proof-reading: Caroline Bletterer, Alison Hindhaugh, Amélie Lépinard, Bernard Ludewig, Stefan Pooth, Erwin Rohwer, Roland Schmid, Dorothée Seifert-Willer, Nathalie Sturlèse, Candelaria Van Strien-Reney, Anne-Lise Weidmann

Typesetting/Print: Pointillés, Hoenheim (France)

Publisher: European Audiovisual Observatory 76 Allée de la Robertsau F-67000 Strasbourg Tel.: +33 (0)3 88 14 44 00 Fax: +33 (0)3 88 14 44 19 E-mail: [email protected] http://www.obs.coe.int

Please quote this publication as: Susanne Nikoltchev, Ed., IRIS Special: Searching for Audiovisual Content (Strasbourg, European Audiovisual Observatory, 2008)

© European Audiovisual Observatory, 2008. Opinions expressed in this publication are personal and do not necessarily represent the views of the Observatory, its members or the Council of Europe.

IRIS_Special_Searching for Audiovisual Content_Cover [6mm].indd 2 2/12/08 16:35:39 IRIS Special Searching for AudiovisualContent

In 1989/90 Tim Berners-Lee developedthe (WWW) forthe European Organization forNuclear Research (CERN). Barely 20 years old,ithas made an unprecedented impact on theway we deal with information. TheWWW has becomeakey resource forprivate life andbusiness.Toeach of us it offers an unimaginable variety andquantity of contentand it calls on each of us to also use it in ordertoshareour own creationswith others.The WWW nurtures our expectationsofwhat information we can find, it influences our habits of howand wheretolook forcontent, anditgenerates fears as to whether we master themachineorthe machinemasters us.

CERN was notonly theWeb’s midwife,itmight also serve as its role modelfor generating giganticamounts of information. CERN expects to producemorethan 15 millionGigabytes of data each year as aresult of acceleratinghundreds of millionsofsubatomicparticles in theLarge HadronCollider(LHC), theworld’slargest particle accelerator. This bears some resemblancewith theswellingmass of information floatingonthe WWW.What drives both systemsisour desire to gather information.

When CERN launched its recentexperimenttoreconstruct thebig bangthat gave birth to our planet, critics remindedusthat notall thattechnology can do is without risk. Certainly the responsible managers andscientists at CERNwill do all they can to make their experimentsafe, yet they also admit that asingle mindcan no longercontrol andoverlook an experimentofthat magnitude.Accordingly,supervision is carried out by ateam with each of themanyteam members handlingonly asmall part of theproject.

Arethere similar developments to be noticed forthe WWW? Arewestill in aposition to handle it or have we created an information monster that asingle humanbeingcan no longercontrol? Arewegradually beingwashed away by adelugeofinformation? These questionsmust lead us to reflect about theprincipal role or even several principal roles that theelectronicinformation supply plays in our lives: I. Theamountofinformation electronically processed is mind-boggling. Oneofthe contributionstothis publication mentions that in 2008 electronically created,stored and replicated information reached theequivalentofthree milliontimes theinformation in all books ever written andthat theamountwill multiply by six over thenexttwo years. No doubt somebodylookingfor relevantaudiovisual contentfaces abig challenge.The fact that we still manage to findwhat we arelookingfor is largely owed to electronichelpers such as search engines andelectronicprogrammingguides.

In short: theabundanceofelectronicinformation compels us to develop andtrust in new search facilitatingtools.

II. Themorecontentfromanincreasingnumber of sources is stored on arapidly growing number of platformsand made availabletomoreand more users withincreasingly diverse backgrounds,the more difficult it is to provide acoherentsystem that will enable everybodyto achieve satisfactory search results.The successful search foraudiovisual contentisaheightened challenge because most existingsearch tools aretext-based.

Acoherentapproach to render data searchable requires acoherentmethodology for identifyingand classifyingthem. We need systematicorderingand taggingjust likewedid when we could still tidy away printed material in bookshelves.

In short: theabundanceofelectronicinformation mightnecessitateuniversal content identification systems.

III. Theease with which anybodycan put anythingabout anysubject onlinehas a significantimpact on howwecan intrude upon each others lives.Wemay findonlinethe most irrelevantand even manipulated but also themost pertinentinformation on anysubjectand any person. Moreover,weourselves mightbethe targetofsuchinformation without even knowing that it exists.

In short:the abundanceofelectronicinformationrenders anycontrol about personal information impossible.

IV.The need to channel or filter thevast array of information available on theInternet so that each mayfindthe relevantpieces mayclash withthe righttofreedom of expression. When search engines andelectronicprogrammingguideslist relevantsources they also always decide what information to exclude from that list.

In short: Should theuser be contentthat theselection amongthe abundanceofelectronic information is made by someoneelse andwith thehelp of an automated system?

V. Given theimportance of search engines andelectronicprogrammingguidesfor finding andaccessing information, isn’t thereaneed to control their functioning? Who, if anybody, steps in if information is excludedthrough thealgorithm of asearch enginefor no apparent reason? Accordingtowhich rules do search engines andelectronicprogrammingguidesproduce their results?

In short, does theEuropean regulatory framework address audiovisual search tools?

VI. Notbychance has theWorld Wide Webalso been referred to as theWorld Wild Web. It seemsthat regulators arenecessarilylimpingbehindtechnical developments andcannot move as fast as theWorld Wide Webexpands.What can we,indeed,what must we do in ordertoensure that theonlineinformation supply is in linewith theEuropean value system? Howdowe, for example,securethe delicate balance between thefreedom of information andthe rightto privacy? Isn’t thereastrongneed to regulate? Could self- or co-regulatory mechanismsachieve what traditional regulation maynot?

In short: Who, if anybody, can adequately delineate thelegal framework forsearching audiovisual contentinanelectronicworld of abundance?

These six aspects arefurther developed in thesix chapters of this IRISSpecial emphasising thelegal context in which they becomerelevant. Thehighly complexmatter of this publication also provoked an intense exchange of views duringaworkshop co-organised by theInstitute for Information Law(IViR) of theUniversity of Amsterdamand theEuropean Audiovisual Observatory in April 2008. Thegist of thediscussion is summarised in theWorkshop Report with which this IRIS Special begins.

We would liketothank all participants to theworkshop fortheir committedcontributionsto thediscussion. Their namesare listed at theend of this publication. Especially,wewould like to express our gratitude to our colleagues at IViR whowereinstrumental in puttingtogether andleadingusthrough theworkshop as well as to theauthors of thesix articles andthe summarising report. Nico vanEijkand Christina Angelopoulos of IViRprovidedadditional support by pre-editingthe original (English) texts.And we liketopointout once again that the success of theIRISseries is also thesuccess of highly qualified andmotivatedtranslators and proofreaders.Wethank them fortheir valuable work on this edition.

Strasbourg,November 2008

Wolfgang Closs Susanne Nikoltchev Executive Director Head of Department forLegal Information

IRIS Special Searching for AudiovisualContent

SEARCHING FORAUDIOVISUAL CONTENT

CONTENTS

Workshop on Audiovisual Search Summaryofthe Discussion ...... 1 Christina Angelopoulos,Joris vanHoboken

Techno-economic Challenges of AudiovisualSearch Engines ...... 13 Ramón Compañó

The Future of Content Navigation ...... 31 JanetGreco

Privacy and the Drive for the Perfect Audiovisual ...... 37 Michael Zimmer

Freedom of ExpressionImplications for the Governance of Search ...... 49 Joris vanHoboken

In Search of the Audiovisual Search Tools in the EU Regulatory Frameworks ...... 65 Peggy Valcke

Regulating Search Engines? On the Use of Self- and Co-regulation in the Field of InternetSearch ...... 79 WolfgangSchulz

SEARCHINGFOR AUDIOVISUAL CONTENT 1

Workshop on AudiovisualSearch Summary of the Discussion

ChristinaAngelopoulos,Joris vanHoboken Institutefor Information Law (IViR), University of Amsterdam

Introduction

On 12 April 2008, theInstitute forInformation Law of theUniversity of Amsterdam (IViR) andthe European Audiovisual Observatory held ajointexpertworkshop in Amsterdam on thetopic of “Audiovisual Search -Regulatory Challengesfor Audiovisual Abundance”.The aim of theworkshop was to stimulate an exchange of ideas on thefutureofaudiovisual search andthe relevant regulatory issues.Tothis end, eightseparate presentationsweremade,eachfollowedbyaroundtablediscussion. Thefirst set of presentationselucidated thecurrentpractical challengesfacingthe contentnavigation business.The secondconcentrated on an analysis of thenormativeand regulatory framework within which this business operates.This workshop report provides asummary of theopinionsexpressed and conclusionsreached duringthe discussion rounds.Athematic,rather than achronological, approach has been taken.

What becameapparentfromthe presentationsand subsequentdiscussionsonaudiovisual search services,the market andthe lawsgoverningthe audiovisual contentvalue chain wasthat this is a rapidly developingand extremely dynamic environment. Oneofthe greatest challengesfor thefuture will be to keep existinglawsand regulation up to date andtofacilitate thestill developingmarket in audiovisual search technology andservices.Fundamental questionsabout thefreeand effectiveflow of informationbetween information providersand end-usersand theimportantissue of privacy protection shouldinform thedebate about thepossibleneed to regulate audiovisual search engines. As search services move to thecentreofthe information environment, futureinformation law and policywill havetoaddress these new entities more adequately.

1. Current PracticalIssues in Audiovisual Search

1.1.The BusinessModelsofSearch Engines andEvolving Search Technology

Oneofthe first topics to catch theattention of participants wasthat of thecurrenttrendsin audiovisual search, as well as in generalsearch services,and theoperation, in particular,ofthe relevant market. Special attention waspaid to thequestion of entry barrierspreventingnew players from joining. It wasagreed that themain barriersare theoperational costsofcrawlingand indexing, on theone hand, andquery answering, on theother,assuggested by presenter Ramón Compañó.1

1 )See thearticlebyRamón Compañó in this publication.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 2 SEARCHINGFOR AUDIOVISUAL CONTENT

Followingonfromthis,the theory wasput forth that,apart from financial costs, other elements, such as detailed informationtobetter targetaudiences,mightalsocreate addedhurdles.Established search engines arecompilingwhat havebeen called “ of intentions”,2 that enablethem, through profiling, to match an individual’ssearch patternswith those of others, helpingthemtoenhance their efficiency andpopularity andoffer acompelling wayoftargetingaudiences forthe benefit of advertisers. If this emergesasanessential item forprovidingasuccessful search tool, newcomerswho havenot hadthe chance to gather similar data will findthemselves at adisadvantage.Consequently, it wassuggested that personalised servicesmighttherefore result in locked-in effects: at themoment, changingservice provider is only aclick away.If, however, asearch engineholds personal data on its users, this can entrench them, creatingobstacles forcompetitorstoprovidecomparableservices.

Does this mean that theInternet search market is oneofnatural monopoly? Participants were inclined to reject theidea, particularly in viewofthe following:

1. Google itself, now thepredominantplayer,was in factalate-comer.The first search engines,such as AltaVista, were eventually squeezed out of themarket. 2. Googleisindeed themarket leaderinthe US andEurope,but this is nottrue forother partsof theworld. It is possiblethat language mightbeanobstaclepreventingexpansion to certain regional markets. 3. Finally,subject-specific search engines arecurrently appearing, indicating niches in themarket abletoaccommodate additional players.

Thepresentation of Thomas Roukensintroduced thebusiness modelofthe largest Belgian cable network provider Telenetand thus provided amoreconcrete basis foranalysingexistingbusiness practices.Aninitial observation concerned thepractice of vertical integration that Telenethas been pursuing. Although someparticipants were critical of such integration, it wasconceded that, due to thelinguistically-boundedFlemish market forwhich thecompanycaters, thecase in point is exceptional. It wasnoted that initially local broadcastersand contentproducerswerewary of such digital services,believingthat theInternet wouldwipe away their viewers andrevenue. Telenettried to overcomethis attitude by creatingacollaborativemodel with contentprovidersand broadcasters andhighlightingthe complementary natureofthe additional services it provided. As aresult, notonly did thetransactionsfor Telenet’sVideo on Demand(VoD) service,iDTV(Integrated Digital Television), increase,but revenuefor local broadcastersand contentproviderswas also boosted andtheir position safeguarded. It wasasserted that, in asmall language community,vertical integration through collaboration can be theonly wayofsurvivingthe competition of bigger broadcasters.

Another topic of discussion highlighted theevolvinginteractivedesign of Electronic Programme Guides (EPGs). Beforelaunchingits ownservice,Telenet wassceptical, questioningwhether themodern consumer wouldnot be more attracted to somethingmoreflashyand elaborate.Yet thesimplicity of theservice offered seemed to be precisely what drew consumerstoit. In addition, consumersmade enthusiastic useoffeatures enablingthe recordingofprogrammes.The EPG eventually becamethe centrepointofTelenet’sinteractiveplatform. Participants were puzzled by this development, asking whether it does notinfacttransform theEPG into arecordingdevice.Telenet, however, has notgained this impression, since theadditional informationonprogrammes is notsimply still availablethrough theEPG, but also widely used.

1.2The ImportanceofMetadata

Janet Greco’spresentationonthe need foraconsistentsupply of metadata foruse in EPGs turned theconversation to therelevant intellectualproperty rights. To what extentdointellectual property rights exist overthe metadata, it wasasked, andwho holdsthem? Participants quickly pointed to the Magill case. 3 Thecase upheldthe imposition of acompulsory license on television companiesto remedy theexercise of their exclusiverights undernationalcopyright legislation that prevented publishersofweekly guides from copyingtheir listings. Thecase indicatedthat therewas indeed

2) Seethe articlebyRamón Compañó in this publication. 3) Joined Cases C-241/91 Pand C-242/91 P RTE andITP v. Commission (Magill) [1995] ECR I-743.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 3

copyrightovermetadata in theform of TV listings, but established that, in exceptional circumstances, theexercise of an exclusiverightbyaproprietor may constitute abusiveconduct. As oneparticipant put it, this reasoningamounted to aconclusion that competition law can in somecases “override” intellectual property.

Thesubject gaverise to an exchange of opinions: on theone hand, it wasfelt that thereare strong incentives to protect intellectual property rights over, forexample, descriptionsand ratings of audiovisual content. Areference to aprogrammeisvaluablesince it reinforces society’scapacity to compareand choose.The generatorofthis addedvalue,therefore,shouldindeed be abletorequest control overhow it is used. Hence, legislation such as theEU’sDatabase Directive, which in essence protects thesweat of thebrow that is invested in compilingsuchinformation, is justified.Onthe other hand, however, it wasalsoasserted that from abusiness perspective, this approach creates complexities.One participant, accordingly,was of theopinion that copyrightoversuchinformation as thetitleofafilm or thenameofits director,asopposed to e.g., asynopsis,shouldnot be recognised, as these aremerepoints of fact. In an environmentwhere copyrightisacknowledged, an aggregator strugglingtoput together consistentand correct metadata has to jugglethe management of business relationswith countless diverse publishing information sources.Concerningourselves with theownership of thecontent, ranthis argument, is missingthe wholepoint: thegoal of enabling consumerstofindcontentthat meetstheir interestsand has all therequiredlabellingfromthe regulatory pointofview.

Yetother participants viewedthe very idea of consistentmetadata as unattainableinaninterconnected world. Organisingprogrammeproducersordistributorstothe extentofenablingsystematic informationinput andclassification accordingtointernational standards is difficult. Moreover, the contents of therelevant databases areconstantly changingand therate of change is increasing. Yet theemergenceofthe couldmakethese problemsirrelevant.Search engines on theInternet operate on an entirely differentbasis,but arearguably better equipped to deal with thedisarranged state in which metadata currently finds itself. Through theuse of algorithms, information retrievalis effective, if messy. Small inconsistencies,suchasthe useornot of an actor’smiddleinitial, no longer compromise search resultssignificantly.Asone of theparticipants explained, search resultsdonot havetobe100% accurate to be serviceable: They simply havetobegood enough to enablearobust search tool. In this setting, it wasasked whether EPGs arenot an artefactofapre-Internet world.

Ralph Traphöner’spresentation, after all,which analysed theworkings of theTHESEUS project,4 demonstrated that theability of new search technology to deal with inconsistentdata setsand extract informationand metadata from an uncontrolled decentralized information environmentisone of the coreinnovationsinmodern information retrievaltechnology.Furthermore, audiovisual data, from a technological pointofview, is notfundamentally differenttotext: in ordertoobtain metadata that describes text, keywordsare extracted from thestrings. When it comes to audiovisual material, the paradigm is thesame, although more sophisticated algorithms andcomputer powermightbenecessary. Theirrelevanceoftraditional EPGs will becomeevengreater,if, as oneofthe participants believedwill happen, it becomes possiblenot only to search for(moving) imagesthrough theinput of metadata, but also through thecomparison of theimagesthemselves.

Theprecedingobservation turned thediscussion to theroleconsumersthemselves can play in categorizingand valuingaudiovisual content. Theexamples of theCompact Disk (CDDB)5 and theInternet Movie Database (IMDb)6 were cited. In such models, thecollection of data is partly placed

4) TheTHESEUS project is aresearch programmeinitiated by theGerman FederalMinistry of Economyand Technology with the goal of developinganew Internet-based infrastructure, so as to better useand utilize theknowledge availableonthe Internet. At thecurrenttime, 30 reasearch institutions, universities andcompanies havejoined theprogrammeunderthe coordination of empolis GmbH. 5) TheCompactDisc Database (CDDB) is aInternet-accessibledatabase that enables softwareapplicationstolook up information on audiocompact disks. Theoriginal softwarebehindCDDB wasreleased underthe GNU General Public License andwas based on voluntary contributionsfromusers. Theproject waseventually incorporated as CDDB LCCin1998. It wasthenboughtby high-tech multimedia electronics manufacturer Escientand,in2000, renamed Gracenote. 6) TheInternet Movie Database (IMDb) is an onlinedatabase of information andphotosrelated to film, television shows, actors, production crew personnel andvideo games.The IMDb waslaunched on 17 October 1990 andwas acquired in 1998 by Amazon.com.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 4 SEARCHINGFOR AUDIOVISUAL CONTENT

in thehands of theconsumers, whohaveagreat incentivetoprovideinformation andthe best knowledge of their needs.Suchsystemscouldtherefore prove highly operable. Again, however, the question of ownership surfaces: it is indicativethat CDDB waseventually incorporated, soldand relaunched as Gracenote,awholly ownedsubsidiaryofSonyCorporation of America, whilethe IMDb is now ownedbyAmazon.

In anycase,itwas suggested that thesolution can only arise from themarket. Database companies arelikely to takethe lead andevolveinto thegatekeepersofmetadata. Oneparticipantevenposited that thesolution couldfinally be foundincompetition: in contrast to thesinglepan-European PEGI classification system forvideo-games,inthe realm of film Europe boasts27differentclassification boards.Possibly this is an advantage that gives consumersaninsightinto amorebalanced view, insteadofobligingthemtoaccept asingleperspectiveaswritten in stone.

2. Fundamental Rights Perspectives on Audiovisual Search

2.1. Audiovisual Content Online and Privacy

After Michael Zimmer’spresentation on theFaustianBargain we currently face, as audiovisual searchtechnologysimultaneouslyboth enhances reachand recall of information andjeopardises users’ privacy, 7 an animateddiscussion ensued. Duringthe discussion, average users’ knowledge of the threatstotheir privacy andtheir interest in protectingthemselves wasquestioned. Theobjection raised to this lineofthinking, however, wasthat privacy violationscan be difficult forindividualsto perceiveonanabstract, de-contextualized level. On thecontrary,ifthe problem affectsthemina personal manner,i.e., when their ownsense of privacy is compromisedinreality,privacy concernsare muchbetter understood. After all, search engines havebecomesointegrated into our lives that the average user’ssensitivity to theprivacy threat they mightpose has been dulled due to overexposure. Thetrust with which search engines inspireusisthe key to their success.

It wasfurther suggested that this problem is aggravated in thecase of theso-called “information have-nots”.The concern wasexpressed that those whoare notconnected andwho thereforedonot enjoy thebenefitsoftechnology mightfindthemselves,inaddition,inaweaker position vis-à-vis privacy on thenet. Thenotice andtakedownprocedureadopted by therelevant regulatory authorities of most countries quickly reachesits limit whereone does nothaveorcannot master thetechnical meansthat enableawareness of invasionsinto privacy.

Finally,the efficacy of thecurrenttrend foruser education as aform of defence wasdebated. It was pointed out that this presupposes auser notonly awareofthe existence of theproblem, but also capableofretainingvast volumes of highly specialised information on everythingfromcopyrightand privacy legislation to consumer protection.

Turningtothe possiblesolutionsavailable, certain participants discerned an obligation forboth academia andpublic advocacy to intervene. It wassuggested that public advocates oughttoraise issues anddefendrights that concern all, but offer to no oneindividual apersonal motivefor immediate action. Their role couldbeespeciallybeneficialfor theweaker membersofthe information society. Additional regulatorysolutionswerealsoenvisioned: theimposition of obligations such as face- blurringfor services likeGoogleStreet View 8 formsacase in point; such measures,ofcourse,takethe process to theopposite pole, establishingprotection as thedefault. Theprovision of educational information as apublic service,tothe endoffacilitatinguser self-edification, wasalsoadvanced.

It waspointed out that manyofthe privacy issues that arise from theincreasingly sophisticated audiovisual search services relatingtopeopleare,inone wayoranother,coveredbyEUdata protection law.Inthis context, therecentopinion on data protection issues relatingtosearch engines issued by

7) See thearticlebyMichael Zimmer in this publication. 8) TheGuardian, “GoogleBlursthe Privacy Issue” (May 2008), availableat http://www.guardian.co.uk/business/2008/may/13/google.digitalmedia (accessed 15 June2008).

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 5

theArticle29Data Protection WorkingParty wasmentioned, as havingreached noteworthy conclusions, forinstance,with regardtothe useoffacial recognition softwarebysearch engines: “Search engineprovidersthat specialise in thecreation of value addedoperations, such as profiles of natural persons(so called ‘peoplesearch engines’) andfacial recognition softwareonimagesand audiovisual contentmusthavealegitimate groundfor processing, such as consent, andmeet all other requirements of theData Protection Directive, such as theobligation to guarantee thequality of data andfairness of processing”9 .This again places theburdenofprotection on theservice provider.Ona relevant note,another solution proposed wouldinvolveprivate companies crawlingthe webinsearch of privacy-endangeringdata. Thedrawback of this solution, it wasnoted, is that it mightonly benefit thefinancially robust.

In amoretechnical realm, solutionsfor control overpersonal data andits processing can also be anticipated. Alongthis lineofthought, apersonal type of encryption andDRM that enablecontrol of uploadedcontentwithout limitingthe ability of sharinginformation with acommunity of online contactswas proposed. In themeantime, theapplication of machinereadablelicenses that protect data may also emerge (in thesameway,for example, that theCreativeCommonslicensingsuite is designed to protect copyrightedcontent). Nevertheless, multiple challengeswereforeseen. Enforcement represents onestumblingblock (how to obligethe end-user to respect designated restrictions?), while achievingthe necessary levelofregularity also seemsdifficult(how can oneimagineevery context in which an image mightbeused in ordertoencodethat into aprotectiveinformation system?). In the long run of course,ifthe idea is that privacy is becomingincreasingly contextualised, then patterns mighteventually be discerned that allow forsomedegree of automation.

On adifferentnote,excessiveprivacy protection shouldalsobeavoided andthe free flow of informationrespected, given that onlinecontentconstitutes part of our public sphere. As amatter of fact, theviewwas also posited, duringthe discussion, that no type of privacy-protectingdesign is warranted: search engines may enableinsightinto another’sprivate life,but, at thesametime, users shouldbeaware that information posted onlineisposted into apublic realm. Detailed regulation might be an excessivereaction with only minimal gainsfor actual privacy rights, whereasimplebehavioural adjustmentcouldachieve more effectiveresults.

Finally,itisworth notingthat theaforementioned concept of a“database of intentions”10 also caused disquiet aroundissues of privacy.The reply offered wasthat trade-offsofthis type area necessaryfeatureofthe information society.Atthe endofthe day,abalancingact has to takeplace between thevalue of effectiveinformation provision andthe invasion of privacy that this might involve. In mobilesearch, forexample, at themoment, themost popular queries relate to theweather, other local informationand maps,which obviously havethe capacity to reveal theperson’slocation, yetequally providevaluableinformation on an immediate basis.The suggestion, therefore, wasthat our concept of privacy will in thefutureevolve, in ordertoaccommodate possibilitiesthat technology is only now enabling.

In fact, it seemed apparenttothe majority of participants that theprotection of privacy anddata protection in particular shouldnot focusonablock mode of protection, but rather amoreflexible approach. An over-archinglegal or eventechnological regimethat accommodates all differentkinds of interaction between peopleisdifficult to envision, especially in thefaceofsearch functionswhich ignorethe traditional barriersthat allow forthis differentiated interaction, thus doingawaywith “privacy through obscurity”. 11

2.2.AudiovisualContent Online andFreedom of Expression

In view of theabove,the need foraclarification of search engines’ legal position as concerns freedom of expression provisionsbecomes all themorepressing. Joris vanHoboken’spresentation opened thedebate on these issues.

9) Article29WorkingParty,Opinion on data protection issues related to search engines,WP148, 4April 2008. 10) See above,page 1. 11) See thearticlebyMichael Zimmer in this publication.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 6 SEARCHINGFOR AUDIOVISUAL CONTENT

Firstly, questionswereraised as to thecurrentlegal framework’ssuitability to deal with freedom of expression issues in theonlinecontext. After all, Art.10 ECHR12 has often been accused of being anachronistic; As muchasthe European Court of Human Rights has striven to maintain adynamic and evolutiveapproach, it has notalwaysbeen abletokeep up with technological changes. With regard to thequestion of theprotection of access forinformation providerstosearch engines,itwas pointed out that Art. 10 does notentail ageneral righttoreach an audience. 13 Yetthis approach arguably could exclude someinformation providersfromequitableaccess to someofthe most effectivemeansof communication, i.e., search platforms. After all, access to awillingaudience is,inthe final analysis, apremise forthe exercise of freedom of expression: without it, your voice fallsinto thevoid. As an alternativetostrict legal rulesonaccess,the idea wasput forth that value mightbehad in considering varioussoftlaw sources,suchasCouncil of Europe recommendations.14 Finally,the conjecturethat the duties andresponsibilities with which freedom of expression is coupled in Art. 10 mightprovideabasis forestablishingsearch engineliability,atleast in aco- or self-regulatory manner,was proffered.

Secondly,attention waspaid to therecentCouncil of Europe’srecommendation on Internet filters.15 Thesuggestion, at thebeginningofthe very first guidelinelisted, that “usersmustbeinformed that afilterisactiveand,where appropriate,beabletoidentify andtocontrol theleveloffilteringthe contentthey access is subject to”caused someperturbation. Oneparticipantfelt that such a requirementwouldbeexcessive, particularly in viewofthe widespreaduse of,for instance,spam filtering. It wasasserted that muchofsuchfiltering is undertaken without theuser’sawareness or consent. Nevertheless,for themost part, it is considered abeneficial service.Noreasonableneed could thereforebediscerned foruser briefingonthe operation of such filtering.

Finally,fromthe user sideofthe equation, therighttoaccess information freely gainsadded importanceinanonlineenvironment. In fact, Article19ofthe Universal DeclarationofHuman Rights16 provides forarightto“seek[…] information andideas through anymedia”,an underdeveloped elementthat is often overlooked. Obviously,the user has an interest in freely searchingthe informationthat is available online. What is importanttounderstand, however, is that what Art.19 guarantees,i.e., thefreedom to seek information without hindrance,differssignificantly from an actual righttoaccess information. As oneofthe participants observed, theimportance of this distinction becomes clear when onecontemplates themain elements of theprivacy discussion outlined above:whenasociety has developed such sophisticated search toolssoastoenableloggingofits members’ every move,arighttoactual access can emerge as apowerful andharmful weapon. Tensions between freedom of informationand privacy thus emerge:eachright, in an optimal situation, providingthe necessary counter-balancingmechanismstorein in theexcesses of theother.

3. RegulatoryAspects of Audiovisual Search

3.1. The Place of Audiovisual SearchinCurrentLawsand Regulation

As farasthe regulatory treatmentofsearch andnavigation toolsisconcerned, anumber of proposalswereput forwardastothe appropriate method of approachingthe subject. Peggy Valcke’s presentation elucidated thefragmented treatmentthat is currently in force: ThecurrentEURegulatory Framework forElectronic Communications(ECNS –Electronic CommunicationsNetworksand Services

12)Convention forthe Protection of Human Rights andFundamental Freedoms(European Convention on Human Rights, as amended) (signed 4June1950, entered into force3September 1953) (hereinafter: ECHR) Art 10. 13) See Staffordv.the United Kingdom,Judgment of theEuropean Court of Human Rights (GrandChamber) of 28 May 2002, para. 68; Appleby andOthersv.the United Kingdom,Judgment of theEuropean Court of Human Rights (Fourth Section) of 6May 2003, paras.47and 48; In addition, similar issues arise in respect of broadcastingin, inter alia: VgT Verein gegen Tierfabriken v. Switzerland ,Judgment of theEuropean Court of Human Rights (SecondSection) of 28 June2001; Haider v. Austria,Decision of inadmissibility of theEuropean Commission of Human Rights (First Chamber) of 18 October 1995, Application No. 25060/94. 14) See thearticlebyJoris vanHoboken in this publication. 15) Recommendation CM/Rec(2008)6 of theCommittee of Ministerstomember states on measures to promote therespect for freedom of expression andinformation with regardtoInternet filters, availableat https://wcd.coe.int/ViewDoc.jsp?Ref=CM/Rec(2008)6 (accessed 1July 2008). 16) Universal Declaration of Human Rights (adopted 10 December 1948) UNGARes 217 A(III), Art. 19.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 7

Directives) is inappropriatefor theregulation of audiovisual search engines andsearch engines in general. Art.2(c) of theFramework Directivestates categorically that theterm“electronic communicationsservices”excludesinformation society services.Search engines,therefore,are not covered. Likewise,they cannot be considered associated facilities,which Art. 2(e) Framework Directive currently defines as “facilitiesassociated with an electronic communications networkand/or an electronic communicationsservice which enableand/or support theprovision of services via that network and/or service.”17 EPGs,onthe contrary,are specifically mentioned by theDirectiveas constitutingassociated facilities.Inany case,the ECNS, in termsofits broadcastingaspects, was intendedfor theregulation of vertically integrated TV platforms, whereone operator controlsaseries of facilities.Inthe case of theInternet, on thecontrary,the value chain is structured differently.

Oneplace in thecurrentregulatory framework wasidentified whereroom forsearch engines might exist, at least as concernsconsumer protection. Theproposed text forarevised Art. 20 (5) of the Universal Service Directive 18 as it stands wouldreadasfollows:“Member States shall ensurethat where contracts areconcludedbetween subscribersand undertakings providingelectronic communications services and/or networks, subscribersare clearly informed in advanceofthe conclusion of acontract andregularly thereafter of anylimitationsimposed by theprovider on their ability to accessor distribute lawful contentorrun anylawful applicationsand services of their choice.” Thesuggestion wasthat, with theopportunity of theongoingreform of theECNS, legislativechangescouldbemade to theeffect of broadeningthe notion of associated facilities,asdefined above,soastoinclude search engines.Inaddition, theproposed Art. 20 (5) shouldberewritten to apply also to thesituation where no contractexistsassuch, but an individualismaking de facto useofaservice.The waywouldthen be open to obligesearch engines to provideaddedtransparency concerninge.g., rankingmethods or sponsored links. Reactionstothis suggestion were hesitant, especially in viewofthe factthat thesame effect couldbeachievedthrough search engineself-regulation.

Finally,the extenttowhich thetermsofthe E-Commerce Directive 19 can influence search engines wasalsoinvestigated. As wasremarked, however, although this Directivedoes in generalapply to search engines,infactitoffersvery littleintermsofactual obligations. Articles 12 to 14 of theE- Commerce Directiveprovidesafe harboursfor three types of intermediaries: caching, hostingand mere conduit.20 Art. 21 of theDirectivemakes it clear that these provisionsdonot cover“theliability of providersofhyperlinksand location tool services”.Onthe contrary,the Articledismisses search engines to further examination in abiannual report on theapplication of theDirective, in which particular consideration shall be paid to “theneed forproposalsconcerningthe liabilityofproviders of hyperlinksand location tool services,‘notice andtakedown’ procedures andthe attribution of liability followingthe takingdownofcontent.” 21

At this pointitwas put forwardthat, notwithstandingthe E-Commerce Directive’ssafe harbour for cachinginArt. 13, some of thecachingbysearch engines couldinvolve liability,especiallyconsidering thehabit of search engines of storingcachesofwebpagesfor theexact purpose of offeringthemwhen access to theoriginal page is cut off. Analogies were made at this pointinthe discussion to similar provisionsinthe CopyrightDirective, which containsaprovision on cachingaswell.22 In this context, reference wasmade to the Copiepresse case,currently windingits waythrough theBelgian courts. It waspointed out that theBelgian Court of First Instance23 has held that cachingonthe part of Google amounts to unauthorised reproductionand communication to thepublic andis, therefore, aviolation

17) Directive2002/21/EC of theEuropean Parliamentand of theCouncil of 7March 2002 on acommon regulatory framework forelectronic communicationsnetworksand services (Framework Directive) [2002] OJ L108/33. 18) Directive2002/22/EC of theEuropean Parliamentand of theCouncil of 7March 2002 on universal service andusers’ rights relatingtoelectronic communicationsnetworksand services (Universal Service Directive) [2002] OJ L108/51. 19) Directive2000/31/EC of theEuropean Parliamentand of theCouncil of 8June2000 on certain legal aspectsofinformation society services,inparticular electronic commerce,inthe Internal Market (Directiveonelectronic commerce) [2000] OJ L178/1 (hereafter: E-Commerce Directive). 20) E-Commerce Directive, Arts. 12-14. 21) E-Commerce Directive, Art. 21. 22) Directive2001/29/EC of theEuropean Parliamentand of theCouncil of 22 May 2001 on theharmonisation of certain aspects of copyrightand related rights in theinformation society [2001] OJ L167/10, Art 5. 23) Google Incv.Copiepresse SCRL (RB (Brussels)) Tribunal de PremièreInstance (Brussels) 13 February 2007; [2007] E.C.D.R. 5 2007 WL 1623283.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 8 SEARCHINGFOR AUDIOVISUAL CONTENT

of copyright. Thecourt did findthat, to theextentthat cachingbyasearch engineconsistsofthe “automatic,intermediate andtemporary storage of that information, performed forthe solepurpose of makingmoreefficientthe information’sonwardtransmission to other recipients of theservice upon their request”24,the E-Commerce Directive’ssafe harbour in generaldoes apply.Nevertheless,the ultimate conclusion wasthat, in theparticular case,the issue at stakewas notthe temporary storage of cached pagesaspart of theindexationprocess,but thevisibility of thecachedpagesfor users, which washeldtobeabreachofcopyright.

Of course,the obviousplace foraudiovisual search engineregulation, if necessary,wouldbeaspart of theAudiovisual Media Services (AVMS) Directive’sregulation of audiovisual services. 25 After all, it is this that governscultural andcontentissues,asopposed to theECNS framework that dealswith questionsoftransmission andcompetition. Theproblem here is that theAVMSDirectivecurrentlydoes notcoveraudiovisual search engines, 26 has only very recently,inDecember 2007, been reviewedand is unlikely to be revisitedbythe Commission forsometimetocome. Nevertheless,atheoretical discussion on thetopic ensued. This mainly revolved aroundthe interpretation of theterm “editorial responsibility”.

Editorial responsibility is that requirementfor an audiovisual media service,asdefined by Art. 1of theAVMSDirective, which wasmentioned as providingthe main stumblingblock forincludingsearch engines within itsremit.27 Thesamearticledefines editorial responsibility as “theexercise of effective control both overthe selection of theprogrammes andovertheir organisationeither in achronological schedule[…] or in acatalogue […]”.Duringthe course of thediscussion, it wasestablished that the intention at least of thelaw-maker when draftingthe Directivewas to exclude search engines from thisdefinition. Thepreparatoryworks, it wasfelt, made this clear: what wasaimed at wasthe activities of broadcasters, i.e., theselection of programmes andtheir inclusion in achronological schemeor catalogue.This is distinct from theautomatic generation of alist through theuse of an algorithm. Recital 19 of theAVMSDirectivestates that “thedefinitionofmedia service provider shouldexclude natural or legal personswho merely transmit programmes forwhich theeditorial responsibility lies with thirdparties.” “Carriers”, therefore, includingsearch engines,wereargued to be unavoidably excluded.

Nevertheless,someofthe participants held theviewthat this definition shouldbesufficientto encompass searchengines.Accordingtothis position, theuse of an algorithmisinfacta demonstration of effectivecontrol: after all, it waspointed out, an algorithmthat does nottakeinto accountthe exclusion of incitements to hatred wouldhavetobemodified accordingly.Asopposed to hostingproviders, whoneed nottakeheed of thecontentofthe information stored on their servers to theextentthat it is legal,28 search enginealgorithms do takeaninterest in thenatureand content of theinformationcontained on thewebsites to which they providereferences.Ifnot forany other reason, then this wouldbebecause it is precisely their mission to furnish their userswith information relevant to their query.

Theopposition to this train of thoughtturned on atechnical understandingofthe wayinwhich search engines actually function. Theobservation wasthus made that it is actually thecontentthat controlsthe search engine, rather than theother wayaround. Consequently,asearch engineis differentfromacatalogue or directory behindwhich ahuman editorial team, which definitely does exercise editorial responsibility accordingtoArt. 1AVMSDirective, can be found. Similarly,a distinction shouldbemade with regardtovertical search platforms, as,for example, GoogleNews, seeingasthere theengineoperator makes adecision as to what particular type of contentthe search engineissearching. Accordingly,itwas suggested that acase can be made forthe positionthat platformssuchasYouTube do exercise editorial responsibility.Onceagain, an analogousperception in

24) E-Commerce Directive, Art. 13. 25) Directive2007/65/EC of theEuropean Parliamentand of theCouncil of 11 December 2007 amendingCouncil Directive 89/552/EEC on thecoordination of certainprovisions laid downbylaw,regulation or administrativeactioninMember States concerningthe pursuit of television broadcastingactivities [2007] OJ L332/27. 26) See thearticlebyPeggy Valcke in this publication. 27) See thearticlebyPeggy Valcke in this publication. 28) E-Commerce Directive, Art. 14.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 9

thefieldofcopyrightcouldhavesevere implicationsfor search engineliability.The Viacom case, 29 for example, wasmentioned: if it is accepted that aspecific service provider actually does exert editorial control, then it wouldbeeliminated from safe harbour provisionsinthe DMCA or similar legislation in Europe,makingitresponsiblefor theprevention of possiblecopyrightinfringement by,and the illegal or harmful natureof, thecontenttowhich it links.

Finally,fromthe business perspective,Telenet’sapproach to editorial responsibility wasalso discussed. To begin with, it wasnoted that, due to thebroad Belgian definition of broadcastinginthe law,on-demandservices areindeed included; meaningthat forTelenet editorial responsibility can in anycase notbeavoided. Within this legal framework, Roukensalsoexplained that Telenethas tried to balance theexpectabledesireofacompanytominimize accountability with customer expectations. Research indicates quite clearly that peopleare notfamiliar with on-demandservices.They expect the samekindofservice,therefore,ontheir IDTV platform that they receivefromafilm rental. Telenet has respondedtosuchexpectations, through, e.g., systemsfitted into their set-top boxeslimiting access to filmsrated as unsuitablefor children.

3.2. Self- andCo-regulationand Search Engines

As aresult of thecurrentlackofcentralized andclear regulation on thepart of thestate,search engineprovidersare obliged to set astandardfor themselves.Yet, as WolfgangSchulz asserted in his presentation, this self-regulatory practice is farfromtransparentand of questionableefficacy. Especially from thepointofviewofthe smaller providers, theresultinglegal uncertainty can be difficult to manage.Ingeneral, it couldproduce achillingeffect.30 Theobservation wasmade during theworkshop that search enginesthemselves haveactually expressedadesirefor more precise regulatory stipulationsthat wouldprovidethemwith amorestablefootingfor questionssuchas whether or not, forexample, picturepreviews in search resultscan constitute copyrightinfringement in andofthemselves.

In anycase,tothe extentthat it exists, search engineself-regulation mainly involves notice and takedownprocedures.Itisworth notingthat no put back provision is at themomentineffect, somethingthat couldprove problematic.However, duringthe discussion, thedesirability of this notice andtakedownmethod as adefence against objectionablematerial on theInternet wasbroughtinto question. After all, oneparticipantremarked, if thetargetislimited to themerereference,which is automatically generated usingthirdparty information, whilethe original material remainsonline, what we areengaginginremainsatbest an exercise in futility.Targetingthe original website would surely be more effective. In this context, theproposal of approachingsearch engines as beingsimilar to cachingservices,rather than hostingones, wasput forth.

These reflectionsaside, onequestion raised concerned theGerman Freiwillige Selbstkontrolle Multimedia-Diensteanbieter (FSM) code.31 This involves aself-regulatory codeofconduct, which the major search engines in Germanyhaveadopted since 2004. Someconfusion seemed to exist among participants as to thevoluntary natureofthe scheme. Thesearch engines themselves appear to claim thecontrary in certain contexts, i.e., that theself-regulatory schemeisState-imposed. So, although anumber of thewebsites deemed inappropriate underthe schemeare to agreat extentstill accessible

2 9) TheViacom case ( Viacom International Inc, vs. YouTube,Inc, No. 07 Civ.2103 (S.D.N.Y., 13 March 2007)) is aclosely-watched case currently pendingbeforethe U.S. courts, which dealswith theUSD 1billion lawsuit formassivecopyrightinfringement filed, in March 2007, by themedia conglomerate Viacom (“Video &Audio Communications”) against YouTube andits parent company Google. Viacom claimsthat YouTube hostsnumerousunauthorized copyrighted clipsofits entertainment programminguploadedbyusers. YouTube,onthe other hand, assertsapplicability of thesafe harbour provision of theDigital Millennium CopyrightAct (17 U.S.C. §512(c)(1) (2000)), underwhich ISPs arenot required to monitor their sites for infringingmaterial, but mustremove such material promptly once itsexistence has been broughttotheir attention. Viacom maintainsthat therequirements forthe exercise of thesafe harbour exception arenot met. 30) See thearticlebyJoris vanHoboken in this publication. 31) SubcodeofConduct forSearch EngineProviders(“VK-S”) of theAssociation of Voluntary Self-RegulatingMultimedia Service Providers(Verhaltenssubkodex für Suchmaschinenanbieter, “VK-S”, der Freiwilligen Selbstkontrolle Multimedia-Diensteanbieter) of 21 April 2004, availableinEnglish at: http://www.fsm.de/en/Subcode_of_Conduct_for_Search_Engine_Providers(accessed 1July 2008) andinthe original German at: http://www.fsm.de/de/Subkodex_Suchmaschinenanbieter

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 10 SEARCHINGFOR AUDIOVISUAL CONTENT

via someparticipatingsearch engines,Google, on theother hand, states on itssearch result page that theremovalofsearch resultsisasthe consequenceofalegal request. Theparticipants’conclusion was that theanswerlies in theinterpretation of theword“voluntary”: When providersfilter on aself- voluntary basis,measures such as thefoundingofself-regulation initiatives,the adoption of acode of conduct andthe operation of acomplaints office (as happensinthe case of theFSM) serveto externalise theresponsibility.Inthis manner,compulsion to conform wouldseem to derivefroman external body,which theparties themselves experience as external pressure. That, however, is notto say that an actual legal obligation existsassuch.

3.3. TheActual Need forRegulatory Intervention

In thefinal analysis,however, theconclusion seemed to be that thevery need forregulation of search engines shouldnot be taken forgranted. When faced with WolfgangSchulz’sdetailed list of the riskstriggered by search engines, 32 oneparticipantobjected, claiming that this mightconstitute an overly gloomyapproach. Forexample, theprecise meaningof, anddangers presented by,the fragmentation of thepublic spherewas inquiredinto. Theanswerreferredtostudies that have indicatedthat theuse of so-called “research media” shortensthe agendaofissues upon which an individual gathersinformation. Traditional media, on theother hand, bringreaders facetofacewith subject matter that they wouldnot necessarily havepre-selected themselves,but which might, in effect, be of interest to them or of relevancetotheir needs. 33 Thecounter-argumentpresented was that search media couldbeviewedasfacilitating thecreation of communities of common interests, therebyincreasingsocial cohesion.

What wasagreed upon by all participants wasthat theway in which peopleapproach information is changinginaradical manner.JaneBuckingham, founderofthe Intelligence Group, has reporteda collegestudent’s conviction that, “if thenews is that important, it will findme.” 34 Someparticipants thereforefelt thatthe problem with regulatingsearch engines couldbethat, at themoment, we reside only in an interim phase: theconsequences andeffectsnew media will haveand thepitfallsthey may containare as yetunclear.Inasimilar vein, theideawas also introduced that theuse of search engines couldactually help reduce theoverall need forregulation. Forexample, in thefieldofconsumer protection, search engines can help strengthen theposition of usersthrough easyaccess to knowledge about theproductspresented to them. Likewise,whenitcomes to media pluralism, search engines havethe powertoguidetheir usersprecisely to apluralistic supply of services.Transparency is thus increased andthe need forregulatory intervention abated. Such benefits, however, although real, should notblindone to theneed to address theseparate question of thepossiblereasonsfor regulatory intervention presented by search engines themselves.

Accordingly,other participants saw arosy, notagloomy, pictureinthe aforementioned list of risks. Accordingtothis view, most of thedangers it includesare either already coveredbyexisting legislation, by self-regulationorbyself-organisation on behalf of users. So, thedangerofdistortion of competition is aquestion forcompetition law andaccess to harmful contentcan be limited through self-regulatory codes of conduct. Finally,issues such as thefragmentation of thepublic spherecan be counterbalanced through thepotential of theInternet forbringingcommon interest groups together. In thefinal conclusion, it wassuggested that amuchbetter useofregulators’ timeand effort would be,inrelation to both broadcasters andany associated facilities (such as EPGs) that areavailableon agiven platform, to makethemdeliver complete information about their programmes,properly tagged. In this way, theinformation accessed through audiovisual search toolswouldbecorrect andnocall foran ex ante regulation of these toolsthemselves wouldpresentitself.

32)These wouldinclude thefollowing: Access to harmful content, Access to illegal content, Discrimination of content, Misleadingconsumers, Influence on opinionmaking, Fragmentation of thepublic sphere, Exploitingprotected works, Exploitingpersonal data, Distortion of competition, includingtransfer of market powertoother markets(e.g., advertising). See thearticlebyWolfgangSchulz in this publication. 33) KlausSchönbach et al., OnlineNewspapers: ASubstitute forPrintNewspapersand Other Information Channels? 6 th WorldMedia Economics Conference,Centred’étudessur lesmédias andJournal of Media Economics,HEC Montréal, Montréal, Canada, 12- 15 May 2004, availableat http://cf.uba.uva.nl/nl/handl/googlescholar/ (accessed 9July 2008). 34) Brian Stelter,“FindingPolitical News Online, theYoungPass It On”, availableat http://www.nytimes.com/2008/03/27/us/politics/27voters.html?_r=1&adxnnl=1&oref=slogin&adxnnlx=1215003699- 5hhJsU3ewQ+YzGMGjWskOQ (accessed 2July 2008).

©2008, European AudiovisualObservatory, Strasbourg (France)

SEARCHINGFOR AUDIOVISUAL CONTENT 13

Techno-economic Challenges of Audiovisual Search Engines

Ramón Compañó (*) European Commission -JointResearch Centre(JRC) Institute forProspectiveTechnological Studies (IPTS)

Acknowledgements:

This paper is dedicated to my colleague andfriendBorisRotenberg,who passedawayon 23 December 2007 in atragic skiingaccident. Manyofthe ideas discussed here were conceived during debates with him about theprospectsfor search engines.Iwill alwaysremember him as adynamic andbrightprofessional andanenthusiastic person.

This work has been supported by theEuropean Commission through theIST project “Chorus” (Contractnumber 2006-045480).

(*)Disclaimer Theviews expressed in this publication arethe soleresponsibility of theauthor anddonot necessarily reflect theviews of theEuropean Commission. Neither theEuropean Commission, norany person actingonbehalf of theCommission, is responsible forthe usethat mightbemade of thefollowinginformation.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 14 SEARCHINGFOR AUDIOVISUAL CONTENT

Introduction

Audio-visual (AV) search engines promise to becomekey toolsfor theaudio-visual world, as text search did forthe presenttext-based digital environment. AV search applicationswill enableusto reliablyindex,sift through, and“accredit” (or giverelevanceto) all formsofaudiovisual creation (both individualand collaborative). Moreover, AV search will becomecentral to predominantly audiovisual file-sharingapplications. It will also lead to innovativewaysofhandlingdigital information.For instance,pattern recognition technology will enableuserstosearch forcategoriesofimagesorfilm clips.Likewise,AVsearch couldbeused forgatheringall thepast voice-over-IP conversationsinwhich certain keywordswereused.

However, if these kinds of applicationsare to emerge,search technology mustdeveloprapidly in scaleand type.There will be agrowingneed to investigate new AV search techniques built, for instance,arounduser behaviour.Therefore,AVsearch is listed as oneofthe toppriorities of thethree major US-basedsearch engineoperators–Google, Yahoo! andMicrosoft. In Europe,the settingupof theFrenchQuaeroinitiative,1 forthe developmentofatop-notch AV search portal, or theGerman Theseus research programme, 2 providefurther evidence of theimportantpolicydimensionsofAV search.

Thereare avariety of AV search engines fordifferenttypes of applications. They differ by type of audio-visual content(text, audio, video), by technological platform (mobile, PC), or by application (e.g., publicly availableinformation on theweb versus proprietary databases). From thegeneral public’spointofview, thebest knownonesare “web search engines” .The term “web search engine” refers to aserviceavailable on thepublic Internet that helps usersfindand retrieve content or information3 from thepublicly accessibleInternet.4 Well-knownexamples of websearch engines are Google, Yahoo!, Microsoft andAOL’s search engineservices.Web search engines may be distinguished from search engines that retrieve information from sources that arenot publicly accessible. Examples of thelatter include those that only retrieve information from companies’ large internal proprietary databases (e.g., those that look forproductsineBay or Amazon, or search forinformation in Wikipedia), or search engines that retrieve information that cannot be accessed by websearch engines. 5 Similarly,wealsoexclude from thedefinition those search engines that retrieve data from closed peer-to-peer networks or applicationsthat arenot publicly accessibleand do notretrieve information from thepublicly accessibleInternet. They can be forgeneral purpose search, likeGoogle, Yahoo! or MS LiveSearch, or specialized forparticularthemes. Thematic or vertical search portals include those specifically designed forchildren,6 forpeoplesearch,7 those focused on health-related topics, 8 forretrievingsoftwarecodes, 9 forradio search, music search andfor science,etc. Aregularly updated by domain is availableatPandia.10

1) http://www.quaero.org/ (last visited: 6June2008). 2) http://theseus-programm.de/ (last visited: 6June2008). 3) It is more accurate to refer to search resultsas“content” or “information”, rather than webpages, because anumber of search engines retrieve information other than webpages. Examples include search engines formusic files,digital books, softwarecodeand other information goods. 4) Forasimilar definition, see James Grimmelmann, TheStructureofSearch EngineLaw (draft), 13 October 2006, p.3, at http://works.bepress.com/james_grimmelmann/13/ (last visited: 6June2008). It is acknowledgedthat many of the findings of this paper may be applicabletodifferentkinds of search engines. 5) Part of thepublicly accessibleweb cannot be detected by websearch engines,because thesearch engines’ automated programmes that index theweb,crawlersorspiders, cannot access them due to thedynamic natureofthe link, or because theinformation is protected by security measures.Although search enginetechnology is improvingwith time, thenumber of webpagesisincreasingdrastically too, renderingitunlikely that the“invisible” or “deep” webwill disappear in thenear future. As of March 2007, theweb is believedtocontain 15-30 billion pages(as opposed to sites), of which one-fourth to one-fifth areestimated to be accessiblebysearch engines.See andcompare http://www.pandia.com/sew/383-web-size.html (last visited: 6June2008) and http://technology.guardian.co.uk/online/story/0,,547140,00.html (last visited: 6June2008). 6) E. g., http://www.fragfinn.de/ 7) E. g., http://www.spock.com/ 8) E. g., http://www.healia.com/ 9) E. g., http://www.koders.com/ or http://www.krugle.org/ 10) http://www.pandia.com/powersearch/index.html (last visited: 6June2008).

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 15

While websearch engines scan publicly availableinformation, this is notthe case forsearch on private desktops or in closed proprietary multimedia databases.The former is popular with individuals whodownloadthe freely-availabletoolsprovided by major search engineproviderstosearch fordata on their personal computers. Although thereare someprivacy concerns, they arepopular because they arefree. Companies with sensitivedata makeuse of business solutionsthat takecareoftheir specific needs.National libraries areanexample. Another type is search in peer-to-peer networks,which is aimed at findingaudio-visual contentonthe sites of private PCs (for example, search toolsoperating on BitTorrentexchange platforms). Another emergingtrend is mobile search.

In thefollowingsection, somefundamental technological aspectswill be introduced, in orderto understandthe major technological challenges.

1. Technological Challenges

1.1.SearchEngine Basics

Thestate-of-the-art in search enginetechnology depends on thespecific retrievaltool (audio, imagesorvideo) andthe platform (e.g., distributed computer architecture, peer-to-peer network, etc.). Thepurpose of this section is to presenttechnological elements in ordertounderstandimportant technological challenges, rather than to offer an exhaustivedescription.

Gathering Providing content Indexing Querying Retrieving Displaying content

Figure1:Search EngineProcesses

In essence,asearch engineiscomposed of anumber of technical components that fulfil four basic exchangesofinformation:informationgatheringand indexing, user querying, information provision anduser information access.The gatheringofinformation is undertaken by automated softwareagents called robots, spiders, or crawlers.11 Once thecrawler has downloadedapage andstoreditonthe search engine’sown server,asecondprogramme, knownasthe indexer, extracts variousbitsofinformation regardingthe page.Importantfactors include thewords theweb page contains, thelocation of these key words(e. g., title), theweightattributedtospecificwords (e.g., rare wordshaveahigher search value than frequently-used ones), theproximity of words, or thehyperlinksonthe page.Importantly, theindex is notanactual reproductionofthe page or somethingauser wouldwanttoread. Theindex is further analysed andcross-referenced to form theruntimeindex that is used in theinteraction with theuser.

When auser sends aquery,the enginebrowses theindex on theserver.The query algorithmisthe “soul” of thesearch engine. This algorithmundertakes twomajor processes.First, it defines the matchingprocess between theuser’squery andthe contentofthe index.Second, thealgorithmsorts andranksthe varioushits. Therefore, thedegree of relevancetothe user depends on thealgorithm. Theuser obtainscontenteither from thelink to theoriginal source or from thesearch engine’scache, i. e.,the “temporary archive” on thesearch engine’sown server.The “cache”isacopyofthe last time thesearch engine’scrawler visited thepage in question. Thus, it is notnecessarily up to date,but may be useful forthe user,ifthe server or page is temporarily unavailable, or if theuser wishes to findout what thelatest amendments to theweb page were.

Currentsearch engines arepredominantly text-based,evenfor AV content. This meansthat non- textual contentlikeimages, audio andvideo files areindexed, matched andranked accordingtotextual

11) Thereare also non- or semi-automated alternatives on themarket, such as theopen directory project wherebythe webis catalogued by users, or search engines that tap into thewisdom of crowdstodeliver relevant information to their users, such as Wiki Search, thewikipedia search engineinitiative(http://search.wikia.com/wiki/Search_Wikia) (last visited: 6June2008), or http://www.chacha.com/ (last visited: 6June2008). See Wade Roush, NewSearch Tool Uses Human Guides, Technology Review, 2February 2007, at http://www.techreview.com/Infotech/18132 (last visited: 6June2008).

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 16 SEARCHINGFOR AUDIOVISUAL CONTENT

clues such as filenames, tags, text near imagesoraudio files (e.g., captions) andeventhe anchor text of linksthat pointdirectly at AV content. Truveo is an exampleofthis forvideo clips, 12 andSingingFish foraudio content.13

Whiletext-based search is efficientfor text-only files,this technology has importantdrawbacks when it has to retrieve information from contentother than text. Forinstance,imagesthat arevery relevant forthe subject of enquiry will notbelistedbythe search engineifthe fileisnot accompanied by therelevant tags or textual clues.For example, although avideo may contain ared mountain, the search enginewill notretrieve this video when auser insertsthe phrase “red mountain” in his search box. Thesameistrue forany other information that is produced in formatsother than text. In other words, alot of relevant information is systematically left out of thesearch enginerankings, andis inaccessibletothe user.This,inturn, affectsthe production of all sortsofnew information. 14

Thus thereisahugegap in our information retrievalprocess.This gap is growingwith theamount of non-textual informationthat is beingproduced at themoment. Researchersacross theglobe are currently seekingtobridge this gap. Onestrandoftechnological developments revolves around improvingthe production of meta-data that describes theAVcontentintextformat. Asolution could be foundby, forinstance,developing“intelligent” softwarethat automatically tags audio-visual content.15 However, though technology is improving, automatic taggingisstill very inefficientdue to complex algorithms andhigh processingorcomputationalrequirements. Another possibility is to create asystem that tags pictures usingacombination of computervision anduser-inputs.16 However, manual taggingistimeconsumingand thus costly.

1.2. Challenges

Technical challengesare multiple andcomplex, as they concernboth fundamental theoretical principles andimplementation issues.For thesakeofsimplicity,these challengescan be grouped into themodules presented in Figure1,namely: collectingthe maximum amountofinformation(data gathering), understandinguser queries,adjustingtothe user’scontext (user interaction), supplying themost relevant results, supplyingresultswhich help user interaction (returningresults). Another challenge is related to how to manage themultiplicity andinteroperability of devices andplatforms. This will be discussed briefly in thefollowingsection.

1.2.1. GatheringInformation:Dealingwith theData Explosion

Theamountofinformationcreated,storedand replicated in 2000 was3billion gigabytes (Exabyte), 24 Exabyte in 2003,and 161 Exabyte in 2008.17 Forthe sakeofcomparison, thelast figureisequivalent to three million times thetotal information contained in all bookseverwritten. This figureisexpected to reach988 Exabyte by 2010.18 We can fairly say that we arewitnessingadata explosion trend.

This huge amountofdata comes in avariety of formats, andcontenthas evolved farbeyond pure text description. In fact, though data wasstill largely text-based in theearly daysofdigitization, it is believedthat now 93% of thedata on theInternet is in multimedia format andthis will approach 99% in 2010. It is also interestingthat 70% of this data is created or achievedbynon-professional

12)http://www.truveo.com/ (last visited: 6June2008). 13) SingingFish wasacquired by AOLin2003, andhas ceased to exist as aseparate service as of 2007. See http://en.wikipedia.org/wiki/Singingfish (last visited: 6June2008). 14) See Matt Rand, Google Video’sAchilles’ Heel,Forbes.com, 10 March 2006, at http://www.forbes.com/2006/03/10/google-video-search-tveyes-in_mr_bow0313_inl.html (last visited: 6June2008). 15) In this regard, see James Lee, SoftwareLearnstoTag Photos,Technology Review, 9November 2006, at http://www.technologyreview.com/Infotech/17772/. See Chris Sherman, TeachingGoogle to See Images,Search Engine Land, 5April 2007, at http://searchengineland.com/070405-172235.php (last visited: 6June2008). 16) See Michael Arrington, Polar Rose: Europe’sEntrant Into Facial Recognition,Techcrunch, 19 December 2006, at http://www.techcrunch.com/2006/12/19/polar-rose-europes-entrant-into-facial-recognition (last visited: 6June2008). 17) See andcomparehttp://www.pandia.com/sew/383-web-size.html (last visited: 6June2008) and http://technology.guardian.co.uk/online/story/0,,547140,00.html (last visited: 6June2008). 18) See Andy McCue, Businesses face data ‘explosion’,ZDNet, 23 May 2007, at http://news.zdnet.co.uk/itmanagement/0,1000000308,39287196,00.htm (last visited: 6June2008), referringtoIDC/EMC Study TheexpandingDigital Universe.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 17

users.19 This underlines theviewthat in thecomingyears, userswill gradually shift from beingmere consumerstobecomingproducers, providersand mediatorsofcontent. It is fair to assumethat search engines,inordertocope with thegrowingamountofaudiovisual content, will increasingly become AV search engines.Futureaudiovisual search engines will havetodeal with twomajor challenges. First, they will need to be scalabletodeal with more andmoredata. Theamountofdata is due notonly to theaforementioned creation of original data to store, but also, andtoanevengreater extent, to the huge adjacentprocessingand meta-data necessary to makeaneffectiveretrievalofAVcontent possible. In fact, theindexingofaudio, imagesand video requires extensiveresources in termsof computerprocessingpowerand storage capacity.Second, they will need to be abletoprocess different kinds of audiovisual material. This will possibly requireless deterministic search concepts, e. g., movingfromexact to fuzzy matching. Theneed foranincreasingscalability andless determined contentwill lead to new computer architectures andsearch concepts, which aresketchedinthe followingfigureand will be briefly discussed in thefollowingsection.

D e g r e e

s c a l a b i l i t y ze d peer o- rgan i -t o parallel f- l distributed centralized e peer s

d e t e r m i n i s m T e c h n o l o g i c a l M a t u r i t y w e l l - c u t t i n g - b a s i c e s t a b l i s h e d e d g e r e s e a r c h Figure2

Themultimedia contentproduced daily by Internet userswill exceed theprocessingpoweroftoday’s searchengines to index andsearch AV contentwithin associated text andmeta-data. This has implications forcomputerarchitecture(from centralised to future self-organised clusters) andthe way searches aremade –fromvery to less deterministic searches,e.g., from exacttofuzzy matching,or from precise to approximate searches. 20

In early times,contentwas stored in centralised servers. As long as theamountofinformation was reasonably large andcontentmostly text-based, search wasreasonably good. As contentsize grew, centralized computer systemscouldnot respondtouserswithin an acceptabletime. Centralised systemshad to be replaced, first by parallel computer architectures andlater by distributed systems. In fact, manyscientistsbelieve that state-of-the-art distributed processingwill notbesufficientin thefuturetocope with theincreasingamountofdata. They think that futureweb search engines will be peer-to-peer (P2P) based, andinthe long-term –given thecomplexity of operatingsuchP2P computer architectures efficiently –adegree of self-organization will be adopted to deal with complexity autonomously.

Today’sstate-of-the-art computer architectures forsearch engines aredistributed systems. P2P systemsare in an advanced phase of development, research prototypes haveshowntheir potential21 andsomecompanies,likeFaroo,22 arealready offeringbeta versionsofP2P websearch engines with distributedindex,crawler andranking. P2P retrievalcouldbecomeascalablealternativetoclustered

19) FautoRabitti, Sapir presentation at Concertation meetingVilamoura(Faro,Portugal) 16 April 2008. 20) http://www.sapir.eu/papers/sapir-poster.pdf (last visited 6June2008). 21) http://www.sapir.eu/ (last visited 6June2008). 22) http://www.faroo.com/ (last visited 6June2008).

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 18 SEARCHINGFOR AUDIOVISUAL CONTENT

retrievalengines if state-of-the-art P2P search techniquescan be showntooperate reliably in alarge number of peer machines,i.e., if they aresupported by alarge community.Itisexpected that forfull- text retrieval, P2P approaches will contribute to documentpartitioning(unstructured overlay network forsearch, e. g., Gnutella) andterm partitioning(structured overlay network forsearch e. g., Chord, P-Grid), whilefor audiovisual search, P2P will mostly contribute to theindexingtask andalsoto similarity research. It is unclear when -orindeed if -P2P architectureswill be readytoreplace distributedsystems. Some(isolated) solutionsfor P2P similarity search forsingleAVfeatures (e.g., colour or shape) arealready available, but systemscombiningtextand multiplefeatures (e.g., colour AND shape) arenot yetefficient.

From an operational pointofview, it is expected that futuresearch engines will be abletoperform better when accessingthe deep weband retrieve anysort of audiovisual content. Theincreased amount of data poses challengeswith respect to data storage andprocessing, particularly forindexingand rapid response to user queries.Ithas to be noted that multimedia formats(images, video, audio, 3D, etc.)makethe indexingtask farmorecomplex than text–based formatsdo. Until new architectures areinplace andsolutionsfor non-indexedcontentare available, it is possiblethat therewill be a wideningAVinformation retrieval“gap”.Itislikely that search engines will rely foraconsiderable period on twocomplementary AV search approaches,namely annotation-based technologies,where indicesare generatedautomaticallyormanually,(e. g., tagging) andcontent-basedtechnologies,e.g., similaritysearch. To help thesearch processatalater stage,indexingrightfromthe sourcelevelwould be beneficial. Forinstance,commercial devices wouldhavearange of data already built in, such as authorship, productionprocess,geographical data, etc. In addition, therewouldbethe option foruser- suppliedindexing, introducingtagging, comments, etc. Such AV-related meta-data wouldthenhelp thefindingofrelevant content.

1.2.2. UnderstandingUser Queries: Bridgingthe Semantic Gap

Typingkeywordsisnot theideal waytointeractinasearch forinformation. Gettingcloser to akind of natural speakinglanguage wouldhelp overcome“keywordese”; but this solution is alongway off. While“simplephrase” search is already possibletoday, slightly more complicatedsentences are difficult formachines to handleascontextual search is notpossible. Forinstance,ifweintroduce the term “Jaguar” thesearch engineisunabletodistinguish whether we mean theautomobileorthe animal, as processingthe user’squery is linked to themachine’sability to understandthe semantics of thequery.Asmentioned before, progress in machinelearningand artificial intelligence has been steady,but thesemantic webasareality is still alongway off.

Retrievalperformance decreases with the“complexity” of content, as will be illustrated in thecase of images. From aretrievalpointofview, thesimplest imagesare those with clear boundaries,colours, geometry (e.g., no deformation), shape,etc. Examples are“artificial” images, likelogos or symbols. They arerelatively easytoretrieve by search engines because ”similarity” approaches work very well. Therefore, search engines areused in applicationstodetect trademark infringements, as trademarks areoften imageswith low-levelfeatures.

Thenextgroup of imagesare those with specificsimilarity measures.They aremorecomplex than thepreviously mentioned “artificial”images, but they haveparticular features that rendertheir retrievalfast. Oneexampleisfrontal pictures of faces, 23 because their individualfeatures (eyes, mouth andnose) arewell defined, as arethe spatial relationships between thefeatures. Another exampleis fingerprints.Their colour contrast is pronounced andthe matchingshapesimilarity is done by comparingthemwith fingerprints already stored as pictures in adatabase.

23) Softwareisabletodetect faces based on thehigh contrast areas such as theeyes, nose andmouth. In optimal situations- near frontal faceportraitsinhigh-quality photographs-the systemsperform at over95% accuracy with minimal false positives.For side(non-frontal) views, low-quality pictures or olderimagesfromcultural heritage collectionsthe performance is still farabove average.For an overview, see: Yang,M.H., Kriegman, D.J., andAhuja. N. 2002. DetectingFaces in Images: ASurvey.IEEE TransactionsonPattern Analysis andMachineIntelligence 24(1), 34-58.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 19

Figure3:“Semi-artificial” pictures

Semi-artificial pictures aremorecomplex from aretrievalpointofview. These areimagestaken underwell-defined conditions, such as onesingleobject per pictureagainst ahomogeneous background. Thesetsoffish andbraceletsinFigure3are examples that fulfil these conditions. Searchingfor this kindofpictureisuseful foreducational purposes,particularly forencyclopaedias, andalsofor commercial purposes (i. e.,searchingfor particular products).24 Themost difficult pictures to retrieve arereal images, likethose in photoalbumsoronthe Internet. They contain manydifferent types of objects, (e.g., landscapes, people, art, artificial objects, etc.), renderingthem“unpredictable” andcomplex and, thus, difficult to analyse andprocess.

Technologically speaking, machines can retrieve low-levelfeatureimages, likeartificial images, fairly well by analysingbasic elements likethe object’scolour,textureand shape.Therefore,similarity search engines haveprovenuseful in specific contextslikesearchingtrademark databases, 25 detecting copyrightinfringement, 26, 27 findingvideo shotswith similar visual contentand searchingfor music with similar rhythms. In all these applications, thebasic features such as colour andtextureinimages andvideo, or dominantrhythm, melody or frequency spectruminaudio, aretightly correlatedtothe search goalsofthe particular application. Forinstance,inSongTapper, 28 auser can search forasong by typingthe rhythmofthe songonthe space bar of thekeyboardor, in Midomi,29 by hummingthe songinto amicrophone.

High-levelfeatures arefar more complex,asan“understanding” of thecontentisoften required. Forinstance,humanscan immediately identify atwo-dimensional projection of athree-dimensional object, i. e.,wecan easily recognize adifferentpictureasjustanother projection (from adifferent angle) of thevery sameobject. Machines,however, areunabletorecognise the2Drepresentation of 3D objects. They lack theunderstandingofthe image contentand thereforeofthe wayahuman would describe an object. Thedifference between twodescriptionsofanobject by differentrepresentations is called thesemantic gap. Closingthe semantic gap is oneofthe biggest challenges in computer science.

This is notjustanacademic challenge,asbridgingthe semantic gap wouldhavepractical consequences which wouldallow more user-friendly interfaces.These wouldbeabletotranslate contextual knowledge in natural language (high-level) into an elementary andreproducibleoperation of acomputing machine(low-level). Designingsuchuser-friendly search engines couldbringthe vast amounts of multimedia knowledge from libraries,databases andcollectionscloser to citizens. These systemswouldneed to understandthe semantics of aquery,and notjustthe underlyinglow-level computational features alone. ImageScape30 wasanearly content-based retrievalsystem addressing semantic gap issues in thequery interface, indexingand results. This early search enginealloweda

24) E.g., http://www.riya.com/ (last visited 6June2008). 25) Eakins, J.P., Riley,K.J., andEdwards,J.D. 2003. Shape FeatureMatchingfor Trademark . CIVR, 28-38. 26) Jaimes,Aand Chang, S-F.2002 Duplicate Detection in Consumer Photographyand News Video, ACMInt.Conf. on Multimedia, 423-424. 27) Joly,A., Buisson, O.,and Frelicot, C. Robust content-based copy detection in large reference database,Int.Conf. on Image andVideo Retrieval, 2003. 28) http://www.songtapper.com/ (last visited 6June2008). 29) http://www.midomi.com/ (last visited 6June2008). 30) Lew, M.S. 2000. Next Generation WebSearches forVisual Content. IEEE Computer,November,46-53.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 20 SEARCHINGFOR AUDIOVISUAL CONTENT

user to query multiplevisual objectssuchassky,trees,water,etc.,usingspatially positioned iconsin an index containingmorethan 10 million imagesand videos.Automatically detectingall semantic contentwithin an image such as aface, tree,animal, etc.,with emphasis on thepresence of complex backgrounds,isstill ahugechallenge.Researcherstryingtoclassify wholeimagesoften getstuck on technological challengesand on practicalities,suchasthe picture’sgranularity beingtoo coarse to be useful in real worldapplications.

Progress in research is steady,although thesemantic gap problem is farfrombeingsolved. Current approaches towardslearningsemantics aremultiple, andinclude conceptslikeexamininghidden associationsduringimage indexing, developingvisual dictionaries which group similar coloursand textures,orintroducinglearningapproaches,and manyothers.

1.2.3. ReturningMoreRelevantResults: Gettingthe User into theLoop

Thelevelofrelevancy depends on thequality of thesearch algorithm. Onedifficulty searchengine developershavetofaceispollution andnegativeexternalities -for example, Spamdexingorlink- bombingare well-knownexamples of unauthorized search engineoptimization toolswhich search engineprovidershavealready hadtofaceupto.

Oneideaabout how to increase thequality of similarity-based search systemsistoenablethemto learn through use, by integratingcontinuous feedback from theuser query.This interactiveprocess is called “relevance”, “query refinement”, or “interactivesearch”.Here, alist of candidate imagesis presented to auser whoisasked to rank therelevanceofthese imagestohis search. Oneofthe major problemsinrelevancefeedback is how to address small training sets, as atypical user may want to label farfewerimagesthan thealgorithmneeds.One option forbetter user feedback is to offer an adequate representation of theinformationretrieved. Somesearch engines arestartingtointroduce visual interfaces in thestructureofthe information, andthe link to other pieces of information is visualized. Clusty31 is an exampleofasearch enginethat clustersresults, andKartOO32 offersavisual interfaceofthe linksand relationships of theretrievedresults. Another wayofincreasingrelevanceis to makeuse of social bookmarking, likeMister-Wong,33 or gettinguser’sfeedback into search process likethe engineChaCha.34

Search services wouldbecomemoreefficientand relevant if likely queries couldbeanticipated,at least to someextent. This wouldrequirethe personalisation andcustomisation of thesearch engine to theuser.Inextremecases,high-levels of personalization wouldmean that systemswouldhaveto storeequally large amounts of personal informationonindividuals. Consequently,the aggregate of all theuser’ssearches wouldget us closer to what John Battellecallsthe “Database of Intentions”. 35

Themotivation forpersonalisation andcustomization depends on theactor involved. Theuser expectsareduction of irrelevant results, andhopes to receivefew but relevant linkstoinformation. From thesearch engineprovider’spointofview, personalization permitsmoretargeted advertising, permittingadvertisementsubscriberstosell “personalised eyeballs”.Itwouldincrease theclick- throughrate andpossiblyreduce theriskofclick-fraud.Fromatechnological pointofview, personalization represents atrade-off between processing performance andspeed of response.Onthe onehand, more user-feed generated information contributes to narrowingthe search domain, andon theother,moreinformation has to be processed. In viewofthe factthat audiovisual search andpattern recognition is highly process intensive, beingabletointroduce aquery in natural language may slow downthe response.

Customization may be at thelevelofindividual search, likeA9, Jeeves, Google, followingcollective search engines,likeEurekster,orbydomains(search history,tagging, notes anddiary sharing, etc.

31) http://www.clusty.com/ (last visited 6June2008). 32) http://www.kartoo.com/ (last visited 5June2008). 33) http://www.mister-wong.com/ (last visited 5June2008). 34) http://search.chacha.com/ (last visited 9June2008). 35) Battelle, John (Sept. 2005). TheSearch: HowGoogle andIts Rivals Rewrote theRules of Business andTransformed Our Culture. NewYork, ISBN 1-59184-088-0.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 21

Practically all major search engines useanumber of approaches to optimize search. Aprominentway is to usecookies with unique IDs linked to thebrowser.Itpermitsthe server to keep records on the logofkeywords, IP address,language,timeand date,type of browser,URL of page requested, redirection, etc. Other relevant data couldbedrawn from varioussources,namely desktop search, toolbar or location-based search. Onerisk is that this information may be linked andcrossed with data originating from other applications andsources -for instance,combiningthe user’straces in with theinformation extracted from his Gmail account.

1.2.4. TowardsContent-based Retrieval

AV search often refersspecifically to techniques better knownascontent-based retrieval. These search engines retrieve audio-visual contentrelyingmainly on pattern or speech recognition technology to findsimilar patternsacross differentpictures or audio files. 36 Pattern or speech recognitiontechniquesmakeitpossibletoconsider thecharacteristicsofthe image itself(forexample, itsshape andcolour), or of theaudio content. In thefuture, such search engines wouldbeableto retrieve andrecognise thephrase “red mountain” in asong, or determinewhether apictureorvideo filecontainsa“redmountain”, despitethe factthat no textual tagattached to thefiles indicatesthis. 37 Content-based methods arethus indispensablewhentextannotationsare non-existentorincomplete. They couldalsoimprove retrievalaccuracy evenwhentextannotationsare present. Theresearch challenge is how to improve multimedia retrievalusingcontent-based methods.Herethe performance of theretrievalprocess andthe quality of theresultsdependonthe type of content, as Lewetal indicate in an overviewpaper. 38

Manyconcepts(i.e., algorithms)and systemshavebeen developed forautomatically structuring, indexingand finding speech andaudio content .Possibly oneofthe first applications forspoken documentretrievalwas the“THISL broadcast news retrieval” system in the90s. 39 Here,speech was transformed into text with thehelp of alarge vocabulary continuous speech database.For each word of theautomatically transcribed word sequences,atimecodewas attached. This allowedasearch based on theusage of astandardtextretrievalmechanism. Since then, progress has been steady as the benchmarkingexercises of theUSNational Institute of Science andTechnology (NIST) show.NIST carries out periodic benchmark exercisestogaugethe quality of spoken documentretrieval evaluations. Comparingthe performance of indexingfor broadcast news duringthe subsequenttext retrievalconferences (TREC, 1997–2000) it becameclear that it has becomeincreasingly challenging to improve significantly thesystem quality.One reason is that theretrievalofspeech recordings is very complex:The overall performance depends on aconsiderablenumber of factors, includingthe number of speakers, backgroundnoises,type of speech (broadcast or colloquial), etc.,and each of them is a challenge in itself.

Equally difficultisthe case of music retrieval. Oneofthe first music indexingand retrievalsystems wasbased on low-levelfeatures of audio processing. 40 Theseweremainly spill-overs invented forMPEG7 Audio standards andAudio-IDsystemsdeveloped by severalgroups.The Audio-ID technology generates afingerprintofasegmentofmusic andprovides fast matchingalgorithms to findthis fingerprintin alarge pre-processed archive. Recently,the focusofmusic retrievalhas been changedtogenreand mood classification.

36)Pattern or speech recognition technology may also provideacogentway to identify content, andprevent thepostingof copyrighted content. See Anick Jesdanun, Myspace Launches Pilot To Filter CopyrightVideo Clips,UsingSystem From Audible Magic,Associated Press Newswires,12February 2007. 37) See Dr.Fuhui Long, Dr.HongjiangZhangand Prof.David Dagan Feng,FundamentalsofContent-Based Image Retrieval, at http://research.microsoft.com/asia/dload_files/group/mcomputing/2003P/ch01_Long_v40-proof.pdf (last visited: 6June 2008). 38) Lew, M. S., Sebe,N., Djeraba, C., andJain, R. 2006. Content-based multimedia information retrieval: State of theart and challenges.ACM Trans. Multimedia Comput. Commun. Appl. 2, 1(Feb. 2006), 1-19. DOI= http://doi.acm.org/10.1145/1126004.1126005 39) D. Abberley,D.Kirby, S. Renalsand T. Robinson, TheTHISL broadcast news retrieval system,Proc. of ESCA ETRW Workshop on AccessingInformation in Spoken Audio, Cambridge (UK), April 1999. 40) Foote,J.1999. An Overview of Audio Information Retrieval. ACMMultimedia Systems7(1), 42-51.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 22 SEARCHINGFOR AUDIOVISUAL CONTENT

For video retrieval,the research focusinthe mid 90s wastofindaway to detect object boundaries, andthe distance between colour histogramscorrespondingtotwo consecutiveframesinavideo.41 Since then, progress has been made in automatizingthe process of boundary detection in an objective way, andaway has been foundofusingthe motion within thevideo to determinethe shot boundary locations, andperformingsemantic classification of thevideo shotsinto categories such as zoom-in, zoom-out, pan, etc.42

Thesearch enginesector is currently thriving, andexamples of beta versionsacross those various strands areavailable, both forvisual andaudio information. Tiltomo 43 andRiya 44 providestate-of-the- art content-based image retrievaltoolsthat retrieve matches from their indexesbased on thecolours andshapes of thequery picture. Pixsy 45 collectsvisual contentfromthousands of providersacross the weband makes these pictures andvideos searchableonthe basis of their visual characteristics.Using sophisticated speech recognition technology to create aspoken word index,TVEyes46 and Audioclipping 47 allow userstosearch radio, podcasts, andTVprogrammes by keyword. 48 Blinkx49 and Podzinger 50 usevisual analysis andspeech recognition to better index rich media contentinaudio, as well as video, format. However, themost likely scenario forthe near futureisaconvergenceand combination of text-based search andsearch technology that also indexesaudio andvisual information. 51 Forinstance,Pixlogic52 offersthe ability to search notonly themeta-data of agiven image,but also portionsofanimage that may be used as asearch query.

Thefollowingtablesummarizes somecurrenttechnological approaches andtheir challengesin multi-media analysis together with currentand futurepromisingapplications. It is split into different media formats, namely text, audio andvideo analysis.The audio domain is divided into speech and music,given thedifferentapplicationrangesand therelated challenges. Someimportantresearch challengescut across differentdomains, such as thepreviously discussed semantics andthe information overflow.

41) Flickner,M.Sawhney,H.Niblack, W. Ashley,J.Qian HuangDom, B. Gorkani, M. Hafner,J.Lee,D.Petkovic,D.Steele, D. Yanker,P.1995. Query by image andvideo content: theQBIC system, IEEE Computer,September,23-32. 42) Lienhart, R. 2001. ReliableTransition Detection in Videos: ASurvey andPractitioner’sGuide. International Journal of Image andGraphics 1(3), 469-486. 43) http://www.tiltomo.com/ (last visited: 6June2008). 44) http://www.riya.com/ (last visited: 6June2008). 45) http://www.pixsy.com/ (last visited: 6June2008). 46) http://www.tveyes.com/ (last visited: 6June2008); TVEyes powers aservice called Podscope (http://www.podscope.com/) (last visited: 6June2008) that allows userstosearch thecontentofpodcastsposted on theWeb. 47) http://www.audioclipping.de/ (last visited: 6June2008). 48) See Gary Price, SearchingTelevision News,SearchEngineWatch, 6February 2006, at http://searchenginewatch.com/showPage.html?page=3582981 (last visited: 6June2008). 49) http://www.blinkx.com/ (last visited: 6June2008). 50) http://www.podzinger.com/ (last visited: 6June2008). 51) See Brendan Borrell, Video SearchingbySightand Script,Technology Review, 11 October 2006, at http://www.technologyreview.com/read_article.aspx?ch=specialsections&sc=personal&id=17604 (last visited: 6June2008). 52) http://www.pixlogic.com/ (last visited: 6June2008).

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 23

Table1:Overviewofthe approaches,applications andchallengesfor content-based retrieval technologies by type of format.

Text Speech Music Images Video Technological Approach -Named entity -Speech Music -Low-levelimage -Shotdetection, recognition, recognition: segmentation: processing keyframe SVM, PLSI spoken document spectral flatness, (histograms, generation Bayesian retrieval, genetic shapes,textures, -Object tracking semantic subword indexing algorithms, MPEG7-visual, based on motion- reasoning -Speech music retrieval SIFT), image based features, -Caption segmentation: and similarity closed captions augmentation speaker recommendation measurements recognition,etc. clusteringand -Relevance -Object recognition feedback detection and speech-to-video recognition transcoding -Video annotation and summarization -Video event detection Applications -Classification of -Indexingof -Indexingof -Content-based -Indexing news and broadcast music collections retrievalin broadcast documents in news/archives -Query by image collections material,media companies -Podcast/ humming -Object observation, -Email filtering videocast search -Audio-music recognition,face indexingof -Text-based (Potzinger, identification recognition videocast search engines, Blinkx) audio recommendation (security,photo material, semantic analysis archives engines collections), recommendation of multimedia (Parliamentdata, automatic engines,video (automatic) historical annotation of fingerprinting, annotations archives) image collections logo detection, with keywords security,etc. andtextual -3Dvideo descriptions Challenges -Semantics, -Variability of -Genre -Semantic gap, -Detection of ontologies design content(e.g., classification image complex concepts, (e.g., background -Polyphonic segmentation, thousands of enrichment) noise) domain instrument sensory gap differentobjects. andlanguage recognition, Segmentation dependency, -Affective into more scalability of analysis semantic-based subwords units(i.e. approaches complex scenes), multimodality, fusion

Source: Adapted from Sebe/Koehler53

53) N. Sebe,J.Koehler,“State of theart in audio-visual contentindexingand retrievaltechnologies” in ChorusReport 2.1 “State of theArt on Engines”, p.29 et seq.,availableathttp://www.ist-chorus.org/documents/D21_final.pdf (last visited 10 June2008).

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 24 SEARCHINGFOR AUDIOVISUAL CONTENT

2. EconomicConsiderations

Thedata explosion in thedigital realm has renderedsearch acriticalfunctionality.Data abundance is at theorigin of search enginedevelopments, but thefactthat search is context dependent makes a difference in applications andtheir markets. Thewell-knownweb search engines offer aservice to userstofindpublicly-availablecontentonthe Internet. Thebusiness modelispredominantly supported by advertising.This is very differentfromsearch engines forbusiness solutions, forinstance. Here,searchismostly carried out in closed, proprietary,multimedia databases (e.g., national libraries). System providersdeliver aretrievaltool specifically tailored to theneeds of thecustomercompany, adapted to their security restrictions, access rights andcontentsecurities,etc.

Although these retrievaltoolsdosharesimilar technologicalchallenges, thenatureoftheir applicationsdiffersand so do their markets. Therefore, thefour most prominentdomains, namely web search, search forbusiness solutions, peer-to-peer search andmobilesearch, will be discussed in the followingparagraphs.

2.1. WebSearch

Although close to onehundred searchengines areoperational, thebulkofthe searches are performed by only afew service providers. Accordingtothe consultancy firm Nielsen/Netratings, the topthree operatorscontrol more than 80% of themarket. In particular,onlinesearches by engine performed in theUSinAugust 2007 were executed by Google53.6%, Yahoo! 19.9%, MSN12.9%, AOL 5.6%, Ask1.7% andthe rest 6.3%. Thesesearches include local searches,image searches,news searches,shoppingsearches andother types of vertical search activity.Morethan 5.6 billion searches were carriedout in that monthalone (August2007).54 Therankingofthe topthree players is undisputed.AccordingtocomScoreNetworks, in December 2006 Googlesites captured 47.3% of the US search market, Yahoo! 28.5% andMicrosoft 10.5%. Americansconducted 6.7 billion searches in December 2006. With respect to thesamemonth last year,this represents an annual increase in search query volumeof30%. This rate is considerableand explainswhy onlineadvertisingonsearch engines is expected to be apromising growth market.

European Internet usersmakeasmuchuse of search engines as their counterpartsonthe other side of theAtlantic.The intensiveuse of search engines explainswhy they areamongstthe most visited pagesonthe Internet andattractalot of traffic.Googleisthe most visitedsearch engineinpractically all countries of theEuropeanUnion. Forinstance in June2007, Googlereached 88.8%ofthe UK, 69.5% of theFrenchand 69% of theGerman onlinepopulation. Google’saudience is notably largerthan for theMicrosoft sites (83.3% UK, 62.3% France,54% Germany) andYahoo! (65.9%UK, 39.6% France and 36% Germany), accordingtothe Internet audience measuringcompanycomScore.55 Thesearch engine market consolidationbecomes evidentwhenobservingthe of hitsoverlongertimeperiods.

Figure4and Figure5show theevolution of thesharefor Germanyand France,respectively.The evolution of Germanyand France is similar to other European Member States.Inparticular,less than ahandful of search engineprovidershaveamarket shareofover90% andGooglehas amuchbigger sharethan itscompetitors.

54) See http://www.nielsen-netratings.com/ 55) comScorePress releases,availableathttp://www.comscore.com/

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 25

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Jul-01 Jul-02 Jul-03 Jul-04 Jul-05 Jul-06 Jul-07 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Jan-02

Google Yahoo! MSN T-Online AOL Other

Figure4Evolution of WebHitsfor search engines in Germanyinthe period 2001-2007. Source: WebBarometer, 56 HSpeck57 andown calculations.

100%

80%

60%

40%

20%

0%

Oct-01Apr-02Oct-02Apr-03Oct-03Apr-04Oct-04Apr-05Oct-05Apr-06Oct-06Apr-07

Google Yahoo! Voila Live AOL Others Figure5:Evolution of WebHitsfor search engines in France in theperiod October 2001-September 2007. Source: BaromètreSecrets2Moteurs 58 andown calculations.

These data highlightthe factthat thesearch enginemarket is highly concentrated andalsothat searchengines havebecomepart of our lives. Theaverage German,for instance,uses Googlemorethan forty times amonth59 andthree-quarters of Internetusers access Internet offersthrough search engines. 60 Although theaudience forthe search engineprovidersmay vary from onecountryto another,the user experience is similar formost western countries.

Consultancy firmscalculatingthe market share, such as Nielson/NetRatings, Compete,Hitwise or comScore, measurethe search behaviour of internet usersbyinstallingreal-timemetersontheir computers(Nielsen states 500,000 peopleworldwide). Themarket shareretrievedbythese consultancy firmsmay differ to acertain extentfor each of thesearch engines,due to thefactthat they employ differentmetrics formeasurementand theaccuracy of thedata is notsufficiently clear.This may partly explain whycomScore’straffic data forGermanyand France differsfromthe hit counts recordedby theGerman WebHitsand theFrenchSecrets2Moteurs.fr.

Theconcentration of theweb search enginemarket appearstobeageneral trendinthe USAand most EU Member States.There may be anumber of reasonswhy GoogleismoredominantinEurope

56)http://webhits.de/ 57) Hendrik Speck. Search.Market.10th InetBib Conference.9-11 April 2008, Würzburg(Germany) availableat http://www.egs.edu/faculty/speck/files/presentation2008inetbibsearchenginemarkets.pdf (last visited 5June2008). 58) http://www.secrets2moteurs.fr/ 59) comScoreGerman data June2007. 60) InternetverbreitunginDeutschland: Potenzial vorerstausgeschöpft? Birgit vanEimeren, Heinz Gerhards andBeaste Frees, Media Perspektiven, Vol8,pages350 –370.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 26 SEARCHINGFOR AUDIOVISUAL CONTENT

thaninthe USA, includingstrongerbranding, national marketingstrategies,better adaptationto market size,better technological adaptation to language,lackofpowerful national search engines,etc. An interestingcase is Russia, eventhough it is notpart of theEuropean Union. Here,Googleisonly thirdbymarket shareafter Yandex61 andRambler.Yandex claimstohavesuperior technology as it mastersbetter thedeclensionsand conjugationsofthe Russian language than other search engines. Other Slavic search engines,likethe Czech Morfeo62 or thePolish NetSprint,63 also claim to havebetter technology.How muchthe Yandex high market shareofover55% in Russia can be attributed to better linguistic performance is,however, notobvious, as thesamesearch engineprovider achievesonly 16% in theUkraine, although theRussian andUkrainian languagesare,linguistically speaking, very close.

Onefactorthat has favoured Google’sdominantposition today is therate at which innovative services havebeen introduced. Manyofthese havebeen proposed to theaudience at development phase (beta versions), rather than offeringfinished services to theusers. This user involvementinthe developmentstage is part of thecompany’scultureoflearning-by-doing. Thecompanyhas benefited from usingthe dominantInternet language,English, when testingservices andapplicationsinthe huge Anglo-Saxon environment, beforeintroducingand adaptingthese to other cultures.

Operational Costs

Theoperational cost of providingweb search is immense.Basically, these costshavetwo major components. Thefirst is thecost of contentcrawlingand indexing, which is practically proportional to thevolumeofthe content. Digital data has grownimmensely andthere arenosignsthat thenumber of sites anddata will notcontinue to grow immensely in thefuture. Consequently,the costsrelated to thegatheringand indexingofcontentwill also further increase.Crawlingisaresource consuming task, but notahugedistinguishingfactoramongstsearch engineprovidershavingtheir ownindexing system. As of 1June, forinstance,the ‘modest’ websearch engine, Exalead, hadcrawled more than 8,064 million sites,similar to thenumber of sites crawled by thebig search engines.The second, more important,component of theoperational costsisrelatedtothe operational costsofprocessinguser queries. Today’susersare very demandingand expect quasi-immediate responses to their queries.The processing powerneeded to answer queries is proportional to thetraffic,i.e., theuser population. So tooare the costs. Forthe big search engines,with alot of traffic,these costsare immense.Inawell-designed business,not only thecosts, but also therevenues shouldbeproportional to traffic anduser population. Therefore, thetraffic to asearch enginecouldbegradually ramped up by balancingincomeand expenditure. However,this kind of organic growth wouldneed to be supported by otherfactors likehaving thetop search technology,agood brandname, well-designed advertisingcampaigns, financial stability etc.,renderingitvery difficult foranewcomer to catch up with themarket leaders. This entry barrier will probably becomemorepronounced themoresearch movesfromtext-based search to audio-visual contentretrieval, as it is technologically more demandingand theoperational costswill further increase.

Progress in technology can contribute to reducingoperational costs. These challengesinclude quicker algorithms, more efficientserver architectures,cheaper hardwareand less energy consuming processors. In fact, minimizingthe powerconsumption of server farmsisareal challenge.State-of- the-art server farmsconsumesomuchenergythat cheap electricity andaccess to cooling(e.g., rivers, lakes) areamong themost importantcriteria fortheir location.

Centrality of Search between Specializationand Integration of Services

Themoresearch has becomeacentral technology forcitizens andbusinesses, themorestringent arethe requirements in termsofperformance in retrievingrelevant contentfor theuser.The need to deliver increasingly good relevancepartly explainsthe concentration of generalsearch engines andthe emergenceofthematic search engines.Inthe first case,only afew providerscan maintain thepressure of high operational costs. These costscomprise thecreation of server farms, their maintenance and operational costs, thedevelopmentcostsfor better performance andalsonew adjacentservices.

61) http://www.yandex.ru/ 62) http://morfeo.centrum.cz/ 63) http://www.netsprint.pl/serwis/

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 27

Thereforeonly afew generalsearch engines aresuccessfully operatingworld-wide. In viewofthe fact that evenmoreprocessingpowerfor audiovisual contentisneeded than fortext, theentry barrier is unlikely to be loweredfor new incumbents of futureaudiovisual search engines.Aconsolidation aroundafew generalweb search engines is likely.Analternativeisspecialisation in anichearea. This is thecase forthe thematic search engines.They providerelevant retrievalinspecific domainsand the entry barrier is more modest that forgeneral websearch engines.

Search engines arebecomingincreasingly integrated with other applicationsand services.Major search engines offer apackage of free services,suchasemail, chatting, instantmessages, voice services,etc. Manyofthese services arenot profitableindividually,but makesense at acorporate level to keep userswithin the“realm” of theservice engineoperator.Assearch engineprovidershavean interestinincreasingthe “stickiness” of their users, they try to personalise services. Through personalization, they aremorelikely to be abletoanticipate queriesand this will enablesearch engines to operate proactively.Ratherthan just “pulling” information, they couldalso“push”, informationand thus targetadvertisements evenbetter.

2.2.Business Search Solutions

In early times, companyinvestments were largely devoted to theestablishment of an efficientIT infrastructure. With time, companies havegathered thenecessary resources andtechnologies to capture, storeand transfer theinformation theenterprise needs forits operation. Oneremaining bottleneck is to provideaconsolidated user-centred viewfor employees to maketheir jobs easier and allow them to be more efficient. This shift from abasically storage-oriented infrastructuretoinformation consumption goes alongwith auser-centric modelrather than atechnology-based one. Providingan efficient,interactiveand secureway to presentuser-specific contentiscomplex,because it has to take into accountdifferentoperational systems, fileformats, schemas,etc. Oneexampleofbusiness search solutions is intelligenttoolsfor providersofyellow (andwhite) telephonepagesservices. These business search toolsfirst retrieve information from thecompanies’ databases,but can cross-check them with other sources.Through analysis andcomparison, data that is missingorinconsistentcan be detected.Another exampleisinthe healthcare sector.Medical records,images, radiographies,laboratory analyses,computertomographyand other relevant medical information is increasingly “born” digital. Thereisaneed forautomatic toolstoretrieve,analyse andinterpret such medical information.

Therefore, tailored search solutionsfor business andenterprise areanemergingfield. Identifying andenablingspecificcontentacross an enterprise to be indexed, searched anddisplayedtoauthorized usersgives addedvalue to companies. Followingareport by theconsultingfirm IDC, theworldwide market forenterprisesearch andretrievalsoftwarein2005 wasUSD 976 million. This hadgrownby 32% with respect to thepreviousyear.The size of this sector is notably smaller than theaforementioned websearch advertisementmarket. Thethree big players, Google, Yahoo! andMicrosoft havesome activity in this field, but their revenues from licensingtechnology aremodest compared to their core business.Incontrastwith theweb search market, thebusinesssolutionmarket is fragmented andthere areanumber of highly specialised companies on themarket. Thereare also someEuropean companies: amongthe most prominentofthese arethe Norwegian companyFAST (recently acquired by Microsoft), Autonomy in theUnited KingdomorExpert System SpA in Italy.These companies sell knowledge management toolswhere thesearch function is an increasinglyimportant software module.Withthese toolscompanies hope to effectively retrieve information andalsotouncovermeaningarisingfromany enterprise informationincludingdocuments, e-mails, entries in relational databases,etc.

Today,the market for‘knowledge management’toolsisvery distinct from theweb search engine market. ThefutureInternet, however, will becomeless text-based andmoreaudio-visually driven. The more websearch engines need to deal with audio-visualcontent, themorethe technological interest in theneed to developsolutionsfor conceptual search, documentclassification, text mining, and information analysis andcorrelation will overlap. This may drivecurrentweb search engines to penetrate the“knowledge management”market further.The factthat Microsoft recentlyacquired FAST may be an indicator of this trend.64

64) http://www.01net.com/editorial/368946/microsoft-s-achete-la-place-de-numero-un-de-la-recherche-en-entreprise/

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 28 SEARCHINGFOR AUDIOVISUAL CONTENT

2.3. Mobile Search

MobileSearch referstoinformationretrievalservices accessiblethrough mobiledevices likephones or PDAs. European telecom operatorsalready providesomesearch optionsfor their 2G, 2.5G and3G services.For these,they rely on technology provided by external companies likeGoogleorFAST and manysearch portalsofferingadedicated interfacefor handheldservices,likeMetaGer.

Although prospectsfor business models arenot yetconsolidated,the mobilesearch market is likely to differ significantly from theweb search enginemarket. Thetechnological context (e.g., small screens, limited bandwidth), thereduced amountofsuitablecontentfor mobiledevices,the role of themarket players(e.g., as telecom operatorstothe Internet by mobiles haveamorepowerful role as providersthan Internet service providershavefor accessingthe internet via acomputer), user behaviour (e.g., type of search requested on themove), may demandadifferentsearch enginebusiness model. Walled-gardenmarketsseem to be theprevailingmodel, but themarket may becomemoreopen in thefuture. Thepossibility of flat-rate pricingisbeingdiscussed andbandwidth restrictionsmay forcepaymentbybit download. This wouldmakeadifference notonly forbandwidth-intensive downloadingsuchasvideo (e.g., theremay be pricingbyvideo per resolution), but also forlocation- based services,which areregardedasbeingvery promisingand wouldallow us to findthe nearest restaurantbytypingthe question, or simply speaking, into our mobiletelephone.

Although themarket is in itsinfancy andfragmented, thetrendsare highly promisingand closely linked to thenumber of subscriberstomobilebroadbandconnections. Although still tinyinabsolute terms, thegrowth rates areimpressive.The GSM Association reported 32 millionmobilebroadband connectionsinMarch 2008, up from only 3million broadbandconnectionsinMarch 2007!65 More importantly,mobileInternet andsearch is an expandingmarket.Its main attractions arelocal information,likeweather,maps or directions. It is estimated that in 2008 thereare 2.5 billion mobile usersworld-wide, of whom roughly 40% have2.5 Gand 10% 3G technology.Asthe number of mobile subscribers is still increasingworld-wide(particularlyinhighly populated, less developed countries wherethe penetration rate is modest) andthe shareof2.5G mobiles or higher will be increasing, the number of subscriberswill also increase.Today,489 million peoplehaveaccess to mobileInternet (not necessarily broadband), andthis may doubleby2011. As thevast majority of themobile Internet users will also be searching, they arepotential customersfor search engineproviders. In fact, eMarketer considersthat themobilesearch advertisementmarket has reachedUSD 221 mio. in 2008 andwill increase to USD 2,361 mio. by 2011.66

Forthis to comeabout, however, anumber of challenges, which arise from thefactthat mobility imposes specific requirements with regardtouser interaction,retrievingdata anddisplays, need to be resolved. Another issue is adaptingorcreatingcontentsuitablefor mobiledevices.Contentsearch particularly will haveaddedvalue when contentisadapted to theuser andcombined with other technologies,e.g., location-based services.This will renderthe search experience more personalised.

Expandingthe Search Experience

Search toolsare tailored to retrieve informationfromspecific platforms. Currentweb search is mostly performed on personal computersorlaptops which access oneofthe more than 70 websearch engines world-wide.One challenge is to expandthe search capabilityfromother platforms, particularly formobiledevices.Technical possibilities forplatformsother than computersare still in their infancy. Forinstance,for mobilesearch, devices havetechnical limitationsand businessmodelsare not consolidated. Theideal from theusers’ point of view wouldbefull interoperability,astheir sole interest is to retrieve relevantcontentirrespectiveofthe technology.

65) www.gsmworld.com 66) www.eMarketer.com

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 29

3. Summary

Currentretrievaltools(mostly text-based) arekey technologies forthe Information Society.Given theexplosion of audiovisual data, futureAVsearch engines will becomeevenmorecentral to society than they aretoday.The provision of relevant resultsinAVsearch is farmorecomplex than in text- based search,and progress in AV search will dependonnew innovationsand also on improvingexisting concepts. Forinstance,retrievingrelevant audio-visualmaterial will benefit from new content-based search (audio, video, images) but it will also rely largely on meta-data (text) concepts. The technological challengesrange from basicscience to developmenttasks. Semantic approaches for search or noveltechnological conceptstomaster thegeneration andflow of huge amounts of data are examples of fundamental research tasks. Interactivesearch concepts, relevancefeedback systems, multi-modalanalysisorimprovedretrievalalgorithms areexamples of bottlenecks that mustbe removedinordertoimprove currentapplications.

Technology andbusiness considerationsgohandinhand. How to deal with theprocessing, storage andtraffic of huge amounts of audiovisual data is atechnological challenge that will requirenew computer architectures anddistributed search solutions. Thereare also importantfinancial implicationsthat couldinfluence considerably theentry barriersfor newcomersonto theAVsearch market. Thedeployment of AV search technology is,therefore,likely to reinforce manyofthe current techno-economic trends of theweb search market. Examples arethe concentration effectofgeneral- purpose websearch engines,orthe emergenceofthematic search networks. At themoment, thereis aclear distinction in theAVsearch market with regardtoweb search, business solutionsormobile search, due to thenatureofthe industry,the clientstructureand thebusiness models. In thefuture, these differences may becomeblurred. Websearch engineprovidersare already startingtoacquire companies offeringbusiness solutionsand non-walled gardenbusiness models may emerge in the mobilesearch sector.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France)

SEARCHINGFOR AUDIOVISUAL CONTENT 31

TheFuture of ContentNavigation

JanetGreco* BroadcastProjects,London

Introduction

As consumersare increasingly faced with amultitude of audio-visual andother contentchoices vyingfor their attention, how can they be surethey arenot missingthe contentthat really matches their interests? In this increasingly complex scenario contentnavigation mechanismssuchas Electronic ProgrammeGuides andsearch engines arekey.

1. ElectronicProgramme Guides at aCrossroads

It used to be simpletofindout what wasontelevision: aquick check of therelatively fewchannels listed in thedaily newspaper or thelocal weekly TV guidesufficed. Printed guides provided thetitles anddetailsofscheduled programmes in asimpletoread, column format, organized by time. Additional editorialcommentary andaphoto flagged noteworthyprogrammes of theday.Asthe number of channelsbegan to increase,printed “grid” format TV listings were introduced, andbythe mid 1980s they becamethe preferred andultimately dominantformat. Since theintroduction of Electronic ProgrammeGuides (EPGs) fordigital television systemsinthe mid-1990s,the grid format has continued to be themain structural buildingblock forcontentnavigation in thepay-TV environment. Over the last ten yearsEPGs havebecomethe essential tool fornavigatingcontentinthe mass market TV landscape.But with so manychannels, insufficient“screen real estate” andnoway to convey much of theuseful detail that couldbetter inform decisions, grid-format guides havereached their limitations.

More recently,Internet-based toolshavebecomecommonplace forlocatingcontentbeyond the television set. With theproliferationofTivo-likedevices andPersonal/Digital Video Recorders (PVRs), we havechangedour habitstonavigatingwhat is “alwayson” or “alwaysavailable”.Choices have increased manytimes over, rangingfromprofessionally produced mainstream entertainmenttothe avalancheof“user-generated content”, thankstovideo-sharingphenomena likeYouTube.The key challenge has now fundamentally changed: how will consumerslocate thecontentrelevant to their personal tastes? Mightmoreactiveparticipation andcomments from usersofsocial networks fundamentally affect how peopleshareand receiverecommendations?

Most peopleprobably havenot thoughtmuchabout theeffort it takesfor specialized intermediaries (e.g., TV “data aggregators” andpress agencies)toobtain regularly andupdate theprogramme schedules that they receivefrombroadcast TV channels. Today,this information is still disseminated

*Janet Greco is founderand principal consultantofBroadcast Projects(www.broadcastprojects.com).

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 32 SEARCHINGFOR AUDIOVISUAL CONTENT

by themarketingdepartments of broadcastersand channelsinavariety of data formats. It is useful to think of these “TV listings”intermsofhow youmightimport names andaddresses into your contact management software: thefieldnames havetobecompatibleand thefieldcharacteristics haveto match (e.g., youcannot import charactersinto a“numeric” field, norwoulditbeofmuchuse to import aperson’s nameinto the“country” field).Most of thetime, only narrativedocuments aresent, so these cannot be easily imported into adatabase.Moreover, publishersneed onestream of information fora multitude of channels, in asingledata format that supportstheir specificsystem requirements. They cannot manage multipleinboundformatscomingfromeachand every channel.

So eventhough international “metadata standards” exist, it is extremely difficult to ensurethat multipleproviderswouldspell, classify andverify this descriptive information in aconsistentway.As aresult, thedata aggregation intermediaries that perform this vital data processing, standardization andverification role remain importantplayersinthe dissemination of this information to professional publishers(onlineand in print).

Themain critical flaw in this process is that theintermediary is notdirectly linked (except in rare cases) to thetransmission andbroadcast play-out centres of thechannels. Therefore, thereisanatural disconnect between thedata disseminated by thedata aggregatorsand theactual broadcast transmission, which can giverise to errors. Nevertheless,the data harmonization functionsthey perform will continue to ensurethat these companieswill play an importantrole. Thecomplicating factors are, of course,the proliferation of onlineand on-demandcontent, without anyfurther need for“scheduled programmes”, andfindingaway to offer consumersamoreaccurate andpersonalized navigation system, coveringall formsofavailablecontent, no matter what thesource.

Addressinghow consumer-facingmetadata can be better organized requires significant organizational change.But it is notoften easytochange work habitswithin theprofessional broadcast environment, which has been making ahugeeffort to migrate to digital file-based systems. Moreover, theinternational metadata standards (such as TV Anytime), that havebeen so diligently andcomprehensively created, haveresulted in avery large number of possible“tags”that couldbeused to annotate thecontent that wouldfacilitate search andnavigation.

2. The Problem withMetadataStandards

With such standards, thereisanexpectationthat multiple contentproducers, TV channels, broadcastersand otherswouldtag theinformation in aconsistentway,enablingthe seamless interchange of information amongstavast array of services anddevices. In this context it is interesting to note that where“tagging”occursmost frequently andconsistently is onlineand by consumers, who originatedfor examplethe Internet Movie Database,1 andthe Compact Disc Database2 (thesystem that drivesiTunesmetadata).User-tagged content, often referred to as “folksonomies”,offersaninteresting angletoconsider in thecontext of this huge joboftaggingcontent. Such systemsare often self- correcting, in that intelligentsearchsystemscan determinesimilar groupings of wordstointerpret the intentionofasearch request. Metadata will remain akey driver of search engineresults. Video search innovators areattemptingtoaddress andwork aroundthese issues in manyvaried ways. But thesimple realityisthat they areattemptingtofill in, with very advanced technology,what is missinginthe simple form of consistentmetadata. As an extension of thesocial-networkingfocused Web2.0, the semantic weband beyond,itmay be more likely that consumers help with thejob of creating metadata once,tobeconsistently maintained in perhaps a”wikipedia”-like way.

Thechallenge today is that major contentownersdonot agreewith (or perhaps do notunderstand) theconcept that thedata that describes their contentshouldbefreely availabletoconsumersand inter-operablewith other systems. This is mainly apointofeducation:metadata is likeproduct packaging. But product packaginginthe digital marketplace is more amatter of data distribution and processing(an Information Technology issue) than it is about “marketing” or “advertising” in the traditional splashyway it has alwaysbeen understood.

1) http://www.imdb.com 2) http://www.cddb.org

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 33

International standards that deal with theunique identification of audio-visual works(e.g . ISAN: International StandardAudiovisual Number,www.isan.org )exist anditisprobably in thepublic interest that worksare registered once andcorrectly.ISAN, however, requires paymentfor the registration of all audio-visual works, whereas themusic recordingindustry took an entirely different approach and, likeISBN numbersfor books(which is also amucholder, but also paid-for system) it is unheardoffor major labelsnot to uniquely identify recordings. Underthe rightconditions, the application of ISAN(or anyother unique identification system) couldconsiderably un-complicatethis story. But evenifthe system amounted to theequivalentofaUniversal Product/“bar code” with which we areall familiar,the challenge persistsofchangingwork habits to integrate those unique codes into thelarge number of existingdatabases!

Manyindividualsilos of information remain today that replicate thesamedescriptiveinformation on audiovisual contentoverand overagain. These include thedata aggregators, contentowners, copyrightsocieties,audience research companies,TVguides andmanyothers. Whileprogramme synopses may differ accordingtothe language,writer or promoter of apiece of content(such as the descriptivesummary of an episodeof“Grey’s Anatomy’), other “static” detailsneed notbemaintained in separate databases.These include itemsofmetadata that neveralter: theoriginal language titleand episodetitle, thenames of theactors, thedirector,producer,and date of production. So aunique identifier makes economicsense,asthere is so muchduplication of effort in termsofmaintaining these databases.

So thechallengesare many, andnew legal issues arenow addedinto themix. With thepending implementation of theECAudiovisual Media Services Directive(by December 2009), andconsidering thevast array of other European regulatory andpolicyinitiatives,new requirements formandatory contentlabellingfor Television and“Video on Demand” services areemerging. Newchildprotection initiatives such as theUK’sByron Reviewwill result in thecreation of aUKCouncil on ChildInternet Safety (bySeptember 2008). Its recommendations, which arewholly supported by theUKgovernment, already provideone clear exampleofhow new contentlabellingrequirements will affect businesses. In this case,video games in theUKmustnow carry theBritish BoardofFilm Classification’s(BFFC) labellingsystem. TheBBFC also provides another interestingpiece of “metadata” in theform of a contentadvisory synopsis,which describes in quite somedetail thecontentofavideo game(or movie, etc). Whileother self-regulatory initiatives by video games producerssuchasPEGI3 sufficed until now as thelabellingmechanism to be employedinthe UK, theintroduction of theBBFC system underUK law represents just oneexampleofhow conflictinglabellingsystemswill ultimately affect the implementation of effectiveand coherentcontentnavigation systems. Theimplementation of this will also becomeincreasingly complex forcontentdistribution businesses workingacross multiple territories andthat will be required to adhere here to multiplelegal regimes.

3. The Importance of Consistent Data Input

In themeantime, consumersincreasingly struggle to identify contentofreal interest across a variety of platforms. Media companies andconsumer electronics firmsnow seek to fill in thegaps with new “advanced guide” productsand “contentnavigation” services.But today these only work well in awalled-gardenenvironment, wherethe data set forthe EPG is clearly defined (e.g., 500 channelson BSkyB, plusVOD andSky Anytime–this is amanageableset of data).

However, mergingsuch“walled gardens”with Internet contentisanentirely differentmatter.On theInternet, video search engines employ manytechnologies to identify content–e.g., those engines identify what is not thereinthe form of consistentmetadata (byanalysingaudio, subtitletracks, face recognition, etc). Andwhere watermarking/fingerprintingtechnology is used, these arenot yetfully linked up with arobust contentmetadata service (linkingtoany form of Unique IDs) –e.g., these representadditional silos of information.

It wouldbemucheasier if every piece of contentwereidentified -once-and couldthenbedigitally accessiblebyall. Audio-visual contentcouldbedescribed once in muchthe sameway as awordis

3) PanEuropean GameInformation (www.pegi.info).

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 34 SEARCHINGFOR AUDIOVISUAL CONTENT

identified anddescribed once each timeinadictionary.Static elements do notchange (e.g., the grammaticalform: noun, verb,etc)inmuchthe sameway as foraudiovisual content: title,director’s name, country of origin, date of production, etc, do notchange either.Descriptive(subjective) text couldbehandled on thebasis of linkstothe main record(differentlanguage versions, different synopsis lengths, differenteditorial perspectives,etc).

Thesemantic webpromises anew generation of information services.XML4 mark-up formatslikeTV Anytime 5 requireagreat deal of effort to implement. This is but oneofmanymetadata specifications. Foritall to work well everyonehas to do it andthe data input has to be consistent(imaginesomeone in Germanytaggingaprogrammeusingthe word “ Komödie” 6 ,whileusingthe term “sitcom” in theUS to describe thesameprogramme).

Forevery good argumenttosupport metadata, taggingand contentnavigation movinginthis direction, therewill also be correspondingcounter-arguments. Thepotential commercial motivation forcontentownerstoopen up suddenly their databases freely to theworldwouldhavetoimply:

•a better understandingofthe product packagingroleofmetadata in an IT environment(as opposed to itstraditional marketingfunction); •awillingness andanability to findaninteroperablesolution forthe protection of content(e.g., interoperabledigital rights management or DRM).

Fornow,the acquisitionand useofprofessional –robustly searchable–consumer-facingmetadata is aservice that has to be specifically commissioned andpaid fortoday,unless we arecontenttoleave thejob to thesearch engines.

4. Future Content Navigation Systems

In “search”, theonlineworldcontinues to havethe upper hand. Search engines arefar more advanced andabletoquickly (if sloppily) pinpointcontentbythe mere entry of asinglewordor phrase.And evenmorepertinentisthe factthat theyoungergeneration is accustomed to andhighly adept at manipulatingboth softwareand services to findwhat they want onlineand multitask while doingit. Thechallenge is plain forthe traditional pay-TV operatorsand,whenyou comparethe interfaces that exist in theSet-top-box-driven pay TV world, thegap is staggering.7

Advanced “guide” productsare coming, but accessingconsistentmetadata across both onlineand broadcast sources remainsahugechallenge.The futureofcontentnavigation is heavily dependent upon gainingaccess to arich source of accurate andconsistentmetadata. Especially importantare:

•One master record •Uniquelyidentified •Accessiblebyall (industry andthe public) •Correct in spelling, grammar andpunctuation •Multi-lingual •Extensively tagged •Extensible-sothat user generated comments anddescriptions(andphotos, videos,etc)can be appendedusingthe unique ID as anchor.

It remainstobeseen how themarket formetadata will evolvetoenablenew contentnavigation services to emerge.Advanced guideproductshavealready been designed, but havenot yetbeen deployedbecause theissue of accessingrobust metadata across both broadcast andInternet sources is still so difficult.

4) XML(ExtensibleMarkup Language)isaflexibleway to create common information formatsand shareboth theformat and thedata on theWorldWideWeb,intranets andelsewhere. 5) www.tv-anytime.org 6) In English “comedy”. 7) Forreferences see www.epgmetadata.com

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 35

Even themost professional metadata aggregation companies cannot cope with theincreasing requirements that theindustry now makes (DVD coverart, photos,video clips,keywords, international film ratings systems, contentadvisoriesand access information forthe disabled, etc). Andlabelling requirements willincrease as regulatory policies evolve. Theway forwardrelies upon education, incentives,toolsand public sector support to create an environmentwhere thedigitaleconomycan thrive. In such ascenario, aunique array of new productsand services can findtheir wayonto the market.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France)

SEARCHINGFOR AUDIOVISUAL CONTENT 37

Privacyand the Drive for the Perfect Audiovisual Search Engine

Michael Zimmer School of Information Studies Center forInformation Policy Research University of Wisconsin-Milwaukee

Introduction

Websearch engines haveemergedasubiquitousand vital toolsfor thesuccessful navigation of the growingonlineinformation sphere. Theworld’slargest search engine, Google, captures thegoal of all general-purpose search engines in itsmission statement: to “organize theworld’sinformation and makeituniversally accessibleand useful”. 1 To this end, Google, likemost search providers, strives to create “a perfectsearch engine[that] will process andunderstandall theinformation in theworld”2 andunderstands “exactly what youmean andgivebackexactly what youwant”. 3 Thedrivefor such a perfect search enginehas resulted in therise of powerful search toolsthat offer access to tensof billionsofonlineinformation sources,processingmillionsofqueries every day,and providingrelevant andoften personalized results–all forfree.

With therise of increased bandwidth speeds,cheaper digital technologies,the growingubiquity of webcamsand cellphonecameras,and innovativewebsites forcontentstorage andsocial interaction (such as Flickr or Facebook), theWeb is increasingly becomingaplatform forthe creation, storage and sharingofuser-generated multimedia content. As aresult, search engineprovidersstrivetoexpand thepowerand reachoftheir crawlerstoinclude image,audio andvideo contentintheir searchable indexes, offer their ownplatformsfor thestorage,retrievaland sharingofimagesand videos,and perhaps specialize in thedelivery of multimedia contentexclusively (such as theEuropean search engineproject, Quaero). In short, theperfect search engineisincreasingly becomingaudiovisual.

Thegrowingpowerand ubiquity of search engines havenot escaped controversy,includingconcerns about search enginebias,censorship (or itscircumvention), copyrightand trademark infringement, andfreespeech online. Perhaps themost potent, however, arethe privacy concernsrelated to power search tools.4 By strivingtomakeaudiovisual contentas“universallyaccessibleand useful” as its

1) Google. (2005). Companyoverview.Retrieved3May 2006, from http://www.google.com/corporate/index.html 2) GooglePress Center.(7June1999). Google Receives $25 million in equity funding [press release] (June1999) http://www.google.com/press/pressrel/pressrelease1.html (accessed 18 August 2006) 3) Google. (2007). Our philosophy .Retrieved27March 2007, from http://www.google.com/intl/en/corporate/tenthings.html 4) See,for example, Sullivan, D. (2 April 2003). Search privacy at Google &other search engines.Retrieved31March 2007, from http://searchenginewatch.com/showPage.html?page=2189531; Associated Press.(17 July 2005). Google Growth yields privacy fear.Retrieved31March 2007,fromhttp://www.wired.com/politics/security/news/2005/07/68235; Mills, E. (3 August 2005). Google balances privacy, reach.Retrieved7January 2007, from http://news.com.com/Google+balances+privacy,+reach/2100-1032_3-5787483.html; Zimmer,M.(2008). Privacy on planet Google: Usingthe theory of “contextual integrity” to clarify theprivacy threats of Google’squest forthe perfect search engine. Journal of Business &Technology Law, 3 (1), 109-126; Zimmer,M.(2008) Thegaze of theperfect search engine: Google as an infrastructureofdataveillance.InA.Spink, &M.Zimmer (Eds.), Websearching: Multidisciplinary perspectives.(pp. 77-99). Dordrecht, TheNetherlands: Springer.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 38 SEARCHINGFOR AUDIOVISUAL CONTENT

textual cousins, theemergenceofpowerful audiovisual search toolshas renewedconcernsoverthese issues amongacademics andregulatorsalike.5 Because of theunique natureofaudiovisual content, theprivacy-related issues that stem from thedrivefor aperfect audiovisualsearch enginedeserve particular attention. This contribution will describe someofthe latest trends in audiovisual search technologies –suchasthe integration of user-generated imagesinto searchableindexes, rich photo taggingand metadata, anduser-targeted facerecognition tools–and bringinto focusthe threatsto privacy they engender.

1. The Drive for the Perfect Search Engine

Since thefirst search engines started to provideaway of interfacingwith thecontentonthe Web, therehas been adrivefor the“perfect search engine,”one that indexesall availableinformation and provides fast andrelevant results.6 Aperfect search enginewoulddeliver intuitiveresultsbased on users’ past searches andgeneral browsinghistory. 7 Forexample, theperfect search enginewouldknow whether asearch forthe phrase “Paris Hilton” is meanttohelp auser locate hotel roomsinthe French capital or obtain thelatest celebrity gossip about theyoungsocialite,and deliver only theappropriate results in each case.Searchenginecompanies haveclear financial incentives forachievingthe “perfect search”: receiving personalized search resultsmightcontribute to auser’sallegiance to aparticular search engineservice,increasingexposuretothat site’sadvertisingpartnersaswell as improving chances that theuser woulduse fee-based services.Similarly,search engines can charge higher advertisingrates when advertisements areaccurately placed beforethe eyesofuserswith relevant needs andinterests(i.e., someoneseekingtravelinformation rather than celebritygossip).8

JournalistJohn Battelle 9 summarizeshow such aperfect search enginemightwork:

“Imaginethe ability to ask anyquestion andget notjustanaccurate answer,but your perfect answer –ananswerthat suitsthe context andintentofyour question, an answer that is informed by whoyou areand whyyou mightbeasking. Theengineprovidingthis answer is capableofincorporatingall theworld’sknowledge to thetask at hand–beitcaptured in text, video, or audio. It’s capableofdiscerningbetween straightforwardrequests–who wasthe third presidentofthe United States? –and more nuanced ones –underwhat circumstances did the thirdpresidentofthe United States foreswear his views on slavery?”

This perfect search also has perfect recall –itknows what youhaveseen, andcan discern between ajourney of discovery –where youwanttofindsomethingnew –and recovery –where youwantto findsomethingyou haveseen before.

When asked what aperfect search enginewouldbelike, Google’sSergey Brin replied, “likethe mind of God”. 10 To attain such an omnipresentand omniscient ideal, theperfect search enginemusthaveboth “perfect reach” to provideaccesstoall availableinformationonthe Web, and“perfect recall” to deliver personalized andrelevant resultsthat areinformed by theprevioushabitsofthat particular searcher.

5) Forexample, in September 2007, theJan vanEyckAcademie (Maastricht, TheNetherlands) held acritical forum on the European audiovisual search project Quaero, see http://www.janvaneyck.nl/0_2_3_events_info/arc_07_quaero_thinktank.html. This wasfollowedinApril 2008 by an expert workshop on theregulatory challengesfaced by audiovisual search engines,hosted by theInstitute forInformation Law at theUniversity of Amsterdam. 6) See Kushmerick, N. (23 February 1998). Thesearch engineers .The Irish Times,p.10; Andrews, P. (7 February 1999). The search forthe perfect search engine .The SeattleTimes,p.E1; Gussow, D. (4 October 1999). In search of…St. Petersburg Times,p.13; Mostafa, J. (24 January 2005). Seekingbetter Websearches.Retrieved30January 2005, from http://www.sciam.com/print_version.cfm?articleID=0006304A-37F4-11E8-B7F483414B7F0000 7) Pitkow,J., Schütze,H., Cass,T., Turnbull, D., Edmonds,A., &Adar,E.(2002). Personalized search. Communicationsofthe ACM, 45(9), 50-55; Teevan, J., Dumais,S.T., &Horvitz, E. (2005). Personalizingsearch via automated analysis of interests andactivities. Proceedings of the28th annual international ACMSIGIR conference on Research anddevelopmentin information retrieval, 449-456. 8) Hansell, S. (26 September 2005). Microsoft planstosell search ads of its own .The NewYork Times,pp. C1, C8. 9) Battelle, J. (8 September 2004). Perfect search.Searchblog. Retrieved16May 2006, from http://battellemedia.com/archives/000878.php 10) Ferguson, C. (2005). What’snext forGoogle? Technology Review, 108(1), 38-46.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 39

1.1. Perfect Reach

To achieve thereach necessaryfor therealization of Search 2.0, websearch engines amass enormous indexesofthe Web’scontent. Expandingbeyond just HTML-based Webpages, search engines providers haveindexedawidevariety of media foundonthe Web, includingimages, video files,PDFs andother computerdocuments. Forexample, in 2005 Yahoo! claimed to haveindexedover20billion items, includingover19.2 billion Webdocuments, 1.6 billion imagesand over50million audio andvideo files. 11 Furthermore, search engines areincreasingly digitizingthe “material world,” addingthe contents of popular books, university libraries,maps andsatellite imagestotheir growingindexes. In addition to their expansiveand diverse searchableindexes, today’ssearch engines also obtain a “perfect reach” by developingvarioustoolsand services to help usersorganize anduse information in contextsnot considered traditionalWeb searching.These include communicationand social networkingplatforms, personal data management,financial data management,shoppingand product research, computer filemanagement andenhanced Internet browsing.

Combining these twoaspectsofthe perfectreach –expansivesearchableindexesand diverse informationorganization products–the perfect search engineenables userstosearch, findand relate to nearly all formsofinformation they need in their everyday lives.The reachofthe perfect search engines allows userstosearch andaccess nearly all contentonthe Web, andalsoenables them to communicate,navigate shop, andorganize their lives,both onlineand off.

1.2.Perfect Recall

Complementingthe perfect reach of theperfect search engineisthe desireofsearch engine providerstoobtain perfect recall of each individual searcher,allowingthe personalization of both services andadvertising. To achieve this perfectrecall, websearch engines mustbeabletoidentity andunderstandsearchers’ intellectual wants, needs anddesires when they perform information seeking tasksonline. In ordertodiscern thecontext andintentofasearch for“Paris Hilton,” forexample, the perfect search enginewouldknow if thesearcher has recently showninterest in travel to France,or whether shespends significant timeonlinesearchingfor celebrity gossip.

Theprimary meansfor personalizingsearch resultsistorely on users’ search habitsand history. 12 To gather users’ search histories,most Websearch engines maintain detailed server logs recordingeach Websearch request processed through their servers, theIPaddress makingthe request, thepages viewedand theresultsclicked.13 Search engines also rely heavily on Webcookies to help differentiate usersand trackactivity from session to session, andincreasingly push thecreation of user accounts to help associate particular userswith their onlineactivity.The motivation behindgatheringthis user information is explained to theusersintermsofimprovingtheir search experience.Google, for example, states,“We usethis informationtoimprove thequality of our services andfor other business purposes”,14 whilethe search engineAsk.com also presents itseconomic motivationsfuelingthe need forthis perfect recall in pursuit of theperfectsearch: “Wecollect…anonymousinformation to improve theoverall quality of theonlineexperience,includingproduct monitoring, product improvement, targeted advertising, andmonetizingcommercially oriented search keywords”. 15 Given theexpansive array of productsand services offered by search engineproviders, their server logs potentially contain muchmorethan simply auser’sWeb search queries. 16 Search queries forvarioussortsofonlinecontent –images, news stories,videos,books, academic research, products, music,and so on –are almost

11)Mayer,T.(8August 2005). Our blog is growingup–and so has our index.Yahoo! Search Blog. Retrieved25November 2006, from http://www.ysearchblog.com/archives/000172.html 12) See,for example, Speretta, M. (2000). Personalizingsearch based on user search histories. University of Kansas; Pitkow,J., Schütze,H., Cass,T., Turnbull, D., Edmonds,A., &Adar,E.(2002). op.cit.;Teevan, J., Dumais,S.T., &Horvitz, E. (2005). op.cit. 13) See,for example, Google. (2005). Google Privacy faq .Retrieved3May 2006, from http://www.google.com/privacy_faq.html; IACSearch &Media. (13 July 2005). Privacy policy forAsk.com.Retrieved6January 2007, from http://sp.ask.com/en/docs/about/privacy.shtml ;Yahoo! (11 November 2006). Yahoo! Privacy policy.Retrieved6January 2007, from http://info.yahoo.com/privacy/us/yahoo/details.html 14) Google. (2005) Google Privacy faq. Op. cit.. 15) IACSearch &Media. Op.cit. 16) Wikipedia listsover100 differentproductsand services offered by Google, andover50byYahoo!

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 40 SEARCHINGFOR AUDIOVISUAL CONTENT

certainly logged, as well as linksclicked andrelated usage statistics from within non-search related services provided by search properties,suchase-mail, discussion groups,mappingservices andthe like.

In summary,the drivefor theperfect search enginehas ledsearch providerstoexpandtheir reach to index andmakesearchableawidevariety of information sources (both onlineand off), as well as enhancingtheir recall of particular users’ habits, needs andwants. Consideringthe steps taken to achieve theperfect reachand perfect recall noted above,search engines seem to be approachingthis omniscientand omnipresentideal andare poised to “becomethe perfect, all-seeing, all-knowing, all-powerful forceofthe 21st century”. 17

2. Privacy and the Perfect Search Engine

Thedrivetocreate theperfect search engine–one that promises breadth, depth, andrelevancy – is notcompletely benign. Thepath leadingtothe perfect search enginehas been fraughtwith concerns, includingbias,censorship (or itscircumvention), copyrightand trademark infringement,and free speech.18 Perhaps most potent, however, arethe privacy concernsabout thedrivefor theperfect search engine. 19 Both thereach andrecall necessarytoachieve theperfect search enginehave particular implicationsfor personal privacy.

2.1. Privacy andthe Perfect Reach

Achievingthe “perfectreach” requires search engines to index as manyWeb pagesand other online sources as arenecessary to providethe largest possibledatabaseofpotential search results. Among thebillionsofpagesindexedbysearch engines areWeb pagescontainingpersonal informationabout individuals, such as personal homepages(both activeand abandoned), discussion forum postings, onlineresumes,minutes of public meetings, property tax records andcourt records.Few peopleare notaffected by the“long arm of Google’sWeb crawler,” explainsjournalist Neil Swidey,referringto thebroad scope of Google’sWeb index:

“Maybe it wasastupid fraternity prank or acareless postingtoanInternet newsgroup in college. Perhaps youoncewentonarantataselectmen’smeetingorsigned apetitionwithout stoppingtoreadit. Or maybe youendured abitter divorce.You may think those chaptersare closed. Googlebegstodiffer.

Whilemost of your embarrassingbaggage wasalready availabletothe public,itwas effectively off-limitstoeveryonebut theprofessionally intrepid or supremely nosy.Now,in states wherecourt records havegoneonline, andthankstothe one-clickease of Google, you can read all thesordid detailsofyour neighbor’sdivorce with no more effort than it takes to check your e-mail.” 20

Engagingina“vanity search” –aWeb search forone’s ownname–can reveal asurprising amount of personal information: tax assessments,court documents, marriage licenses, deeds andvoter registration information, forexample. Thenotion of “Googling” someonebeforeablinddate has

17) Ayers, C. (1 November 2003). Google: Could this be thenew God in themachine? TheTimes,p.4. 18) See,for example, Grimmelmann, J. (forthcoming). Thestructureofsearch enginelaw. Iowa Law Review; Introna, L. & Nissenbaum, H. (2000). Shapingthe Web: Whythe politics of search engines matters. TheInformation Society,16(3), 169- 185; Elkin-Koren, N. (2001). Let thecrawlerscrawl: On virtual gatekeepersand therighttoexclude indexing. University of Dayton Law Review, 26, 180-209; Gasser,U.(2006). Regulatingsearch engines: Takingstock andlookingahead. Yale Journal of Law &Technology, 9 ,124-157; Goldman, E. (2006). Search enginebias andthe demise of search engineutopianism. Yale Journal of Law &Technology,188-200. 19) See,for example, Sullivan, D. op.cit.; Associated Press. op.cit.; Mills, E. op.cit.; Zimmer,M.(2008). Privacy on planet Google. Op.cit.;Zimmer,M.(2008) Thegaze of theperfect search engine. Op.cit.. 20) Swidey,N.(2February 2003). Anation of voyeurs: Howthe internet search engineGoogle is changingwhat we can findout about oneanother -and raisingquestionsabout whether we should.The Boston Globe Sunday Magazine, p. 10. 21) Lobron, A. (5 February 2006). Googlingyour Friday-nightdate may or may notbesnooping, but it won’t let you peek inside anysouls.The Boston Globe Magazine, p. 42.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 41

becomecommon practice. 21 Almost oneinfour Webusershavesearched onlinefor information about co-workersorbusiness contacts,22 andemployers areGooglingprospectiveemployees beforemaking hiringdecisions.23 In less than an hour,one reporter uncoveredavariety of personal detailsofGoogle CEO Eric Schmidt’slife:

•Schmidt doesn’t reveal muchabout himself on his home page.But spending30minutes on the Googlesearch engineletsone discoverthat Schmidt, 50, wasworth an estimated USD 1.5 billion last year.Earlier this year,hepulled in almost USD 90 million from sales of Googlestock andmade at least another USD 50 millionsellingshares in thepast twomonthsasthe stock leaped to more than USD 300 ashare.

•Heand his wife Wendyliveinthe affluenttownofAtherton, Calif., where, at aUSD 10,000-a-plate political fund-raiser fiveyearsago,presidentialcandidate Al Goreand his wife Tipper danced as Elton John belted out “Bennie andthe Jets.”

•Schmidt has also roamed thedesert at theBurningMan art festival in Nevada, andisanavid amateur pilot.24 25

Theexpansion of both thedepth andbreadth of their searchableindexestoobtain the“perfect reach” has reduced users’ sense of “privacy through obscurity”. 26 Amassiveamountofinformation about individuals, which wasoncedifficult to findand evenmoredifficult to cross-reference,isnow readily accessibleand collectiblethrough theuse of search engines.Asaresult, thereisagrowing ability to exploitthe increased publication of personalinformation online in ordertoengage in cyberstalking,27 perform aggregation anddata miningofpersonal information across otherwise disparate databases, 28 andtobuilddigital dossiersofindividuals.29 Ethicist Herman Tavani has written specificallyabout theease with which personalinformation can be routinely collected, aggregated and analyzed by Websearch engines:

“Because thevariousnews groups contain linkstoinformation posted by aperson, they can providesearch-engineuserswith considerableinsightinto that person’sinterestsand activities. 30 So it wouldseem to follow that notall of thepersonal information currently includedonWeb sites accessibletosearch engines wasnecessarily either placed therebythe personsthemselves or explicitly authorized to be placed therebythose persons.” 31

An individual mightnot be awarethat hernameisamong those includedinone or more of those databases accessibletosearch engines,let alonefluentinhow thesearch engines themselves work and their ability to retrieve personal information from avariety of onlinesources.

22) Sharma, D. (21 October 2004). Is your boss Googlingyou? Retrieved6January 2007, from http://news.com.com/Is+your+boss+Googling+you/2100-1038_3-5421210.html 23) Weiss,P.(19 March 2006). What atangled Webweweave: Beinggoogled can jeopardize your job search.New York Daily News. Retrieved7January 2007. 24) Mills, E. (3 August 2005). Op.cit. 25) Ironically,Googlepunished CNET forpublishingthe personal information about Schmidt –foundvia itsown search engine –with aone-year boycott against answeringany inquirersfromthe news service.Amid public criticism, Googleendedthe boycott twomonthslater. 26) Swidey,N.(2February 2003). op. cit.;Ramasastry,A.(12 May 2005). Can we stop zabasearch –and similar personal information search engines? When data democratization verges on privacy invasion.Retrieved12June2006, from http://writ.news.findlaw.com/ramasastry/20050512.html 27) Tavani, H. &Grodzinsky,F.(2002). Cyberstalking, personal privacy, andmoral responsibility. Ethics andInformation Technology,4(2), 123-132. 28) Garfinkel, S. (2000). Database nation: Thedeath of privacy in the21 st century (1st ed.). Sebastopol, CA: O’Reilly. 29) Solove,D.(2004). Thedigital person: Technology andprivacy in theinformation age (Ex machina). NewYork: NewYork University Press. 30) Asearch formynameuncovers (admittedly forgotten) poststoUsenet discussion forumsfromthe early 1990s on topics rangingoverabortion rights, Catholicism, feminism, marketing, andLotus1-2-3 spreadsheet software. 31) Tavani, H. T. (2005). Search engines,personal information andthe problem of privacy in public. International Reviewof Information Ethics,3,39-45.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 42 SEARCHINGFOR AUDIOVISUAL CONTENT

JohnBattellebest summarizes theanxiety that emergesfromthe perfectreach of search engines:

“What do we do when information that we know,bylaw,shouldbepublic,becomes, well…really public?...What happenswhenevery singlethingthat’severbeen publiclyknown about you–fromamention in your second-grade newsletter (now online, of course) to the vengeful ravings of aspurned lover–trailsyour nameforever?” 32

2.2. Privacy andthe Perfect Recall

Accompanyingthese privacy concernswith theincreasingly perfectreach of theperfect search engineare growingconcernsabout theability of search engines to collect andaggregate awidearray of personal andintellectual information about their usersinpursuit of theperfect recall. Themost obviousconcern involves thetrackingofindividual search queries andthe ability to associate them with particular IP addresses.This practice of collectingand retainingsearch query data in support of attaining“perfectrecall” gained widespreadpublic attention in January 2006 when it wasrevealed that, as part of thegovernment’seffort to upholdanonlinepornographylaw,the U.S. Departmentof Justice hadasked afederal judge to compel theWeb search engineGoogletoturn overrecords of millionsofits users’ search queries33.Googleresisted, but three of itscompetitors, America Online (AOL), Microsoft andYahoo!, complied with similar governmentsubpoenasoftheir search records34. Later that year,AOL released over20million search queries from 658,000 of itsuserstothe public in an attempt to support academic research on search enginequery analysis35.Despite AOL’sattemptsto makethe data anonymous, individual usersremained identifiablebased solely on their search histories, whichincludedsearch terms matchingusers’ names,social security numbers, addresses,phone numbersand other personally identifiableinformation 36.

These cases broughtsearch query retention practices into amorepublic light, creatinganxiety amongmanysearchersabout thepresence of such systematic monitoringoftheir onlineinformation- seekingactivities, 37 andleadingnews organizationstoinvestigate andreport on theinformation search enginesroutinely collectfromtheir users.38 In turn, variousadvocacy groups havecriticized the extenttowhich Websearch engines areabletotrackand collect search queries,often with little knowledge on thepart of theusersthemselves, 39 whileboth European andU.S. governmentregulators havebegun to investigate search enginequery retention practices andpolicies. 40

Alongwith thegrowingpublic attention focused on thepractice of archivingusers’ websearch queries in server logs, new concern is emergingregardingsearch engineproviders’ ability to monitor

32) Battelle, J. (2005). Thesearch: HowGoogle andits rivals rewrote therules of business andtransformed our culture. NewYork: Portfolio, p. 193. 33) Hafner,K., &Richtel, M. Google resists U.S. subpoenaofsearch data.The NewYork Times,pp. A1, C4; Mintz, H. (16 January 2006). Feds after Google data: Records soughtinU.S. quest to reviveporn law.San Jose Mercury News. Retrieved19January 2006, from http://www.siliconvalley.com/mld/siliconvalley/13657386.htm. 34) Hafner,K., &Richtel, M. (20 January 2006). Google resists u.s.Subpoenaofsearch data.The NewYork Times,pp. A1, C4. 35) Hansell, S. (8 August 2006). AOLremovessearch data on vast group of Webusers .The NewYork Times,p.C4. 36) McCullagh, D. (7 August 2006). Aol’sdisturbingglimpse into users’ lives. Retrieved3December 2006, from http://news.com.com/AOLs+disturbing+glimpse+into+users+lives/2100-1030_3-6103098.html?tag=st.num 37) Barbaro, M., &Zeller Jr,T.(9August 2006). Aface is exposed forAOL searcher no.4417749.The NewYork Times,p.A1; Hansell, S. (8 August 2006). op.cit.; McCullagh, D. (7 August 2006). Aol’sdisturbingglimpse into users’ lives.Retrieved 3December 2006, from http://news.com.com/AOLs+disturbing+glimpse+into+users+lives/2100-1030_3-6103098.html?tag=st.num 38) Glasner,J.(5April 2005). What search sites know about you.Retrieved2August 2006, from http://www.wired.com/news/privacy/0,1848,67062,00.html; Ackerman, E. (19 August 2006) What do Google,Yahoo, AOL andMicrosoft’sMSN know about you? San Jose Mercury News. 39) See,for example, Electronic Frontier Foundation. (2007). Privacy andsearch engines.Retrieved28July 2007, from http://www.eff.org/Privacy/search/; Privacy International. (9 June2007). Arace to thebottom: Privacy rankingofinternet service companies.Retrieved10July 2007 from http://www.privacyinternational.org/article.shtml?cmd%5B347%5D=x-347-553961 40) Associated Press.(25 May 2007).EU data privacy officerslaunch investigation into Google’sinternet search engine .Retrieved 28 July 2007, from http://www.iht.com/articles/ap/2007/05/25/business/EU-FIN-EU-Google-Privacy-Probe.php; Lohr,S. (29 May 2007). Google Deal said to bringU.S. Scrutiny .Retrieved27July 2007, from http://www.nytimes.com/2007/05/29/technology/29antitrust.html

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 43

andaggregate activity across their array of productsand services. 41 Theencouragement of thecreation of user accounts, thecross-promotion of services to keep usersfromleavingthe search engine’s website andthe useofpersistentWeb cookies across services,armssearch engines with theability to collect andaggregate awidearray of personal andintellectual information about their users, notjust extendingbeyond thekeywordsthey search for, but also includingthe news they read,the interests they have, theblogs they follow,the booksthey enjoy andother intimate detailsabout aperson’s identity,political interests, health status, sexlife,religion, financial statusand buyingpreferences. 42 Theresult is what John Battellecallsa“database of intentions”:

“This information represents, in aggregate form, aplace holderfor theintentionsof humankind-amassivedatabase of desires,needs,wants, andlikes that can be discovered, subpoenaed, archived, tracked, andexploited to all sortsofends. Such abeast has neverbefore existed in thehistory of culture, but is almost guaranteed to grow exponentially from this day forward. This artifactcan tell us extraordinary things about whoweare andwhat we want as a culture.” 43

Whilemanyofour day-to-day habits–suchasusingcredit cards,ATMs,cell phones or automated toll collection systems–leavecountless “virtual footprints”ofour activities,the drivefor theperfect recallnecessary to buildthe perfect search engineresultsinthe construction of arobust infrastructure of dataveillance with thepowertomonitor andrecordusers’ search histories,e-mails, blog postsand generalbrowsinghabits, providing“an excellentsource of insightinto what someoneis thinking ,not just what that person is doing”. 44

3. Makingthe Perfect Search Audiovisual

With theproliferation of digital imagesand video contentonthe Internet andWorldWideWeb, search engineprovidershavelongsoughttoinclude audiovisual contentintheir searchableindexes. In 1995, AltaVista wasthe first search enginetooffer theability to search forimage,audio andvideo content, afeaturesoon offered by theother major search engines of thetime, includingLycos,HotBot andInfoseek. These early audiovisual search services incorporated access to multimedia files primarily on thebasis of filenameextensions(.gif, .avi, .au) andtextual references extracted from contextual materials, such as thefilename, metadata, hyperlink text or surroundingwords. 45 Today,whilesome specialty search engines survive(Picsearch, forexample, offersthe ability to search overtwo billion images), audiovisual search services aredominated by thelarge general-purposesearch engines. Google, Yahoo, Microsoft andAsk each offer theability to search imagesand video files,usingmuch thesameretrievalmethods developed by AltaVista overadecade earlier.

Whilestillheavily dependent on metadata to identify thecontentofimagesfor retrieval, innovationshaveemergedinaudiovisual searching, includingboth social andtechnological enhancements in audiovisual search services.The rise of photo-sharingwebsites has resulted in an explosion in theamountofpersonal imagesavailableonline, dominated by Flickr.Whileprovidinga robust andeasy-to-use platform foruploadingand sharingphotosonline, Flickr’skey innovation was in providingthe ability forusersto“tag” photos with their owndescriptivelabels, allowingsearchers to findimagesrelatingtoacertain topic such as place nameorsubject matter.Upon acquisition by thesearch engineYahoo, Flickr’s database of over100 million photos wassoon integrated into Yahoo’s ownimage search results, greatly expandingthe reachoftheir audiovisual search offerings.46

41) Glasner,J.(5April 2005). op. cit.;Ackerman, E. (19 August 2006). op.cit.; Waters, D. (31 May 2007). What search engines know about us.Retrieved2April 2008 from http://news.bbc.co.uk/2/hi/technology/6700997.stm 42) Zimmer,M.(2008) Thegaze of theperfect search engine . Op.cit. 43) Battelle, J. (13 November 2003). Thedatabase of intentions .Searchblog. Retrieved16May 2006, from http://battellemedia.com/archives/000063.php 44) Hinman, L. (2005). Esse est indicato in Google: Ethical andpolitical issues in search engines. International Reviewof Information Ethics, 3 ,19-25, p. 23. 45) Schwartz, C. (1998). Websearch engines. Journal of theAmerican Society forInformation Science,49(11), 973-982. 46) Yahoo! (26 June2007). Flickr-izingimage search.Yahoo! Search Blog. Retrieved10September 2007, from http://www.ysearchblog.com/archives/000466.html

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 44 SEARCHINGFOR AUDIOVISUAL CONTENT

Relyingonthe phototags created by theimage’s creator within Flickr helped Yahooincrease the accuracy of itsimage retrievalsearch engine: rather than usingsurroundingtextormerely thename of theimage in ordertodetermineits relevancy to asearch query,Yahoo can now takeadvantage of themultipletags purposely placed by ahuman beingtodescribe theimage.Googlesoon developed its ownway of improvingimage search results: theGoogleImage Labeler. 47 Usersare randomly paired together andpresented with an image from Google’sdatabase.Eachuser provides labelstodescribe theimage,without seeingthe other’ssubmissions. If amatch is foundwithin 120 seconds,points are earned. Though thefeatureisenjoyablefor theusers, it is also acleverway forGoogletoensurethat itsimagesare matched with themost appropriate keyword.

Alongwith thegrowinguse of human andsocially-generated tags to aid theidentification and retrievalofimages, advances in automated content-based image retrievaltechnology havegreatly enhanced audiovisual searching.48 Rather than relyingonhuman-generated metadata, content-based systemsattempt to recognize andcataloguedifferentshapes,colorsand textureswithin images, automatically distinguishing between an image of abrown dogand ared apple, forexample.The ability to automatically identify anddistinguish between faces within image retrievalsystemshas recently received significantattention, both forlaw enforcementand commercial ends. 49 Increasingly,these facial recognition systemsare beingdeployedinthe consumer realm andintegrated into Websearch engines.For example, after acquiringNevenVision, acompanythat develops technology designed to detect andrecognize objectsand personsinimages,50 Googleenhanced their Image Search service to restrict theresultstoreturn only imagescontainingfaces.For example, abasic search for“Zimmer” in Image Search provides varied results, rangingfromthe exotic car of that name, randompictures of rooms(“Zimmer”isthe German word for“room”) andsomephotosofpeople.But selectingthe content- type “faces” in theadvanced image search settings deliversonly resultscontainingpeople’sfaces.

Recently,variousaudiovisual search services havebeen released focusingsolely on facial recognition within images. Launched in early 2006, thephoto search service Riyaprovides facerecognition and text recognition technology to identify faces within auser’sdigital photocollection andtag them accordingly. 51 Once theuser has manually trained thesystem by taggingaperson’sfaceinnumerous photos,the softwaretakes over,and can automatically scan, identify andtag that personinany future photouploadedonto theuser’slibrary. 52 Riyahad amillion photos uploadedinthe first twodaysafter launching, andsevenmillion photos uploadedinthe first sevenweeks.53 Buildingonthis success,Riya announced planstoexpandits facerecognition technologies beyond people’spersonalphoto collection to include anyimage on theWeb.Through itsplanned “visual search engine,”userscan submit an image to Riya(aparticularrug design, apurse seen in aphotograph, or evenaperson’sface), andit will return image resultsthat aresimilar from across theWeb. 54

TheSwedish-basedPolar Rose plans similarinnovationsinusingfacial recognitionfor image retrieval.55 Rather than relyingonuserstouploadtheir personalimage libraries to be analyzed (like Riya), Polar Rose planstorelease aWeb browser plugin that will identify faces in imagesonnearly any website,allowinguserstotag faces they recognize as they surf theWeb. 56 This data is then sentto

47) Lenssen, P. (1 September 2006). Google image labeler.Retrieved11July 2007, from http://blogoscoped.com/archive/2006-09-01-n44.html 48) See,for example, Lew, M., Sebe,N., Djeraba, C., &Jain, R. (2006). Content-based multimedia information retrieval: State of theart andchallenges. ACMTransactionsonMultimedia Computing, Communications, andApplications, 2(1), 1-19; Vasconcelos,N.(2007). From pixels to semantic spaces: Advances in content-based image retrieval. Computer,40(7), 20-26. 49) Zhao, W.,Chellappa, R., Phillips,P., &Rosenfeld, A. (2003). Face recognition: Aliteraturesurvey. ACMComputingSurveys, 35(4), 399-458. 50) Google. (15 August 2006). Abetter waytoorganize photos? Retrieved10July 2007 from http://googleblog.blogspot.com/2006/08/better-way-to-organize-photos.html 51) See http://www.riya.com/ 52) Ribeiro, J. (19 January 2006). Riya photo search site readies forlaunch.Retrieved11July 2007 from http://www.pcworld.com/article/id,124407-c,searchengines/article.html 53) Arrington, M. (15 June2006). Riya 2.0 on theway; major strategy shift.Retrieved11July 2007, from http://www.techcrunch.com/2006/06/15/riya-20-on-the-way-major-strategy-shift/ 54) Arrington, M. (15 June2006). Op.cit. 55) See http://www.polarrose.com/. At thetimeofwriting, Polar Rose is still aprivate beta test. 56) Simonite,T.(19 December 2006). Face-huntingsoftwarewill scour Webfor targets.Retrieved11July 2007, from http://technology.newscientist.com/article/dn10828-facehunting-software-will-scour-web-for-targets.html

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 45

Polar Rose’scentral server allowinganyonelookingatanimage containingthat particular facetotell whoitis, andconfirm or correct itstag.Thus, if they see aphoto that is clearly Bill Gates but tagged as Steve Jobs,they can makePolar Rose “smarter” by fixingthe mistake. Userswill also be ableto search theweb formorephotoscontainingthat face, based on theunique biometric profilecreated forthat facebyPolar Rose.

In summary,audiovisual searchservices havewitnessednumeroustechnological advances,including themergingofuser-generated imagesinto searchableindexes, thewidespreadtaggingofimage files andthe deployment of user-based facerecognition systems. Through these innovations, Websearch providershavebeen abletoextendtheir reach to include theincreasingamountofuser-generated imagesavailableonthe Internet. In fact, somesearch engines focussolely on providingaccess to audiovisual content, such as theEuropean Quaeromultimedia search research project, which includes amongits goalsthe facilitating of access to cultural heritage audiovisual archives anddigital libraries.57 At thesametime, as these new audiovisual search toolsprovideuserswith theopportunity to search forfaces,sounds,videos andplaces,search providerscan gain new insights into users’ wants and desires,enhancingtheir ability to recall users’ intentions. Forexample, knowingifauser typically searches forvideos of celebrities rather than imagesofEuropean landmarkscan help asearch engine providethe most relevant resultsfor thekeyword search “Paris Hilton.” In total, theperfectsearch engineisincreasingly becomingaudiovisual.

4. Privacy and the Perfect Audiovisual Search Engine

Part Two of this articledetailed theparticular privacy threatsthat emerge from theincreased reach andrecallattained in thedrive forthe perfect search engine. As theperfect search engineincreasingly becomes audiovisual, new,and perhaps more potent, privacy threatsemerge.This section will detail how theattemptstoadd audiovisual components to theperfect search enginedetailed above intensify theprivacy concernsoutlined above with regardtothe perfectreach andperfect recall of theperfect audiovisual search engine.

4.1.Privacyand the Perfect Reach of Audiovisual Search

As described above,the perfect reachofsearch engines has reduced the“privacy via obscurity” of individualsbymakingvisiblewebsites,documents andother onlinefiles that were previously difficult to locate or access.The rise of advanced audiovisual components accelerates this disintegration of privacy via obscurity as usersincreasingly lose control overthe accessibilityand identifiably of their likenesses online. Image sharingwebsites likeFlickr havesimplified theability to upload, tagand share images–both of oneself andofothers. Combined with theubiquity of cellphonecameras,almost any moment of one’slife couldbecaptured, tagged andplaced online forall to see.With theincorporation of Flickr imagesinto generalYahoo! image search results, usersnolongerneed to be membersofthe Flickr community (or evenaware that it exists) to access theimagesstoredonthe service.Asaresult, an individual mighthavetheir phototaken, posted online, labeled with their nameand discoverable within theYahoo search engine, without everknowingit.58 Whileindividualsare frequently subject to havingtheir phototaken (especially in public settings), this new ability to havethese imagestagged andsearchablebymillionsthreatensexistingprivacy normsregardingthe availability of one’slikeness to mass audiences.

Theprevalenceofdetailed metadata exacerbates this concern. Digital photos typically makeuse of theexchangeableimage fileformat (Exif), which captures variousmetadata elements within each image,includingdate andtimeinformation, camerasettings, athumbnail forpreviewingthe picture on thecamera’sdisplay screen or in software, andpossibly copyrightinformation. Manyimagesare also geocoded withtags forlocationinformation,either manually by theuser or automatically through abuilt-in GPS receiver on thecameraitself. Such detailed metadata elements areviewable(andoften

57) See http://www.quaero.fr/ 58) Forexample, asearch for“Michael Zimmer” in theYahoo Image Search service revealsaphoto of me deliveringatalk, tagged with my nameand posted to Flickr,all without my knowledge.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 46 SEARCHINGFOR AUDIOVISUAL CONTENT

searchable) within onlinephoto-sharingwebsites,allowinguserstoknow precisely when andwhere particular photos were taken. As aresult, notonly arephotosofindividualsincreasingly available onlinewithout their knowledge,but awealth of detailed information about precisely whereand when that person waswhenthe photowas taken is also freely available, further lesseningthat person’s presumed “privacy via obscurity.”

Therise of robust facerecognition toolstoidentify andsearch forimagesonlinerevealsanother instance wherethe powerful reachofthe perfect audiovisual search enginethreatensuser privacy. 59 By no longerrelyingonuserstorecognize aperson’sfaceand tagitinaFlickr photo, faces arenow easily discoverablebased on their unique biometric features.With services such as RiyaorPolar Rose, auser coulduploaddozensofpersonal photos to train thesoftwaretolearn herown face, makingthe labellingoffutureimageseasier.However, Riyamakes that facial profileavailabletoother usersto search andidentify imagesaswell. Thus, adifferentuser couldsubmit arandomphoto foundonline, andthe Riyasoftwarewill scan itsdatabase formatchingfaces andautomatically label any recognizablefaces within thenew image.Userscouldalsoscan theentireWeb forfaces they have trained Riyatorecognize.The seconduser previously hadlittleway of identifyingthe person in the photo, but with facerecognition toolssuchasRiyaand Polar Rose,identification of unknownfaces is mucheasier.Further,Polar Rose also makes it easytosearch theWeb forother imagescontainingthat person’sface. Thiscan result in findingimagesonsocial networkingsites,personal homepages, employerwebsites,and so on. Whilepreviously all such imageswerepublicly viewable, it wasdifficult forarandomuser to know they all existed, or wheretofindthem. Through these new socially-oriented facerecognition tools, userslose theability to control thecontext andcommunity in which certain imagesare viewable.

4.2. Privacy andthe Perfect Recall of Audiovisual Search

Alongwith thethreatstoprivacy engenderedbythe expandedreach of audiovisual search, similar threatsexist when the recall of theperfect search engineisaugmented by thepresence of audiovisual content. Just as users’ generalsearch queries aretrackedand logged by search engines in ordertohelp predict their wants andneeds,audiovisual search terms–names,places,pornography, andsoon–will be similarly captured andpotentiallyassociated with users’ accounts.Morethan simplybeing embarrassing, somesearch queries couldshedlightonanindividual’ssocial or political views, and perhaps evenbeofquestionablelegality.Justaslaw enforcementofficialshaveshowninterest in trackingorgatheringgeneral search activities, 60 attention mightalsobepaid to audiovisual searches forcontroversial or illegal itemssuchaschildpornography, cartoonsofthe Islamic prophet Mohammed, Nazi memorabilia, or copyrightprotected material. In short, theunique type of query made possible by theaddition of audiovisual contentinweb search services threatenstomakethe “database of intentions” more detailed andpotentially more damaging.

5. Conclusion

In strivingtomakeaudiovisual content, includingmovingimages, as “universally accessibleand useful” as itstextual cousins, powerful new toolshavebeen addedtoWeb search offerings, including theinclusion of user-generated images, robust phototaggingand metadata, andthe emergenceof sophisticated facerecognition toolsplaced in thehands of everyday Webusers. These new audiovisual enhancements bringwith them significantprivacy concerns, as both thereach andrecallofthe perfect search engine now include people’sfaces,biometric identifiers, detailed metadata andthe history of thekinds of multimedia contentthey search for. Thus akindofFaustian bargain emerges: Audiovisual search technologies promise to enhance theperfectsearch enginebymakingavailableawealth of

59) Granick, J. (5 July 2005). Face it: Privacy is endangered.Retrieved8April 2008 from http://www.wired.com/politics/law/commentary/circuitcourt/2005/12/69771 ;Inman, M. (5 January 2007). Face recognition foronlinephoto searches sparks privacy fears .Retrieved5April 2008 from http://news.nationalgeographic.com/news/2007/01/070105-photo-search.html 60) McCullagh, D. (20 December 2006). Police blotter: Google searches nabwireless hacker.Retrieved13June2007 from http://www.news.com/2100-1030_3-6144962.html

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 47

multimedia content–muchofituser-generated –whilealsotappinginto thevalue of metadata and theWeb 2.0 ideology of havingusersidentify andlabel imagesacross theWeb to help futureretrieval. But new privacy concernsalsoemerge,includingthe continued loss of “privacy via obscurity,” thefact that userslose theability to control whether imagesofthemselves areuploaded, whether their faces arelabeled, andthe factthat audiovisual search activities can be addedtothe ever-growing“database of intentions” collected by search engineproviders.

What optionsexist forrenegotiatingour Faustian bargain with these emergingaudiovisual search tools? First, we mustexpandour conceptualization of privacy as it relates to contentavailableonline, includingimagesuploadedbyindividualsortaken of them in public places.Itbecomes vital to embrace amorecontextual notion of privacy, 61 recognizingthat just because aperson wasinapublic venue, it does notmean that an image shouldbetaken, uploadedtothe Web, identified, tagged andperhaps haveabiometric scan of theperson’sfacecreated andshared with all. Someimagesare meantfor a particular context, andperformingthese additional tasks–evenwith thepositiveintentofimproving theperfect search engine–mightstill violate one’sprivacy.

Second, we mustexplorenew legal andregulatory frameworkstoaddress theincreased flow of personal imagesonlineand thegrowth of sophisticated technologies designed to identify,tag and makethemsearchableonthe Web. Somejurisdictionshavetaken steps towardsanoutrightban on thepublishingofpeople’slikeness onlinewithout their explicit consent. Othersrequirethe blurring of anyidentifiablefaceinonlineimageswhere individualshavenot given their consent. Onecould also foresee a“notice andtakedown” process similar to that used foralleged copyrightviolationsunder theUSDigital Millennium CopyrightAct. Here,userswho findtheir likeness online can issue anotice to theuploaderofthe image,orthe website hostingit, to remove theitem in question.

Finally,varioustechnological optionsexist to help mitigate theprivacy concernsabout emerging audiovisual search tools. Even in theabsence of legal or regulatory requirements, theprovidersof audiovisual search couldprovidetechnological toolsfor individualstorequest theremovaloftheir image when foundonline, or create arobust reportingsystem –similar to that in thecredit industry –toinform individualsifanimage of their faceistaggedoruploaded. With regardtothe factthat users’ audiovisual searches mightbetrackedand includedina“databaseofintentions,”search providersshouldgiveusersaccess to anysearch activity collected, allowingthemtoremove anydata they do notwantrecorded, andprevent futurerecordingoftheir search activities.

In sum, therise of powerful audiovisual search toolshas thepotential to greatly enhance thedrive forthe prefect search engine. But only if themultipleprivacy concernscan be addressed –through broadeningour conception of privacy,creatingpolicyprotectionsand engaginginprivacy-protecting design –will we avoid followingthe path of Faustand makeabargain with thedevil in exchange for access to universal knowledge.

61) Nissenbaum,H.(2004). Privacy as contextual integrity. Washington Law Review, 79(1), 119-157; Barth, A., Datta, A., Mitchell, J. C., &Nissenbaum, H. (2006). Privacy andcontextual integrity: Framework andapplications .Paper presented at theIEEE Symposium on Security andPrivacy;Zimmer,M.(2008). Privacy on planet Google.Op.cit.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France)

SEARCHINGFOR AUDIOVISUAL CONTENT 49

FreedomofExpression Implicationsfor the Governance of Search

Joris vanHoboken Institute forInformation Law,University of Amsterdam

Introduction

Search toolsand services havebecomecentral in our informationenvironment. Themost straightforwardexplanation forthis developmentisthe abundance of onlineinformation, in particular on theWorldWideWeb,whileinnovationsinsearch andinformation technology havemade this developmentpossible. As aresult of this abundance,the question is now notsomuchwhether certain information is published, but whether an end-user will be abletofinditinthe midst of all theother material out there. Whilethe Internet andtext-based media standout because of theoverwhelming amountofsources andmaterial they havetooffer to their users, audiovisual contentisfollowing closely on their heels. AccordingtopredictionsofInternet traffic statistics by Cisco, “Video on demand (VoD), IP television (IPTV), andInternet TV will accountfor nearly 90 percentofall consumer IP traffic in 2012.”1 Therefore, thesamegate-keepingpowerisshiftingtowardsnew intermediaries, and questionsabout thegovernance of thetoolsand services that index theavailablecontentand help usersselect thecontentoftheir choice aregainingimportanceinthe policy debateabout the informationsociety.

This contribution will provideinsightinto thedebate about theregulation of search from the perspectiveoffreedom of expression andinformation. It will do so by lookingatsomeofthe governance issues relatingtosearch engines,fromthe righttofreedom of expression to theinterests of theprimary stakeholders in thefreeflow of information facilitated by thesearch medium. Where appropriate,particular attention will be paid to aspectsthat arespecific to theregulatory debate about audiovisual search.

In thefirst section search engines arecharacterized in termsoftheir functionality andafew examples aregiven of their growingimportance as an elementinthe Internet contentvalue chain (1.1). Followingthis,the focusisplaced on theprimary stakeholders with regardtothe governance of search engines (1.2) andthe interestsofthe primarystakeholders with respect to thefunctioning of search engines (1.3).

Thesecondsection will address to what extentthese interestsare protected underthe rightto freedom of expression andinformation,asenshrined in Article10ofthe Convention forthe Protection of Human Rights andFundamental Freedoms. After ashort examination of this generalquestion (2.1),

1) Cisco Visual NetworkingIndex -Forecast andMethodology,2007–2012, white paper,16June2008. Availableat http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-481360.pdf (Last viewed1July 2008).

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 50 SEARCHINGFOR AUDIOVISUAL CONTENT

theanalysis will proceed by comparingsearch engines to Internet filters(2.2) andthe freedom of expression issues they haveraised. In thecontext of this comparison, arecentRecommendation of theCouncilofEurope’sCommittee of Ministersonmeasures to promote respect forfreedom of expression andinformation with regardtoInternet filterswill be discussed (2.3).2 As will be shown, theCouncil of Europe’sRecommendationonInternet filtering andthe underlyingexpert Report3 provideaninteresting startingpointfor theanalysis of theimplicationsoffreedom of expression for thegovernance of theinclusion andexclusion of information providersbysearch engines,inother wordsthe issues surroundingthe collection of information providersinthe search engine’sindex that can be accessedthrough aparticular search service (2.4).4

Section three offersadiscussion on thequestion of theliability of search engines forthirdparty content, an important issue forthe governance of theinclusion andexclusionofinformation providers in asearch platform. Thesection will giveashort overviewofthe presentlegal regimefor search engineliability at theEUleveland characterizethe freedom of expression perspectiveinthis context. Section four concludesthe contribution.

1. Search Engines

1.1. WhatDoesaSearch Engine Do?

Generally speaking, searchengines help theend-users of acertainmedium to findcontent, includingaudiovisual content, which can be enjoyedthrough that medium. Internet search engines likeGoogleand Yahoo! providesuchservices forend-usersofthe Internet. Theprimary stakeholders with regardtoasearch service arethe service provider itself, theend-usersand theinformation providers. Search engines areinthe business of organizinginformation. Commercial search engine providersusually makeprofitsthrough thesales of targeted advertising. Theadvertisements can be linked to specific search actionsofend-users. Most search engines arecommercial andfollow such a business model. Forend-users, search engines areatool that helps them select information from alist availablethroughaparticularmedium. On theother sideofthe search service,wefindthe information providersthat haveaninformation output that couldqualify forinclusion in asearch engine. Thus, forinformation providers, search engines areameanstoreach an audience andthe search platform functionslikeaforuminwhich they compete forprominence andthe end-users’ attention. The importanceofbeingincludedinasearch engine’sindex is captured by thefollowingphrase: “toexist is to be indexedbyasearch engine.” 5 Although most information providerswill aim to be included andranked highly,insomecases information providerswill in factliketobeexcludedfromasearch engine.6

Thefunctional steps that areneeded to set up asearch enginecan be divided into two. First, the provider hastocompose alist or index of information from thefull setofinformationthatcan be made availabletothe end-user.Second, it can usethis list to offer an interactiveselection service,allowing theend-user to search theindex.Itfollows that aWeb search engineallows end-userstosearch their index,not theentireWorldWideWeb.

Forexample, in thecase of thesearchenginefor the Webasprovided by Googlethe available information is all theinformation on theWeb that itsautomatic spidersorcrawlerscan access.The

2) Recommendation CM/Rec(2008)6 of theCommittee of Ministerstomember states on measures to promote therespect for freedom of expression andinformation with regardtoInternet filters. Availableat https://wcd.coe.int/ViewDoc.jsp?Ref=CM/Rec(2008)6 (Last viewed1July 2008). 3) Report of theGroup of Specialistsonhuman rights in theinformation society (MC-S-IS) on theuse andimpact of technical filteringmeasures forvarioustypes of contentinthe onlineenvironment, as it appearsindocumentCM(2008)37 add, Availableathttps://wcd.coe.int/ViewDoc.jsp?Ref=CM(2008)37&Ver=add (Last viewed1July 2008). 4) Schulz, W.,T.Held, et al. (2005). Suchmaschinen als Gatekeeper in der öffentlicheKommunikation: RechtlicheAnforderungen an Zugangsoffenheit undTransparanz bei Suchmaschinen im WWW. 5) Introna, L. D. andH.Nissenbaum (2000). Shapingthe Web: Whythe Politics of Search Engines Matters. TheInformation Society 16(3): 169-186. 6) Seeinfrasection 2.4.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 51

search enginesends these crawlerstolook forinformation that is availableonlineand systematically makes an index based on their findings. Newtechnologies makeitpossibletodetect audiovisual content andmoreand more search engines havedeveloped ways to index audiovisual content. Alot of content on theWeb is notincludedinthis index.Manyinformation providersuse the robots.txt instruction file andother functionalities to instruct search engines’ crawlersnot to index partsoftheir websites and major search engines normally respect that. Other sources mightnot be includedfor other reasons, for instance because thecrawlersdonot havethe technical capabilities to extractthe information from a particular website.Yet another reason whyGooglemightnot include awebsite in itsindex is that it wouldbeunlawful to show thereference,for instance because it has received anotice underthe Digital Millennium CopyrightAct (DMCA). As aresult, Google’sindex containsaselection of theinformation availableonthe Web, aselection that is determined by information providers, thirdparties,Google itself, thetechnicalproperties of theWeb andthe law.

After theindex is made,the search engineletsend-usersinteractwith it by typinginkeywords. The search resultsare organized andranked to help theend-user findsatisfactory results, andtargeted advertisements areplaced nexttothe resultstomakeaprofit. Themethods that areused by major search engines likeGoogletoorganize andrank contentare acomplex combination of statistical methods.One of these methods is thefrequency analysis of keywordsonthe websites.Manysearch engines also perform an analysis of thepopularity andimportance of websites,for instance by makingahyperlinkanalysis. Google’soriginal PageRank algorithmisone of themost widely knownand successful algorithms that use thehyperlinksonthe Webtomakeanestimate of therelevanceofaspecific website.The original PageRank algorithmmeasures therelevanceofawebsite,e.g.,website X, by countingthe linksfromother websites that link to X. Alink carries more weightifitcomes from awebsite that is more linked to itself. Therefore, underPageRank, awebsite is most relevant if it has manylinkscomingfromvery relevant pages. Another factorthat is used by search engines to determinethe rankingofawebsite is thepast viewingand searchingbehaviour of end-user(s). This search andviewinghistory can be used to personalize search resultsand makerecommendationstoend-users. Thedetailsofthe more complex rankingalgorithms areusually kept secret.

1.2.Examples of Regulatory Attention forSearchEngines in Europe

Many havenoted theimportance of search engines andtheir impact on theinformation flows in theinformation environment. Internet search market leaderGooglereceives particular attention, but search engines in generalhavealsostarted to gain in importanceinthe eyesofEuropean regulators andpolicymakers.

In thecontext of Internet contentregulation, Ofcom (theUKregulator with responsibilities covering television, radio andelectronic communications), has started to allocate search engines aseparate segmentinthe Internet value chain.7 It divides this value chain into twobroad categories,namely entities that control theavailability of content, such as contentproducers, contentaggregatorsand Webhosts, on theone hand, andentities that control theaccessibility to content, such as Internet service providers, navigation andsearch engines andconsumer devices,onthe other.Ofcom takes the position that traditional regulatory intervention or strategies with regardtothe availability of content havebecomeless effective, due to theglobal Internet contentmarket andcomplicationsresultingfrom legal jurisdiction. 8 Whereascontentregulation used to address entities in thecontentvaluechain that controlled theavailability of content, Ofcom stressesthe need to focusalsoonthe entities that control theaccessibility of content. Ofcom concludesthat self-regulation, in particular by Internet Service Providers–which havestrongties with thejurisdictionsofend-users–isacrucial elementinapolicy that aimstomanage contentrisksfor children. In thecase of search engines,however, such alackof strongties withjurisdictionsisnotable. Ofcom concludesthat filteringshouldbepromotedasa solution, but that it is notfeasibletolegally requireISPs to filter at thenetwork layer,atype of contentcontrol that is practiced forinstance in China andSaudi-Arabia.9

7) See Ofcom -Office of Communications, Ofcom’sResponse to theByron Review, 27 March 2008, pp. 50-51. Availableat http://www.ofcom.org.uk/research/telecoms/reports/byron/byron_review.pdf (Last viewed1July 2008). 8) Ibid, pp. 50-51. 9) Ibid., p. 51.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 52 SEARCHINGFOR AUDIOVISUAL CONTENT

Another European regulatory debate on thesubject of search engines relates to their effectsas concernsprivacy andthe protection of personal data. Theuser data retention policies of large Internet search engines were investigated by theEuropean Union’sArticle29Data Protection WorkingParty. 10 Search engines tendtoretain detailed data on theuse of their services by end-usersoveralengthy period of time. Thelogsofsearch engines normally include thesearch termsentered, theip-address(es) from which auser is connected to theInternet at specific dates andtimes,aunique cookie identifying theuser andthe consequentuser navigation on thesearch engineplatform after thesearch results havebeen presented to theuser.The Article29Data Protection WorkingParty recently concludedan investigation into thedata processingpracticesofmajor search engines with adetailed official Opinion, which opensbystatingthat “Search engineprovidersonthe WorldWideWeb fulfil acrucial role in theinformation societyasintermediaries.The WorkingParty recognises thenecessity and utility of search engines andacknowledgestheir contribution to thedevelopmentofthe information society.”11

Finally,another reference to search engines can be foundinthe underlyingReport of theCouncil of Europe in thecontext of theimplicationsoffreedom of expression forInternet contentfilters.12 The Report clarifies that: “Most Internet userstoday rely on search engines to rapidly findthe most relevant information fortheir needs.Inthis way, search engines play an importantroleasgatekeepers in determiningand shapingwhichinformation is received through theInternet. This issue is accentuated when, as is thecase today,there is onedominantsearch engineprovider on themarket. Theway such asearch enginelistsand prioritizes search resultsmay in practice haveanimportant impact on what information auser receives.”13 In thesecondsection of this contribution, thefreedom of expression implications forsearch engines andInternet contentfilterswill be further addressed.

1.3. TheInterests of InformationProviders, End-users, SearchEngines andThird Parties

Beforeaddressingsomeofthe implicationsoffreedom of expression forthe governance of content accessibility through search engines, it is interestingtolook at theinterestsofthe primary stakeholders. In thefollowingsection,abroad characterization is given of what is at stakefor the search engineprovider,end-usersand information providerswith regardtothe accessibility of content through asearch platform. As will becomeapparent, these interestsare notalwaysinalignment. The question to be addressed later on in this contribution is to what extentthey areprotected under freedom of expression. Of course,thirdpartiesmay also haveinterestsinthe governance of search engines,suchasthe protection of their privacy andreputation or intellectual property,howevera discussion of these mattersgoes beyond thescope of this contribution.14

To start with thesearch process itself, onecan distinguish different types of issues that arise in the governance of theaccessibility of content.Asexplained above,the compositionofthe index is decisive forwhat can be foundthrough aparticular search engineand thereare manyreasonswhy aparticular source couldbeexcludedfromthe index of asearch engine. However, notonly thecomposition of the index,but also theconsequentrankingofsearch resultshas enormousimpact on what is beingfound in practice.Empirical research shows that end-usersusually do notgobeyond thefirst page of search

10)The Article29WorkingParty consistsofthe national data protection authorities of theEU’sMember States andhas thetask of facilitatingharmonized interpretation of theData Protection Directive(95/48/EC), Directive95/46/EC of theEuropean Parliamentand of theCouncil of 24 October 1995 on theprotection of individualswith regardtothe processingofpersonal data andonthe free movement of such data, OJ L281, 23 November 1995, pp. 31–50. 11) Article29WorkingParty,Opinionondata protectionissues related to search engines,WP148, 4April 2008.Atseveralpoints in theOpinion, theWorkingParty addresses theimportance of thefundamental righttofreedom of expression and information in thecontext of search engines.(“[…] abalance needs to be struck by Community data protection law and thelawsofthe variousMember States between theprotection of therighttoprivate life andthe protection of personal data on theone handand thefreeflow of information andthe fundamental righttofreedom of expression on theother hand.”). Availableat http://ec.europa.eu/justice_home/fsj/privacy/docs/wpdocs/2008/wp148_en.pdf (Last viewed1July 2008). 12) Supra, notes 2and 3and accompanyingtext. 13) Supra, note 3. 14) Foracomprehensiveoverviewofthe interestsinvolved in thegovernance of thevariousaspectsofsearch engines andthe information flows they facilitate,see James Grimmelmann, TheStructureofSearch EngineLaw, Iowa Law Review, Vol. 93, No. 1, 2007. Availableathttp://ssrn.com/abstract=979568

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 53

resultsand tendtoclick more on thefirst coupleofresultslisted: references to afilm on thefirst page determinewhether it will be foundornot.15 Thereforethere is more needed to exist, than just to be indexed. As aconsequence,awhole industry of search enginemarketingand optimisationexiststhat assistsinformation providersinachievingaprominentrankinginsearch resultsand managingeffective advertisingcampaignsinsearch engines.Finally,the search service mightbepersonalized forspecific audiences and/or end-users. We couldend up in asituation in which all end-usershavetheir own personalized search service based on their owninterestsand contextsofuse.

All primary stakeholders havespecific interestswith regardtothese governance issues.The search provider’sinterest is to be in control overits ownservice,includingthe composition of itsindex,its functioningand therelevanceand integrity of itssearch results. Theinterest of manyinformation providersistofeatureeffectively in asearch enginesoastoreach an audience,although it is worth mentioning that, forvariousreasons, somemay in factnot want to be includedinasearch engine. Since thebusiness models of search engines andmanyonlinepublishersare based on revenues from advertising, many publishersalsocompete with search engines foronlineadvertisingrevenue.16 Finally, theend-user’sgeneral interest is to search freely andfindthe information providersthat providethe most value.Thus, he/shehas an interest in beingabletoinform him/herself fully through usingthe medium. In thenextsection therighttofreedom of expression andinformation will be discussed in relation to these interestsofthe primary stakeholders.

2. Freedom of Expression and Information and SearchEngines

2.1.The Implications of Article10European Convention on Human Rights

AccordingtoArticle10ofthe European Convention on Human Rights, everyonehas therightto freedomofexpression, which includesthe freedomtoholdopinionsand to receiveand impart informationand ideas without interference by public authority andregardless of frontiers.17 Article10 para. 2provides that these freedomsare subject to possiblerestrictionswhich areprescribed by law, necessary in ademocratic society andserveone of alisted set of goals, such as theprotection of health or morals andthe protection of thereputation or rights of others. Theprotection of therights of others is understood to include theprotection of intellectual property rights.18

TheEuropean Court of Human Rights (hereinafter referred to as “ECHR”) has notyet addressed the question of search engines in anyofits rulings, so it is necessary to examineother case law instead in ordertoaddress theimplicationsofArticle10for thegovernance of search engines.One judgment that containsanumber of relevant considerationsisthe Autronic judgement,inwhich theEuropean Court of Human Rights reaffirmed that Article10applies notonly to natural persons, but also to profit makingcorporations. It concludedthat:

“neither Autronic AG’s legal statusasalimited companynor thefactthat itsactivities were commercial northe intrinsic natureoffreedom of expression can depriveAutronic AG of the protection of Article10(art. 10). TheArticle(art. 10) applies to “everyone”, whether natural or legal persons.” 19

15) Forthis andother empirical findings about search engineuse see thespecial volumeofthe Journal of Computer-Mediated Communications, Volume12, Issue 3, April 2007, Special ThemeI,The Social, Political, Economic,and Cultural Dimensions of Search Engines,edited by Eszter Hargittai. Availableat http://jcmc.indiana.edu/vol12/issue3/. Last viewed1July 2008. 16) TheEuropean Commission discusses this aspect in itsrecentCommunication on CreativeContentOnline, See Communication from theCommission to theEuropean Parliament, theCouncil, theEuropean Economic andSocial Committee andthe Committee of theRegionsonCreativeContentOnlineinthe SingleMarket. COM(2007) 836 final, Brussels, 3January 2008, pp. 34-35. 17) Council of Europe,Convention forthe Protection of Human Rights andFundamental Freedoms, 4November 1950, Europ.T.S. No. 5. 18) Foradiscussion see P. BerntHugenholtz, Copyrightand Freedom of Expression in Europe, in: R. Dreyfuss et al. (ed.), Expandingthe Boundaries of Intellectual Property.Innovation Policyfor theKnowledge Society,New York: OxfordUniversity Press,2001, pp. 343-363. 19) ECHR, 22 May 1990, A178, §47(Autronic AG v. Switzerland ).

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 54 SEARCHINGFOR AUDIOVISUAL CONTENT

Thus thecommercial andcorporatenatureofmajor search engines does notdeprivethemofpossible protection by Article10. In thesamejudgment theCourt clarified that:

“Article10(art. 10) applies notonly to thecontentofinformation but also to themeansof transmission or reception since anyrestrictionimposedonthe meansnecessarilyinterfereswith therighttoreceiveand impart information.” 20

It is unclear whether search toolsingeneral shouldbeunderstood to be ameansoftransmission andreception of information. Article10does notexplicitly refer to arighttoseek or gather information,but these communicativeprocesses aregenerally notconstrued as beingexcludedfrom itsscope.Interestingly,both theUnitedNationsUniversal Declaration of Human Rights andthe United NationsInternational CovenantonCivil andPolitical Rights do contain an explicit reference to the righttoseek information andideas in Article19.21

VanEijk already noted thedifficulties of qualifyingsearch engines in thelightoffreedom of expression due to their complex role. 22 Because it combines both beinganintermediary between information providersand end-usersand beingacommunicator itself, thesearch engineisacomplex medium. TheInternet access provider of an end-user is an essential part in anyonlinecommunication between that end-user andonlineinformation providers. If theaccess provider does notfacilitate the communication betweenthe end-user andthe information provider,itcannot takeplace.Search engines,however, arenot anecessary part in alinear chain of communicationsbetween information providersand end-users. An onlineinformation provider can be easily accessedwithout theuse of a search engine, either directly by typinginthe URL in abrowser,orindirectly by followingalink to theblogelsewhere,for instance on another website or through theuse of another search or recommendation service.

Forthese reasonsVan Eijk assertsthat thefreedomstoimpart or receiveinformation, as explicitly mentioned in Article10, arenot themain aim of search engines,since theinformation is already presentand directly accessible.23 Asearch enginedoes notand cannot change that. VanEijk concludes that search engines facilitate access to information anddonot offer access to information by themselves.Heassertsthat this activity of making accessible shouldhaveasimilarstatusunder Article10asthe activity of disclosingordisseminatinginformation andideas.

This perspectivefocuses on theultimate communication between information providersand end- usersthat is (partly) facilitated by search engines.Itshouldreceivesignificant weightbecause of the increasingimportance of search engines as mediatorsinour information environment; as pathfinders andasforatoreach audiences.The question, however, is whether Article10does notalsoprotect the communicationsbetween theinformationprovider andthe search engineand thecommunications between thesearch engineand theend-user.Byzoominginonthe communications that aretaking place between theprimary stakeholders, we may be abletodrawamoredetailed analysis of the freedom of expression implications forthe governance of search engines.

Thesearch engine’scrawlersare ameansofreceivinginformation, in somewaysamodern, automated version of theresearcher or journalist delvingfor sources of information. The Autronic judgement of theECHR clarified that Article10alsoprotectsthe meansoftransmission andreception. Since crawlingisameansofreception, crawlingwouldbeprotected underArticle10and interference

20)Ibid. 21) TheUnited NationsUniversal Declaration of Human Rights, Adopted andproclaimed by General Assembly resolution 217 A (III) of 10 December 1948 (Article19: “Everyonehas therighttofreedom of opinion andexpression; this rightincludes freedom to hold opinionswithout interference andtoseek, receiveand impart information andideas through anymedia andregardless of frontiers.”); TheUnited NationsInternational CovenantonCivil andPolitical Rights, adopted andopened forsignature, ratification andaccession by General Assembly resolution 2200A (XXI) of 16 December 1966 (Article19para. 2: “Everyoneshall havethe righttofreedom of expression; this rightshall include freedom to seek, receiveand impart information andideas of all kinds,regardless of frontiers, either orally,inwritingorinprint, in theform of art, or through anyother media of his choice.”). 22) N. A. N. M. vanEijk, “Search Engines: Seek andYeShall Find? ThePosition of Search Engines in Law.” IRIS plus 2006-2, p. 5. Availableathttp://www.obs.coe.int/oea_publ/iris/iris_plus/iplus2_2006.pdf.en (Last viewed1July 2008). 23) Ibid.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 55

by public authorities wouldneed to satisfy thetest of Article10para. 2. Therefore, alaw which would restrict crawlingbysearch engines wouldneed to satisfy this test. On topofthat, theimportance of search engines forthe dissemination of information andideas andthe ability of thepublic to inform itself couldbeone of thedeterminingelements in decidingthe extenttowhich crawlingand indexing areprotected andthe question of whether such interferences areinaccordance with Article10§2. Thus, theprotectionagainst interferences with crawlingbysearch engineswouldbeenhanced because of theimportanceofsearch engines in our information society.

Similarly,the publicationofsearch resultsbyapubliclyaccessiblesearch engineand thequerying andreception of search resultsbythe end-user wouldbeprotected underArticle10. Thepublication of search resultsitself constitutes acommunication between thesearch engineprovider andthe end- users. From this perspective, thesearch enginenot only facilitatesaccess to information of third parties,but also offersaccess to information about thirdparties.Aninterestingquestion in this context is to what extentthe communications between asearch engineand itsusersare in factpublic in nature. Arguably,the queryingbysearch engineusersisnot apublic but aprivate communication. To theextentthat thesearch service is made publiclyavailable, thecommunication of search results to usersisofapublic nature.24

Based on thecase law of theECHR with regardtoArticle10and theinterpretation of the“duties andresponsibilities”inArticle 10 para. 2there areseveralthings worth mentioninginthe context of search engines.First, thepotential impact of themedium andthe natureofthe contentthat can be foundthrough asearch enginecouldbeofimportance here.Severaltimes thecourt has clarified that theparticular impact of audiovisual mediahas to be taken into account when consideringthe permissibility of interference by public authorities.For instance in its Jersild judgment,the Court concludedasfollows:

“In consideringthe “duties andresponsibilities” of ajournalist,the potential impact of the mediumconcerned is an importantfactorand it is commonly acknowledged that theaudiovisual media haveoften amuchmoreimmediate andpowerful effect than theprintmedia […]. The audiovisual media havemeansofconveyingthrough imagesmeanings which theprintmedia arenot abletoimpart.” 25

Therefore, it is possiblethat audiovisual search engines wouldreceiveless protection against interference than text-based search engines.

Secondly,asearch enginemightcarry specific duties andresponsibilities connected to itsrolein theinformation environmentand the(technical) natureofits service.Accordingtothe ECHR the duties andresponsibilities,underArticle10European Convention of Human Rights, also dependon thetechnical meansused forexpressionand dissemination. TheCourt has stated that:

“[…] whoeverexercises his freedom of expression undertakes “duties andresponsibilities” “thescope of which depends on his situation andthe technical meansheuses.The potential impact of those meansmustbetaken into accountwhenconsideringthe proportionality of the interference.The safeguardaffordedbyArticle10tojournalistsissubject, because of those very “duties andresponsibilities”, to theproviso that they providereliableinformation in accordance with theethics of journalism” 26

To which theCourt addsinamorerecentjudgment:

“These considerationsplay aparticularly important role nowadays, given theinfluence wieldedbythe media in contemporary society: notonly do they inform, they can also suggest by theway in which they presentthe information how it [theinformation] is to be assessed. In aworldinwhich theindividual is confronted with vast quantities of informationcirculated

24) See Schulz &Held2005. 25) ECHR 23 September 1994, §31(Jersild v. Denmark). 26) ECHR 14 June2007, §42(Hachette Filipacchi Associes v. France).

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 56 SEARCHINGFOR AUDIOVISUAL CONTENT

via traditionaland electronic media andinvolvinganever-growingnumber of players, monitoringcompliancewith journalistic ethics takes on addedimportance.”27

These last considerationsdoseem to resonate in thecontext of search engines but it remainsan open question whether andhow these considerationstranslate into thecontext of search engines and whether we will see acertain developmenttowardsabody of search engineethics andself-regulation. Undoubtedly,the impact of search engines on our information environmentisconsiderable. It would, however, be prematuretogivethe ultimate judgment on these issues.The nextsection will continue thediscussion by focusingonone specific aspect of search enginegovernance from theperspectiveof freedom of expression, namely thegovernance of thecomposition of theindex of search engines. Beforeaddressingthis question, however, it is useful to consider asimilar problem, that is to say the governance of Internet contentfilters, which has recently been dealt with in arecommendation by theCouncil of Europe’sCommittee of Ministers.

2.2. Internet Filters and Search Engines: aPossible Analogy?

Thereare manytypes of Internet contentfilters(“Internet filters”) andthey aredeployedina variety of circumstances.Ofinterest here is thetype of Internet filter whose purpose it is to limit the accessibility of Internet content. ThetechnicalitiesofInternet filterswill be left aside, but thereare filtersthat are–amongstother things –abletodetect audiovisual content(forexample, Googleis workingonfilteringcopyright protected content on Youtube based on a“fingerprinting” technique).28 Internet filter softwareaimstoblock certain categories of contentthat wouldbeaccessiblefor end- userswith an open, unfiltered Internet connection. Internet contentfilterscan be used andinstalled on Internet terminalsbyend-users. They can be deployedbyprivate actors on their networks, for instance by employers andthe owners of Internet cafés.They can be voluntarily deployedbyInternet Service Providers(hereinafter referred to as “ISPs”) on thenetwork levelasaservice to their customer, givingcustomers control overfiltering. They can be installed by public libraries that want to restrict access to harmful andillegal contentontheir internet access points.29 In agrowingnumber of countries in andoutsideofEurope, ISPs have agreed with publicauthoritiestofilter childpornography at thenetwork levelvoluntarily. In somejurisdictions, public authorities requireISPs by law to use filteringproductsatthe network level.30

Thefunction of Internet contentfiltersistoblock apart of theavailableinformation on theWorld WideWeb that wouldotherwise be accessible. As such, Internet filtersfall into thebroadercategory of selection intermediaries,aclass of intermediaries that select andfilter information travellingonline from information providerstoend-users.31 Examples of selection intermediaries include Internet contentfilters, search engines,recommendation services andISPs that block or filter material on the basis of content. Selection intermediariesfulfil an importantfunction in our informationenvironment. In this environment, which is characterized by abundance,selection intermediaries help end-usersto findand select theinformation they consider relevant or useful andexclude information that they are notwillingorallowedtoaccess,for instance because it is harmful or illegal.

Since search engines andInternet filtersare similar in thesense that they both makedecisions about thecollectionofinformation that is accessiblethrough their services,itisinterestingtosee what can be said about Internet filtersfromthe perspectiveoffreedom of expression. Takinginto accountthe differences between Internet filtersand search engines,this will provideauseful starting pointfor theanalysis of thefreedom of expression implicationsfor search engines.One difference is thatsearch engines aremorecomplicated, in thesense that they notonly makeaselectionof

27) ECHR 10 December 2007, §104 ( Stoll v. Switzerland ). 28) Foradiscussion see: Report of theGroup of Specialistsonhuman rights in theinformation society,and references,supra, note 3. 29) In theUnited States someofthe governmentfundingtopublic libraries has been made conditional on theinstallation of such filteringsoftware. See United States SupremeCourt, United States v. American Library Association,539 U.S. 194 (2003). 30)For acomprehensiveoverviewand discussion of global Internet filteringsee RonaldDeibert, John Palfrey et al. (ed.), Access Denied: ThePractice andPolicy of Global Internet Filtering, Cambridge MA: TheMIT Press,2007. 31) See Chandler,Jennifer A., “A RighttoReach an Audience: An Approach to Intermediary Bias on theInternet”.HofstraLaw Review, 2008. Availableathttp://ssrn.com/abstract=1021344

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 57

informationproviders, i.e., sources on theInternet that can be foundthrough their service,but in addition organize andorderthis selection of information providerswith their rankingalgorithms. Another difference is that Internet contentfiltersare normally placed between theend-user andthe Internet at theend-user’sterminal or at somelevel of thenetwork andthat switchingbetween Internet filtersisdifferentfromswitchingbetween search engines.Although thereisone dominant search engineonthe Webtoday in termsofuse,namely Google, thereare other similar search engines andnormally end-userscan easily switch between them. Finally,the purpose of Internet contentfilters is to prevent contentfrombeingaccessible, whilethe purpose of search engines is to help usersto access content. On afunctional level, these purposes arequite opposite.

Apart from thedifferences between search engines andInternet filters, however, thereremainsan importantsimilarity between thegovernance of Internet filtersand search engines,inparticular when we restrict thefocustothe governance of theinclusion andexclusion of information providersfrom asearch engine’sindex.Asnoted before, both types of services manage theaccessibility of information. Forinformation providersthe decision whether or nottobeindexedbysearch engines or includedonablockinglist of filtershas similar effectsontheir availability to thepublic. Forend- users, thecomposition of theindex of search engines presents similar issues of access to information, transparency andcontrol as in thecase of Internet contentfilters.

Notably,search engines already usefilterswithin their services.The image search functionsofthe major search engines all haveanoptional filter that an end-user can turn on or off. Google, for instance,offersthe option of SafeSearch in itssearch preferences,allowingend-userstoopt fora filtering of explicit imagesfromimage search resultsand afilteringofall search results, includingtext based search.32 Thepreferences forfilteringare saved through theplacementofacookie on theend- user’sbrowser.Googlehas also started to offer filteringfor corporate clients.33 As aresult of the problem of search enginespam (spamdexing) search engines mightalsouse filterstofilter pagesthat havenoother purpose than to manipulate search resultsand receiveanunfairly high ranking. In the context of unlawful search results, arecurringquestion is whether search engines shoulduse filters to prevent unlawful resultsfromshowingup, or be obliged to usesuchfiltersbylaw or regulation.

These findings areinlinewith thefindings of theCouncil of Europe’sGroup of SpecialistsonHuman Rights in theInformation Society,which prepared areport on Internet filtersfor theCommittee of Ministers.34 TheReport notes that the“filtering effectsofsearch engines can […] be said to be more indirect than forfilteringproducts”, “[…] filteringproductsaim to protect usersagainst illegal, harmful or otherwise unwanted content, search engines aim to providethe user with rapid access to informationthat is most likely to fulfil theuser’sinformation needs as expressed in thesearch terms employed.” TheReport also notes theuse of contentfiltersbysearch engines andassertsthat “the findings in thereport apply to search engineproviderstothe extentthat they areusersoffiltering software.”The Report states that “it is likely that at least someofthe best practice guidelines identified in this report couldform thebasis of similar guidelines forsearch engines.”35

From theabove it is clear that although thereare someimportantdifferences,anenquiry into Internet filtersand theimplicationsoffreedom of expression forInternet filtersisjustified.This enquiry will be made alongthe lines of therecentRecommendation on Internet filtersand freedom of expression of theCouncil of Europe’sCommittee of Ministers.

2.3.Recommendation on InternetFilters andFreedom of Expression

On 26 March2008, theCouncil of Europe’sCommitteeofMinisters adopted aRecommendation (CM/Rec(2008)6) on measures to promote therespect forfreedom of expression andinformation with regardtoInternet filters.36 TheRecommendation andthe underlyingReport acknowledge both the

32) GoogleSafeSearch filteringpreferences,viewed1July 2008, availableat http://www.google.com/support/bin/static.py?page=searchguides.html&ctx=preferences 33) ‘Web Security forEnterprise’, Cnet News.com, 8May 2008. Availableat http://www.news.com/8301-10784_3-9938812-7.html?tag=nefd.top (Last viewed1July 2008). 34) Report of theGroup of Specialistsonhuman rights in theinformation society,and references,supra, note 3. 35) Ibid. 36) Supra, note 2.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 58 SEARCHINGFOR AUDIOVISUAL CONTENT

legitimate function of Internet filtersand thewaysinwhich Internet filterscan impact on freedom of expression andinformation.Itexplicitly addresses someofthe requirements of Article10ofthe European Convention on Human Rights in this context,someofwhichrelate to theinterestsof information providersand end-users. TheRecommendation callsupon theMember States of the European Convention to takemeasures with regardtoInternet filtersinlinewith aset of guidelines promoting user notification, user awareness anduser control of Internet filters, andthe accountability of theprivate andpublic parties involved. Theguidelines arecategorized into three sections. Section Igives guidelines about “usingand controllingInternet filtersinordertofully exercise andenjoy the right to freedomofexpression andinformation”.Section II dealswith“appropriate filteringfor children andyoungpeople” andsection III is about the“useand application of Internet filtersbythe public andprivate sector”.

TheRecommendation differentiates between mandatory filteringand theuse of Internet filtersby public entities,suchaspubliclibraries andschools, on theone hand, andtheir usebyprivate entities, such as enterprisesinthe context of internet accessinthe workplace,onthe other.The Recommendationdoes notlimit itself to theimplicationsoffreedom of expression on Internet filters in thepublic sector or legally prescribed filtering, but it also containsspecific guidelines on theuse of filtersinthe private sector.With regardtothe useand application of Internet filtersbypublic entities or mandatory filtering, theRecommendation qualifies theuse of Internet filtersasan interference with therighttofreedom of expression andmakes thetest of Article10§2moreexplicit. With regardtothe useofInternet filtersbyprivate entities,the Recommendation andthe underlying Report refer to positiveobligationsupon Member States as guarantors of pluralism andinmaintaining andenhancinglegal andpractical measures to prevent private censorship.

In section III of theguidelines,the Recommendation specificallyaddresses thedemands of Article 10 ECHR with regardtothe legal requirements relatingtoInternet filters. It callsupon theMember States to refrain from filteringInternet contentinelectronic communicationsnetworksoperated by public actors forreasonsother than those laid downinArticle10§2.Similarly,itstates that Member States areonly allowedtointroduce nationwidegeneral blockingorfilteringmeasures if theconditions of theConvention, in particular Article10§2, arefulfilled. Accordingtothe guidelines in the Recommendation, this meansthat thefilteringhas to concern “specific andclearlyidentifiable content”, “a competentnational authority shouldhavetaken adecision on itsillegality” and“there shouldbeanopportunity to havethis decision reviewedbyanindependent andimpartial tribunalor regulatory body,inaccordance with therequirements of Article6ofthe European Convention on Human Rights”. 37 Furthermore, theguidelines stipulate that Member States havetoensurethat there is an evaluation of theproportionality of filtersbeforeand duringtheir implementation in termsof their possibleeffectsonthe unreasonableblockingofcontent.

Severalofthe guidelines –insectionsIand III –explicitlyrelate to thegovernance of theinclusion andexclusion of information providersbyInternet filtersand therights of end-usersand information providersinthis respect. First of all, theguidelines providethat end-users, whereappropriate,must be abletocontrol theleveloffiltering. Theguidelines further stipulate that end-usersshouldhave thepossibility of challengingthe blockingorfilteringofcontentand seekingclarificationsand remedies.With respect to theend-user’sability,where appropriate,toactivate anddeactivate filters andtobeassisted in varyingthe leveloffilteringinoperation, theguidelines call upon theMember States to ensure, in cooperation with theprivate sector andcivil society,the existence of anumber of more detailed guarantees.Itisstated that end-usersshouldreceiveguidance regardingthe manual overridingofanactivated filter,including, more specifically,whomtocontact when it appearsthat contenthas been unreasonablyblocked, as well as thereasonsthat may allow afilter to be overridden foraspecific type of contentorUniform Resource Locator (URL). Furthermore, theRecommendation states that contentthat is filtered by mistakeorbecause of an error has to be accessible without undue

37) Article6European Convention of Human Rights, first paragraph: “In thedetermination of his civil rights andobligations or of anycriminal charge against him, everyoneisentitled to afair andpublic hearingwithin areasonabletimebyan independent andimpartial tribunal established by law.Judgement shall be pronounced publicly by thepress andpublic may be excludedfromall or part of thetrial in theinterest of morals, public orderornational security in ademocratic society, wherethe interestsofjuveniles or theprotection of theprivate life of theparties so require, or theextentstrictly necessary in theopinion of thecourt in special circumstances wherepublicity wouldprejudice theinterestsofjustice.”

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 59

difficulty andwithin areasonabletime. With regardtothe useand application of Internet filtersby thepublic sector,Member States havetoavoid theuniversal andgeneral blockingofoffensiveor harmful contentfor userswho arenot part of thegroup forwhich afilter has been activated to protect, andofillegal content foruserswho justifiably demonstrate alegitimate interest or need to accesssuch contentinexceptional circumstances,particularly forresearch purposes.Member States shouldalso providefor effectiveand readily accessiblemeansofrecourse andremedy,includingsuspension of filtersincases whereusersand/or authors of content claim that contenthas been blocked unreasonably.

2.4.Relevance forthe Governance of Search

TheCouncil of Europe Recommendation on Internet filteringand theunderlying expert Report provideaninterestingstartingpointfor theanalysis of theimplicationsoffreedom of expression for thegovernance of theinclusion andexclusion of information providersbysearch engines in theform of theissues surroundingthe collection of information providersthat can be accessed through a particular search service.

Followingthe Recommendation, therewouldbeadifference between thefreedom of expression implicationswith regardtolegal obligationsonsearch engines to limit access to certain information providersand theprovision of search engines by thepublic sector,onthe onehand, andthe voluntary limitationsonsuchaccess by search engines in theprivate sector,onthe other hand. Alegal obligation on search engines to filter out certain information providersfromthe search index wouldhaveto respect thetest of Article10para. 2. Followingthe guidelines,this wouldmean that theremovalof informationprovidersfromthe search engine’sindex has to concern specific andclearly identifiable content. Imposinggeneral preventivefilteringobligationsonsearch engines with regardtoillegalor infringingcontentcouldbeproblematic from this pointofview.38 Another interestingpointisthe requirementthat, in cases of mandatory filtering,acompetentnational authority shoulddecideon theillegality of theinformation andanopportunity to havethis decision reviewedbyanindependent andimpartialtribunal or regulatory body shouldbeoffered.

Theguidelines on Internet filtersaim to guaranteeacertain degree of user controloverand transparency of theapplication of Internet filtersand it is interestingtosee how this playsout in the context of search engines.Afirst problem that arises here is that asearch enginecan notbeturned on or offlikeanInternet filter.Inthe context of search engines,the Internet end-user has other optionsofredress.The end-user can, forinstance,decidetouse another search engine, something which is only differentifthe choice between search engines is restricted, which is normally notthe case.The end-user can also access theinformation directly,which requires that he or sheisaware of thelocation of theinformation on theInternet. In manycases,the useofasearch engineimpliesthat theuser is notaware of theinformation’slocation. Theadditional filtersofsearch resultsoffered by search engines,suchasthe image filtersdiscussed above,already giveuserscontrol overfilteringin theadvanced settings. These possibilitiescouldfurther increase due to technological developments.

With regardtothe effectiveand readily accessiblemeansofrecourse andremedy in thecase of an information provider notincludedinasearch engine’sindex,there areseveralpoints worth mentioning. First, searchengines partly rely on thesuggestionsofthirdparties, especially information providers, to include sources in their index.All major Websearch engines haveonlineformsand tools to submitwebsites forcrawling. Notably,the control that information providershaveoverbeing indexedasaresult of exclusion protocolslike robots.txt,implies less control forend-usersoverthe composition of theindex.Byrespectinginstructionslike robots.txt,the major search engines favour theinterestsofinformationprovidersinhavingcontrol overthe accessibility of their contentthrough asearch service,overthe end-users’ interestsinhavingsuchcontrol. Thereare severalreasonswhy generalsearch engines havecometorespect these protocols. Asystematic discussion of these reasons goes beyond thescope of this contribution. Oneofthe main reasonsisthat search engines on theWeb

38)This seemstofit with Article15ofthe DirectiveonElectronic Commerce (2000/31/EC). However, oneshouldnote that Article15ofthe Directiveonelectronic commerce does notapply to search engines,ascan be seen in Article21ofthis Directive. Infra, section 3.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 60 SEARCHINGFOR AUDIOVISUAL CONTENT

do notrestrict themselves to meta-data, but storelarge partsoreventhe complete contents of websites on their servers. By respectingthe exclusion protocols, search engines evade some–but surely notall –ofthe potential conflicts with onlinepublishers.39 In itsOpinion on search engines,the Article29 WorkingParty recommends that search engines respect these protocolsbecause they couldbeused to prevent personal data beingindexed.40

Thesituation mightbedifferentwhenasearch enginehas purposely removedaspecific website from itsindex.One of thereasonsfor such aremovalcouldbethat awebsite is usingillegitimate techniques to influenceits ranking. Another reasoncould be thatthe search engineprovider considers theremovalofcertain references from itsindex as itsparticular responsibility. 41 Yetanother reason couldbethat s/heisrequired by law or orderedbyajudge to do so, because thesearch resultsare unlawful, forinstancebecause they contain or link to illegal information.42 This leadstothe interestingquestion of thecircumstances in which search engineprovidersfindthemselves undersuch alegal obligation to remove references from their index.This legal question about theliability of search engineprovidersfor references to thirdparty contentisclosely related to thequestion of the inclusion andexclusion of information providersinasearch platform.

3. Liability of Search Engine Providers for Unlawful References

Oneofthe reasonswhy search engineprovidersremove references to information providersfrom their index is that it wouldbeunlawfulfor them to show these references to their users(forinstance because they contain or provideaccess to defamatory contentorlink to unauthorised copies of copyrightprotected material). Clearly,the liability of search engineprovidersfor unlawful references can haveasignificant impact on thecomposition of theindex of search engines andthereby on the free flow of information facilitated by search engines.

Thequestion about theliability of search engineprovidersfor thirdparty contentisaquestion about onlineintermediaryliability. 43 In theEuropean Union, theliability of onlineintermediaries for thirdparty contenthas been specifically dealt with by theDirectiveonElectronic Commerce. 44 This Directivecontainsspecial provisionsfor theliability of information society services actingasonline intermediaries in Articles 12-15. TheDirectivedefines three categories of intermediaries,namely “mereconduit” (Article12), “caching” (Article 13), and“hosting” (Article14).45 Foreachofthese

39) Oneofthe most notableconflictsbetween publishersand search engines is theBelgium case Copiepresse et al. v. Googleet al. -Court of First Instance 13 February 2007, Computer Law ReviewInternational, 2007-2, pp.48-50, Summary &Comment by Croux/Sasserath. Thecase is beingappealed. See “Googleloses copyrightcase launched by Belgian newspapers”’, International Herald Tribune, 13 February 2007. Availableat http://www.iht.com/articles/ap/2007/02/13/business/EU-FIN-Belgium-Google-vs-Newspapers.php (Last viewed1July 2008). 40) Supra, note 10, p.14 (“Itisessential that search engineprovidersrespect opt-outsexpressed by website e ditors.”). 41) In Germany, themajor search engines haveentered into aself-regulatory agreementtoremove certain references from their index “toimprove consumer protection andtoprotect children andyoungpeoplefrominappropriate contentwhenusing search engines”.See SubcodeofConduct forSearch EngineProvidersofthe Association of Voluntary Self-Regulating Multimedia Service Providers(Freiwillige Selbstkontrolle Multimedia-Diensteanbieter –FSM) (VK-S), 21 April 2004. Available at http://www.fsm.de/en/Subcode_of_Conduct_for_Search_Engine_Providers(Last viewed1July 2008). Foradiscussion of theagreementsee WolfgangSchulz &Thorsten Held, Der Index auf dem Index? Selbstzensur undZensur bei Suchmaschinen, in Marcel Machill &MarkusBeiler (ed.), Die Machtder Suchmaschinen /The PowerofSearch Engines, Köln: Halem, 2007, pp. 71-87. 42) Foracollection of German andinternational case law on search engineliability forunlawful references see thecollection by Stephan Ott at Links&Law.Availableat http://www.linksandlaw.com/legalresources-publicationsenglish.htm. Last viewed1July 2008. 43) Foradiscussion of onlineintermediary liability see K.J. Koelman, Liability foron-lineintermediaries,Imprimatur,IViR, Amsterdam, 1997; Rosa Julia-Barceló: On-LineIntermediary Liability Issues: ComparingEUand US Legal Frameworks, European Intellectual Property Review, 2000, Vol. 22 (3), 2000, P. 106-119; K.J. Koelman, OnlineIntermediary Liability,in P.B. Hugenholtz (ed.), Copyrightand Electronic Commerce,Information Law Series-8, Kluwer,2000, pp. 7-58. 44) Directive2000/31/EC of theEuropean Parliamentand of theCouncil of 8June2000 on certain legal aspectsofinformation society services,inparticular electronic commerce,inthe Internal Market (DirectiveonElectronic Commerce), OJ L178, 17 July 2000, p.1. 45) Generally speaking, “Mereconduit” applies to transmission of thirdparty content, “caching” applies to thetemporary storage of thirdparty contentfor technical reasonsand “hosting” applies to thestorage of thirdparty contentonservers in general.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 61

categories it contains ahorizontal46 conditional liabilityexemption,often called a“safe harbour”.For instance,inthe case of theprovidersofhostingservices,the condition tied to thesafe harbour is that “(a) theprovider does nothaveactual knowledge of illegal activity or information and, as regards claimsfor damages, is notaware of facts or circumstancesfromwhichthe illegal activity or information is apparent; or (b) theprovider,upon obtainingsuchknowledge or awareness,actsexpeditiously to remove or to disableaccess to theinformation”.The liability exemptionsexplicitly leaveopen the possibility foracourt or administrative authority to requirethe provider to terminate or prevent an infringement.The exemptions do notaffect thelawfulness of theprocessingofinformation by providersofany of these types of intermediary services.The lawfulness has to be determined by applyingthe relevant lawsofthe Member States.Finally,the Directiveproscribes generalobligations on theprovidersofthese services to monitor theinformation that they transmit or storeortoseek facts or circumstances indicatingillegal activity (Article15).

TheDirectivefocuses on thelegal obligationsofinformation society services,which include search engines. 47 But Article21onthe evaluation by theEuropean Commission of theDirectiveshows that search engines (location tool services) andhyperlinksare notcoveredbythe intermediary liability regimeinArticles 12-15 of theDirective. Article21para. 2provides that “[...] in examiningthe need foranadaptation of this Directive, thereport shall in particular analyse theneed forproposals concerningthe liability of providersofhyperlinksand location tool services,[…]”.The exclusion of search engines from onlineintermediary liability regulation at theEUlevelhas ledtoadivergent treatmentofsearch engines by legislatures andthe judiciary in thevariousMember States.Because theDirectiveonElectronic Commerce does notprovidefor full harmonization of intermediary liability exemptions, theMember States couldstill regulate theliability of search engineprovidersand other intermediaries notcoveredbythe Directive. Accordingtothe European Commission’sfirst report on theapplication of theDirectiveonElectronicCommerce,this option has been used by afew Member States,includingSpain, Portugal, Austria andEEA-member Liechtenstein.48 Thenew EU Member States Hungary andPolandalsoextendthe limitation on liability forprovidersofhostingservices to informationlocation tools. Thesafe harboursfor search engineprovidersinSpain, Portugal, Hungary andPolandresemblethe safe harbour forhostingproviders(Article14ofthe Directive). Thesafe harboursfor search engineprovidersinAustria andLiechtenstein resemblethe safe harbour formere conduit (Article 12 of theDirective).

Theresult of theabsence of aEuropean safe harbour forsearch engineprovidersand thelackof harmonization is apatchwork of lawsinEurope governingthe liability of search providersfor unlawful references.The question is whether this situation, which impliesacertain degree of legal uncertainty, is favourablefromthe perspectiveofthe facilitatingroleofsearch engineservices with regardtothe free flow of information. It goes without sayingthat extensiveduties of carefor search engines can result in achillingeffect on legitimate information flows andperhaps theabsence of effectivesearch engines altogether. 49 Ideally,the law wouldset theproper incentives forsearch engineprovidersto remove illegal resultsonthe onehand, without chillingthe legitimate information flows facilitated by search engines on theother.

Thereare someexamples of national courtstakingaccountofthe importance of search engines for therealization of freedom of expression andinformation. ThePaperboyjudgment of theGerman Bundesgerichtshof affirmed thepermissibility of theuse of hyperlink technologies by crawler-based search engines in adecision referringboth to thesocial utility of search engines andthe fundamental rightoffreedom of expression andinformation. 50 In amorerecentruling, theOberlandesgericht

46) Thismeansthat theliability limitationsapply both to civil andcriminal liability,and coverliability forcopyright, trademark, defamatory statements, hate speech, etc. 47) As can be seen in recital 18 of theDirective: “[…] information society services arenot solely restricted to services giving rise to on-linecontractingbut also, in so farasthey representaneconomic activity,extendtoservices which arenot remunerated by those whoreceivethem, such as those offeringon-lineinformationorcommercial communications, or those providingtoolsallowingfor search, access andretrievalofdata; […].”, ibid, p. 6. 48) See “First report on theapplication of Directive2000/31/EC of theEuropean Parliamentand of theCouncil of 8June2000 on certain legal aspectsofinformation society services,inparticular electronic commerce,inthe Internal Market, 21 November 2003, COM(03)0702”, p. 13. 49) Foradiscussion see Koelman in P.B. Hugenholtz (ed.), op. cit.,pp. 40-44 50) Bundesgerichtshof 17 July 2003, IZR259/00.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 62 SEARCHING FOR AUDIOVISUAL CONTENT

Hamburgbuilds on this judgment,concludingthat thenormal liabilitystandardfor publications is too strict in thecase of thepublicationofreferences by search engines andrequiringanexception forthe case of search engines. 51 Thecourt concludesthat theliability of search engineprovidersfor the possibly unlawful contentoftheir references,which is theresult of theautomatic reproductionof unlawful informationfromthe billionsofwebsites on theWeb,shouldbelowered, because of the significanceofsearch engines forfreedom of expression andinformation.

At this pointintimeitisunclear whether somekindofsafe harbour forthe providersofsearch engines will be adoptedatthe European level. Article21ofthe Directiveinstructsthe European Commission to conduct abiannual report on theapplicationofthe Directivewhich has to contain an analysis of theneed forproposalsconcerningthe liability of providersofhyperlinksand location tool services.In thefirst report theCommission concludedthere wasnoreason to amendthe existingintermediary liability rules with regardtosearch engines.The Commission did note thediverginglegislativechoices andwrote thefollowingabout legal developments with regardtosearch engineliability:

“Itisencouragingthat recentcase-law in theMember States recognizes theimportance of linking andsearch engines to thefunctioningofthe internet. In general, this case-law appears to be in line with theInternal Market objectivetoensurethe provision of basic intermediaryservices,which promotes thedevelopmentofthe internet ande-commerce.Consequently,this case-lawdoes not appear to giverise to anyInternal Market concerns.” 52

TheCommission subsequently placed theissue on theagendaofthe nextreview, which, according to Article21ofthe Directive, shouldhavebeen finalizedin2005. In thesecondreport special attention will be paid to thequestion of intermediary liability,includingthe question of liability of search engine providersfor thirdparty content. In itsreview, theEuropean Commission will also havetoaddress the perspective of freedom of expression. Whereas theprimary rationaleofthe Directivelies in the strengtheningofthe Internal Market forE-commerce andrelated services,the Directiveacknowledges that the“free movement of information society services can in manycases be aspecific reflection in Community law of amoregeneral principle, namelyfreedom of expression as enshrined in Article10(1) of theConvention forthe Protection of Human Rights andFundamental Freedoms[…].”53

4. Conclusion

Search engines havebecomecentral in theonlineinformation environmentand play acrucial facilitatingrolewith regardtoonlinepublic communications. They mediate between theinterestsof information providersand end-usersthat relate to thefreeflow of information onlineand havea significant impact on access to information, includingaudiovisual content. Because of their facilitating role in thefreeflow of information,morespecifically,the ability of information providerstoreach an audience andthe interest of end-usersinsearchingfor thecontentoftheir choice,this contribution has focused on theimplications of thefundamental righttofreedom of expression andinformation as enshrined in Article10ofthe European Convention on Human Rights forthe governance of search engines.Someofthe generalquestionsabout these implicationshavebeen examined, in particular theimplicationsonthe governance of theinclusion andexclusion of information providersfroma search engine’sindex.The recentCouncil of Europe Recommendation on Internet filtershas been showntobeaninterestingstartingpointfor studyingthese issues.Asitisarelated topic,the liability of search engines forthirdparty contentatthe EU levelhas also been discussed. This liabilityregime andthe lack of legal certainty couldbeconsidered sub-optimal from thepointofviewoffreedom of expression andinformation. Allinall, thelegal andregulatorychallengeswithregardtosearch engines can be said to be significant. Search engines aremultifaceted andcomplex andsomeofthe questionsraised in this contribution will only be answered overtime.

51) OLG Hamburg, 20 February 200 –AZ.: 7U126/06. 52) First report on theapplication of Directive2000/31/EC of theEuropean Parliamentand of theCouncil of 8June2000 on certain legal aspectsofinformation society services,inparticular electronic commerce,inthe Internal Market, 21 November 2003, COM(03)0702, p. 13. Availableathttp://ec.europa.eu/internal_market/e-commerce/directive_en.htm 53) Directive2000/31/EC, recital 9.

©2008, European AudiovisualObservatory, Strasbourg (France)

SEARCHING FOR AUDIOVISUAL CONTENT 65

In Search of the Audiovisual Search Tools in the EU RegulatoryFrameworks

Peggy Valcke Interdisciplinary Centrefor Law &ICT of K.U. Leuven K.U. Brussel

Introduction

In his contribution to theIRIS plus 2006-02, Nico vanEijk noted that “thesearch enginedoes not haveaplace in law”. 1 He attributes thelimited attention search engines havereceived so farfrom lawmakers 2 to thefactthat they findthemselves in thegrey zonebetween electronic communications (transmission) andbroadcasting(content).Meanwhile, both the“Televisionwithout Frontiers” Directive 3 (see infra1.3.) –the cornerstoneofthe EU’sbroadcastingregulation –and theElectronic Communications Directives 4 (see infra1.2.) havebeen (or arecurrently) subjecttoathorough revision. Has this changed(or is it changing) somethinginthe legal treatmentofsearch engines,orinmore generalterms, contentnavigationtools? It is thepurpose of this contributiontoassess theEU regulatoryenvironment andcheck to what extentthe respectiveframeworks(could) apply to audiovisual search or contentnavigation tools.5 In asecondpart, somepreliminary ideas on “how to fill thegaps” will be formulated.

Warning Notice

Iwouldliketostart my contribution with thefollowingwarningnotice: what makes this analysis extremely complicated is theexistence of awidevariety of audiovisual search tools. Somesearch engines orderaudiovisual material on theWeb,aspart of amoregeneral or horizontal search site (includingalsotext, images, locally stored files,etc.,suchasExalead 6 )orasocial bookmarkingservice

*) Peggy Valcke lectures in media andcommunicationslaw at K.U.Leuven andK.U.Brussel andhas been visitingprofessor at Central European University in Budapest in 2006-2007; sheworksasapostdoctoral researcher at theInterdisciplinary Centrefor Law & ICTofK.U.Leuven (www.icri.be), which is afoundingmember of theFlemish Institute forBroadbandTechnology (www.ibbt.be). 1) VanEijk, Nico. 2006. Search engines: Seek andyeshall find? Theposition of search engines in law.IRIS plus 2006-2; Strasbourg: European Audiovisual Observatory,availableathttp://www.obs.coe.int/oea_publ/iris/iris_plus/iplus2_2006.pdf 2) Though notfromcourts, as is demonstrated by numerous court rulings relatingtosearch engines in anumber of law domains (intellectual property,unfair trade practices,data protection…). 3) Council Directive89/552/EEC of 3October 1989 on thecoordination of certain provisionslaid downbylaw,regulation or administrativeaction in Member States concerningthe pursuit of television broadcastingactivities,[1989] OJ L298/23, as amendedbyDirective97/36/EC, [1997] OJ L202/60. 4) In short: Directive2002/21/EC (Framework Directive) [2002] OJ L108/33; Directive2002/19/EC (Access Directive) [2002] OJ L108/7; Directive2002/20/EC (Authorisation Directive) [2002] OJ L108/21; Directive2002/22/EC (Universal Service Directive) [2002]OJL108/51; Directive 2002/58/EC (ePrivacy Directive)[2002] OJ L201/37. Thesedirectives areaccompanied by Commission Directive2002/77/EC (“Competition Directive” [2002] OJ L249/21) andanumber of regulations, decisionsand recommendations(formoreinformation, see: http://ec.europa.eu/information_society/policy/ecomm/current/index_en.htm). 5) Audiovisual search toolsand contentnavigation toolsare used as synonyms in this article. 6) http://www.exalead.com 7) http://del.icio.us

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 66 SEARCHING FOR AUDIOVISUAL CONTENT

(e.g., del.icio.us 7 ). Othersare dedicated to servingupvideos, sometimes combiningthe search function with (free) hostingcapacity foruserswho want to uploadtheir ownvideos (e.g., GoogleVideo8 ). In somecases,the focusisnolongeronthe searchfunction, butonthe hostingand distribution functions, as is thecase with themushroomingonlinevideo portals. Of these,somefocusonuser- generated content(likeYouTube,Dailymotion, GarageTV or BlipTV)9 andothersonprofessional channels, shows, movies andmusic videos (likeJoost).10 Other portalscombineboth types of content (likeGotuit Media11)and/or offer userscommunicationand ratingtoolstosharetheir experience with friends whilethey arewatching(e.g., instantmessagingsoftwareorablogspace).12

At theother endofthe spectrum,wehavethe traditional (edited) electronic programmeguides (EPGs), used in thecontext of awalled-gardendigital television environment. Nowthat established broadcastersare expandingtheir territory from thetraditional television platformstothe Internet, we findmoreand more of their programmelistings andguides onlineaswell, therebyallowingusersto check what is on achannel at that moment,that night, thenextday,etc. (oneofthe manyexamples is theone offered by theBBC13). Similar onlinelistings arealsooffered by TV guidemagazines that reproduce their printproduct on theInternet –often with additional featuresand usually forfree(i.e., advertisingbased, without subscription from theuser; e.g., TV Guide 14)–orbynew entertainment portals(likethe Irish entertainment.ie website15). It is standardpractice forthese to presentagrid of television programmes that areshownonthe variouschannelsatacertain moment,thereby offering usersthe option of adjustinggrid preferences (e.g., to change theorderofthe channels), searching foraspecific programmeand checkingadditional programmeinformation. Someevenoffer an online scheduling service,i.e., atool to schedulerecordings of programmes on your digital video recorderby simply clickingthe programmeinthe onlinegrid (e.g., TVGuidewhich has partnered with TiVo® DVR forthat purpose). Contrary to theelectronic programmeguides on thetraditionaltelevisionplatforms, they do notoffer16 thepossibility of accessing thechannel or programmedirectly via theonlinegrid. At least that is,not yet; in thelightofthe growingconvergencebetween theInternet andtelevision, it is only amatter of timebeforeonlinetelevision listings will enablesuchaccessingand zappingas well, whileatthe sametime, “traditional” electronic programmeguides will include similar features enablingaccess andsearch of theuser-generated contentavailableonthe Internet.

In other words, thereisanalmost endless continuum of search services andproductstoassist the user in findinghis/her waythrough the“audiovisual abundance”.But that common feature–the search function –does notautomatically makethese services susceptibletoequal legal treatment(it is notbecause avehiclehas wheelsthat youneed adriver’slicense to steer it). It makes ahuge difference whether thesearch function is part of abroadercontentoffering(although in that case the provider may stillbelargely left “untouched” when it looksmorelikethe electronic equivalentofprint or text products), part of abroaderaudiovisual contentoffering(in which case thesearch service may be subject to thesamerules applicable to theaudiovisual media service which it accompanies–unless thecontentservice does notfulfil all criteria of thelegal definition of audiovisual media service; infra ) or part of transmission services (likethe EPGs on traditional television platforms, which areoffered by conditional access providers; infra ).

It is notthe aim of this articletooffer acomplete legal analysis of all possibleformsofaudiovisual search tools. Theintention is to pinpointrelevant provisionsinthe existingEUregulatory frameworks (i.e., provisionscontainingareference to search engines or electronic programmeguides), analyse the conceptsthat they use, as well as thescope of therules they impose.

8) http://video.google.com/ 9) http://youtube.com/ http://www.dailymotion.com/ http://www.garagetv.be/ http://blip.tv 10) http://www.joost.com 11) http://www.gotuit.com 12) Foraninterestinganalysis of these new portalsfromthe perspectiveofcontentregulation, see: Ariño, Monica. 2007. Content Regulation andNew Media: ACase Study of OnlineVideo Portals. Communicationsand Strategies (66):115 –135. 13) http://www.bbc.co.uk/cgi-perl/whatson/search/grid.cgi?MEDIUM=tv 14) http://www.tvguide.com/Listings/default.aspx (a similar exampletaken from theFlemish context is http://www.tvgids.be/) 15) http://entertainment.ie/tv/whats-on-now.asp 16) At least those which we havechecked.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHING FOR AUDIOVISUALCONTENT 67

1. Content Navigation Toolsunder EC Regulation for the Info-communicationsSector 1.1. In SearchofSearch Tools…

Audiovisual search toolshavetocomply with aseries of generalEUlegal provisions, includingthe competition rules in Articles 81 and82EC-Treaty,the rules relatingtounfair commercial practices in Directive2005/29/EC,17 theadvertisingrules contained in, forinstance,Directives 2006/114/EC (misleadingand comparative advertising),18 2003/33/EC (tobacco advertising)19 and2001/83/EC (medicinalproducts).20 These instruments will, however, notbeanalysed further here.

This contribution aimstofocusonthe rules specifically adopted forthe sectorsofbroadcastingand audiovisual services,communications, andinformation –inshort the“info-communicationssector”. EC sector-specific regulation forthe info-communicationssector is based on ademarcation between transmission andcontent. Theelectronic communications directives,adopted in 2002,cover all transmission networksand services (as aresult of convergence), but notthe contentofservices delivered oversuch networks(such as broadcastingcontent, financial services andcertain information society services21). Servicesatthe contentlayer,consistingofthe editing anddeliveryofaudiovisual contentor information, areregulated by the“Television without Frontiers” Directive 22 (recently amendedand renamed theAudiovisual Media Services Directive; infra )and by theE-Commerce Directive.23 Inextricable linksbetween transmission andcontent, however, renderitimpossibletoregulate onecompletely separately from theother or to exclude hybrid or “intermediary” measures.The electronic communications framework, forinstance,containsrules that apply to transmission networksorassociated facilities for content-related reasons(likeculturaldiversity) –think of themustcarry-provisionsinArticle31 Universal Service Directiveorthe rules relatingtobottleneck facilities in digital television ( infra ).

1.2. …inthe Electronic Communications Directives

2002 Regulatory Framework

As mentioned earlier, the2002 Electronic CommunicationsDirectives apply to transmission networks andservices,aswell as theso-called “associatedfacilities”. Since it is obviousthat search toolsdonot constitute anetwork of their own, we canconfine our analysis to thenotionsofelectronic communica- tionsservices andassociated facilities.

Theformer aredefined by Article2(c) Framework Directiveas“services normally provided for remuneration which consist wholly or mainly in theconveyanceofsignalsonelectronic communicationsnetworks, includingtelecommunicationsservices andtransmission services in networksused forbroadcasting”.The definition explicitly excludes“services providing, or exercising editorial control over, contenttransmitted usingelectronic communicationsnetworksand services” as well as “Information Society services,asdefined in Article1ofDirective98/34/EC, which do not consist wholly or mainly in theconveyanceofsignalsonelectronic communicationsnetworks.”The exceptionsseem rather to be written with broadcastingand on-demandaudiovisual services in mind, excludingthose which edit that type of service,aswell as those which provide(or distribute) such

17)Directive2005/29/EC of theEuropean Parliamentand of theCouncilof11May 2005 concerningunfair business-to-consumer commercial practices in theinternal market andamendingCouncil Directive84/450/EEC, Directives 97/7/EC, 98/27/EC and 2002/65/EC of theEuropean Parliamentand of theCouncil andRegulation (EC) No. 2006/2004 of theEuropean Parliament andofthe Council (“Unfair Commercial Practices Directive”), [2005] OJ L149/22. 18) Directive2006/114/EC of theEuropean Parliamentand of theCouncil of 12 December 2006 concerningmisleadingand comparativeadvertising, [2006] OJ L376/21. 19) Directive2003/33/EC of theEuropean Parliamentand of theCouncil of 26 May 2003 on theapproximation of thelaws, regulationsand administrativeprovisionsofthe Member States relatingtothe advertisingand sponsorship of tobacco products, [2003] OJ L152/16. 20) Directive2001/83/EC of theEuropean Parliamentand of theCouncil of 6November 2001 on theCommunity coderelating to medicinal productsfor human use, [2001] OJ L311/67, as amendedbyDirective2004/27/EC of 31 March 2004, [2004] OJ L136/34. 21) Cf. Recital 5Framework Directive. 22) Supra ,footnote 3and infra ,footnote 39. 23) Infra ,footnote 47.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 68 SEARCHING FOR AUDIOVISUAL CONTENT

contenttothe public.Neither theediting, northe provision or distribution of contentvia theInternet or more traditional platforms, likecabletelevision, is considered an electronic communications service. 24 Search toolsact as directory services,facilitating access to thirdparty information, but they do notedit that information or contentthemselves,and it also seemsdifficult to argue that they providethe contentthemselves.Atthe sametime, it cannot be denied that theservice they offer does notconsist wholly or mainly of theconveyanceofsignals. Thecoreoftheir service is notmerely to bringsignalsfrompointAtopointB,but to assist theuser in findingthe informationhe/sheis lookingfor by providingreferences to thirdparty information, ranked, accordingtorelevance, in the lightofsearch termsand/or past user preferences.So, eventhough search toolsmay show some aspects of routingservices,their addedvalue lies in their functionalities as marketinginstruments and content-related services andwill thereforefall outsidethe scope of electronic communicationsservices in thesense of Article2(c) Framework Directive.25

Thenotion of associatedfacilities is defined in Article2(e) Framework Directive as “those facilities associated with an electronic communicationsnetwork and/or an electronic communications service which enable and/or support theprovision of services via that networkand/or service.Itincludes conditional access systems andelectronic programme guides.” Hence, EPGs –aform of audiovisual search tool as explained in our introduction –are explicitly coveredbythe directives.But how fardoes this notion extend? Does it, forinstance,alsoinclude search engines on theInternet? Contrary to the conceptsofconditional access systems(CAS) andapplication program interfaces (APIs),26 EPGs arenot further defined. Only by lookingatthe relevant rules in theAccess Directivedoes it becomeclear27 that thenotion of an EPG is also confined to thecontext of digital radio andtelevision broadcasting services andonly regulated from theangle of itstechnical(transmission) aspects:

•Article5,para. 1(b) andArticle6Access Directive(in conjunctionwith Annex I) contain aspecific access regimefor associated facilities,morespecifically conditional access systems(Article6), application program interfaces andelectronic programmeguides (Article 5), in theareaofdigital radio andtelevisionbroadcastingservices. 28 This makes it unlikely that Internetsearch engines would be considered electronic programmeguides in thesense of theelectronic communicationsdirectives.

•The purpose of these provisionsistoprovidetoolstoensureaccess on fair,reasonableand non- discriminatory (“FRND”) termstothese facilities.Access is to be interpreted in atechnical sense here,coveringtransmission aspectsonly (access to theunderlyingtechnology or routingfacilities only). We can infer this notonly from thedefinition of “access” itself in Article2(a) Access Directive,29 but also from Article6,al. 4, which states that “conditionsapplied in accordance with

24) See also Recital 10 Framework Directive(“Thesameundertaking, forexampleanInternet service provider,can offer both an electronic communicationsservice,suchasaccess to theInternet, andservices not coveredunderthis Directive, such as the provision of web-based content .”) andRecital 20 Authorisation Directive(“Thesameundertaking, forexampleacableoperator, can offer both an electronic communicationsservice,suchasthe conveyanceoftelevision signals, andservices not coveredunder this Directive, such as thecommercialisation of an offer of soundortelevision broadcastingcontentservices,and therefore additional obligationscan be imposed on this undertakinginrelation to itsactivity as a contentprovider or distributor,according to provisionsother than those of this Directive, without prejudice to thelist of conditionslaid in theAnnex to this Directive”). 25) Cf. vanEijk (2006) with regardtosearch engines.See also therecentArticle29WorkingParty Opinion on data protection issues related to search engines,mentioned infra ,footnote 57, “Search engines thereforefall outsideofthe scope of thedefinition of electronic communication services” (which does howevernot prevent thesearch engineprovider from beingsubject to the Electronic CommunicationsDirectives if he also offersadditional services that meet thedefinition of electronic communications service,suchasapublicly accessible email service). 26) Respectively Articles 2(f) and2(p) Framework Directive. 27) Forconditional access systemsand application program interfaces,this is already clear from thedefinitionsthemselves. 28) Article5(1) andAnnex 1Part II of theAccess Directivestate that in theabsence of significantmarket power(SMP), national regulatory authorities areabletoimpose obligations on operatorstoprovideaccess to APIsand EPGs on fair,reasonable, and non-discriminatory termstoensureaccessibility forend-userstodigital radio andtelevision broadcastingservices specified by theMember State. 29) “Access meansthe makingavailableoffacilities and/or services,toanother undertaking, underdefined conditions, on either an exclusiveornon-exclusivebasis,for thepurpose of providingelectroniccommunicationsservices.Itcovers inter alia: access to network elements andassociated facilities,which may involvethe connection of equipment, by fixed or non-fixed means(in particular this includesaccess to thelocal loop andtofacilities andservices necessary to provideservices overthe local loop), access to physical infrastructureincludingbuildings, ductsand masts; access to relevant softwaresystemsincludingoperational supportsystems, access to number translation or systemsofferingequivalent functionality, access to fixed andmobilenetworks, in particular forroaming, access to conditional access systemsfor digital television services; access to virtual network services.”

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHING FOR AUDIOVISUALCONTENT 69

this Articleare without prejudice to theability of Member States to impose obligationsinrelation to thepresentational aspect of electronic programmeguides andsimilar listingand navigation facilities.” Hence, Member States remain free to regulate contentaspectsofelectronic programme guides and“similarlistingand navigation facilities” (likeInternet search engines), as these are notcoveredbythe Electronic CommunicationsDirectives.

Couldsearch tools, other than electronic programmeguides,nevertheless fall within thebroader category of “associated facilities”? Possibly,asthis notion is formulated broadly (not confined to the digital broadcastingsphere) anditcouldbeargued that search tools“enableand/or support the provision of services via that network and/or service” (just likeelectronic programmeguides do in the area of digital broadcasting). In that regard, it is worth notingthat thenotion of associated facilities wasdeliberately extendedduringthe legislativeprocess leadingtothe adoption of theFramework Directivein2002 from facilities necessary to provide“electronic communications services” to facilities enablingorsupporting theprovision of “services” (includingbroadcastorinformation society services).30 Theusefulness of such qualificationwould, however, remain limited. Even though the generalregimeofsignificantmarket powercould, in principle, be applied to associated facilities, triggeringthe application of ex ante obligationsfor providersofthose facilitieswith aposition on the market equal to dominance is highly unlikely to happen in theshort to medium term. Firstly,sucha market is notincludedinthe Commission’sRecommendation on Relevant Markets,31 meaningthat national regulatory authorities whowouldwanttoanalyseitfaceasubstantial burdenofproof,inthat they wouldhavetodemonstrate that theconditionsfor theso-called three criteria-test arefulfilled.32 Secondly,the possibleremedies that can be imposed alwaysrelate to thenotion of “access” in a technical sense (as we havealready explained), coveringtransmission aspectsonly,and arehence not (well) suited to address problemsrelated to contentaspects.

TheReform Proposals

TheElectronic CommunicationsDirectives arecurrently beingrevised. TheEuropean Commission published itslegislativeproposalson13November 2007, andthe European Parliamentheldits first readingon24September 2008.33 No fundamental changeshavebeen introduced with regardto “associated facilities”ordistribution functions, notwithstandingseveralsuggestionsbyconsultants in thecontext of preparatory studies.Wecouldthink of theattention that theCullen Study of 2006 has drawn to theneed formoreclarificationofthe distribution function in thecontext of theelectronic communicationsframework (as it is now largely in thegrey zone).34

30) See,for instance,Report Paasilinna (ITRE) on theProposal foraDirectiveofthe European Parliamentand of theCouncil on acommon regulatory framework forelectronic communicationsnetworksand services (COM(2000) 393 –C5-0428/2000 – 2000/0184(COD)), A5-0053/2001, PE 297.096, 7February 2001, amendment19, p.17-18. 31) European Commission Recommendation of 17 December 2007 on relevant product andservice marketswithin theelectronic communicationssector susceptibletoexanteregulation in accordance with Directive2002/21/EC of theEuropean Parliamentand of theCouncil on acommon regulatory framework forelectronic communicationsnetworksand services, [2007] OJ L344/65. 32) In ordertoidentify marketsthat aresusceptibleto ex ante regulation, three cumulativecriteria need to be fulfilled: (1) the presence of high andnon-transitory barrierstoentry,(2) absence of adynamic market structurethat tends towards effectivecompetition within therelevant timehorizon, (3) application of competition law alonewouldnot adequately address themarket failure(s) concerned. If amarket is includedinthe Commission’slist, thereisapresumption that the market in question meetsthese criteria; if not, thenational regulatory authority has to providethe relevant evidence. 33) European Parliamentlegislativeresolutionsof24September 2008 on theproposal foradirectiveofthe European Parliament andofthe CouncilamendingDirective2002/21/EC on acommon regulatory framework forelectronic communications networks andservices,Directive2002/19/EC on access to,and interconnection of,electronic communicationsnetworksand associated facilities,and Directive2002/20/EC on theauthorisation of electronic communicationsnetworksand services,P6_TA- (2008)0449 (COM(2007)0697 –C6-0427/2007 –2007/0247(COD)) –A6-0321/2008 andonthe proposal foradirectiveofthe European Parliamentand of theCouncil amendingDirective2002/22/EC on universal service andusers’ rights relatingto electronic communicationsnetworks, Directive2002/58/EC concerningthe processingofpersonal data andthe protection of privacy in theelectronic communicationssector andRegulation (EC) No 2006/2004 on consumer protection cooperation, P6_TA- (2008)0452 (COM(2007)0698 –C6-0420/2007 –2007/0248(COD)) –A6-0318/2008. Availableat: http://www.europarl.europa.eu/sides/getDoc.do?pubRef=-//EP//NONSGML+TA+20080924+SIT+DOC+WORD+V0//EN&language=EN 34) Cullen International. 2006. Study on theRegulation of BroadcastingIssues under theNew Regulatory Framework, 11-14.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 70 SEARCHINGFOR AUDIOVISUAL CONTENT

Thenotion of “associated facilities” will remain unchangedwhenitcomes to electronic programme guides.They arenot affected by thechangesthe Commission proposed to Article2(e) Framework Directive: “associated facilitiesmeansthose facilities associated with an electronic communications network and/or an electronic communicationsservice which enableand/or support theprovision of services via that network and/or service or have the potentialtodoso ,and include number or address translationsystems,conditional access systemsand electronic programmeguides, as well as physical infrastructuresuchasducts, masts, street cabinets, and buildings)”. 35 TheEuropean Parliamentadopted an amendmentaddingthe followingfacilities,which again havenolink with electronic programmeguides: “ entries to buildings,building wiring, towers and othersupporting constructions,ducts, conduits,masts, antennae,manholes and cabinets and all other network elements whichare not active ”. Also thesuggested (minor) changestoArticles5and 6Access Directivedonot alter thelegal statusofelectronic programmeguides.

TheCommission, however, proposed to extendthe notionof“access”inArticle 2(a) Access Directive as follows:“access meansthe makingavailableoffacilities and/or services,toanother undertaking, underdefined conditions, on either an exclusiveornon-exclusivebasis,for thepurpose of providing electronic communicationsservices or deliveringinformation societyservices or broadcast content services .Itcovers inter alia: access to network elements andassociatedfacilities,[…] access to physical infrastructureincludingbuildings, ductsand masts; access to relevant softwaresystems includingoperational support systems; access to number translation or systemsofferingequivalent functionality; access to fixed andmobilenetworks, in particular forroaming; access to conditional access systemsfor digital television services; access to virtual network services”. 36 This wouldimply that contentprovidersand/or distributorswouldbeentitled to access andinterconnection underthe sameconditionsasprovidersofelectronic communicationsnetworksand services today andwould resolve ongoingdiscussions aboutwhether contentprovidersand/or distributors can be direct beneficiaries of access obligationsimposed on network operators. Theproposed modification of Article 20 Framework Directive–which clarifies that disputes between “service providers” wouldfall within thepowers of dispute resolutionmechanismsofnational regulatory authorities from themoment“one of thepartiesisanundertakingprovidingelectronic communicationsnetworksorservices”(meaning that theother party couldbeabroadcaster or information society service provider) –goes in thesame direction.

It remainstobeseen whether these Commission proposalswill be retained in thefinal text of the directives.The draft Trautmann Report suggested that thereference to information society services or broadcast content services in thedefinitionof“access” be removed, in order“to prevent the regulation from coveringcontentissues,which wouldopen amuchbigger fieldoflitigation (already coveredinAVMSand eCommerce Directives)”. 37 However, after thevote in theITRE Committeeon7 July 2008, thefinal Trautmann Report contained thepeculiaramendment 91 which suggestschanging thetextinto “makingavailableoffacilities and/or services,toanother undertaking, underdefined conditions, on either an exclusiveornon-exclusivebasis forthe purpose of providingelectronic communications services, including the delivery of information society services or broadcast

35) European Commission, Proposal foradirectiveofthe European Parliamentand of theCouncil amendingDirectives 2002/21/EC on acommon regulatory framework forelectronic communicationsnetworksand services,2002/19/EC on access to,and interconnection of,electronic communicationsnetworksand services,and 2002/20/EC on theauthorization of electronic communicationsnetworksand services,COM(2007)697 rev1, 13 November 2007, availableat: http://ec.europa.eu/information_society/policy/ecomm/library/proposals/index_en.htm Thetextinboldmarksthe suggested modifications. 36) COM(2007)697 rev1, 13 November 2007. Thetextinboldmarksthe suggested modifications. 37) European Parliament(ITRE Committee –Rapporteur: Trautmann), Draft Report on theproposal foradirectiveofthe European Parliamentand of theCouncil amendingDirective2002/21/EC on acommon regulatory framework forelectronic communicationsnetworksand services,Directive2002/19/EC on access to,and interconnection of,electronic communicationsnetworksand services,and Directive2002/20/EC on theauthorization of electronic communications networksand services (COM(2007)0697 –C6-0427/2007 –2007/0247(COD)), PE398.542v02-00, 23 April 2008, Amendment 60, p. 46 (see also Amendment615 of Pleguezuelos; Amendments to theDraft Trautmann Report, vol. 5(amendments 603- 716), PE407.731v01-00, 10 June2008, p. 15). Availableat http://www.europarl.europa.eu/meetdocs/2004_2009/organes/itre/itre_20080707_1900.htm or http://www.europarl.europa.eu/activities/committees/publicationsCom.do?language=EN&body=ITRE

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 71

content services .”38 This amendmentcontainsinour viewacontradiction with thedefinition of electronic communicationsservices (explicitly excludingservices providing, or exercisingeditorial controlover, content, supra ), but wasnevertheless adopted by theEuropean Parliament on 24 September 2008.

Proposalsinthe sphereofconsumer protection (e.g., transparency forusers; Article20Universal Service Directive) remain limitedtocontractual situations(i.e., relationsbetween subscribersand providers) andto“undertakings providingelectronic communicationsservices and/or networks” (and nottoprovidersofassociated facilities).

Unless somedrastic reformsare proposed by theCouncil, the new telecomspackage willnot fundamentally change thelegal position of search toolsinthe context of EU transmission regulation.

1.3. …inthe AudiovisualMedia ServicesDirective

On 11 December 2007, theEuropean legislator adopted itslong-awaited Audiovisual Media Services Directive(AVMS Directive), modernisingand renamingthe “Television without Frontiers” Directive (TVwF Directive).39 Theaim of this Directivehas alwaysbeen (andstill is) thecreation of an internal market fortelevision services and, in this light, theprovision of minimum harmonisation in certain areas of broadcastingpolicy(includingprotection of minorsand of human dignity,cultural diversity andmedia pluralism,access to information andcommercial communications).

Thenew Directivecovers more than just traditionalbroadcastingservices.Its scope of application includesboth linear broadcastingservices,irrespectiveofthe underlyingtechnology or platform, and on-demandaudiovisual media services (such as video-on-demand). These aregrouped underthe common headingofan“audiovisual mediaservice”, which is defined as “a serviceasdefinedby Articles 49 and50ofthe Treaty which is underthe editorial responsibility of amedia service provider andthe principal purpose of which is theprovision of programmes in ordertoinform, entertain or educate,tothe generalpublic by electronic communicationsnetworkswithin themeaningofArticle 2(a) of Directive2002/21/EC”.

Thevariouselements of this definition –which arecumulative criteria 40 –are further clarified in variousrecitals. It wouldfall outsidethe scope of this contribution to engage in an in-depth analysis of these criteria, 41 so we will look at only those aspectsthat arerelevant to search tools. Recital 18 stressesthat thedefinition of audiovisual media service shouldonly covermass media in their function to inform, entertain andeducate thegeneral public andthat it excludesall services whose principal purpose is notthe provision of programmes,i.e., whereany audiovisual contentismerely incidental to theservice andnot itsprincipalpurpose.“Examples include websites that contain audiovisual elements only in an ancillary manner,suchasanimated graphical elements, short advertisingspotsor informationrelated to aproduct or non-audiovisual service.”42 Recital 18 continues that, “for these reasons”, games of chance (involvingastakerepresentingasum of money), lotteries,bettingand other

38) European Parliament(ITRE Committee –Rapporteur: Trautmann), Final Report on theproposal foradirectiveofthe European Parliamentand of theCouncil amendingDirective2002/21/EC on acommon regulatory framework forelectronic communicationsnetworksand services,Directive2002/19/EC on access to,and interconnection of,electronic communicationsnetworksand services,and Directive2002/20/EC on theauthorization of electronic communications networksand services (COM(2007)0697 –C6-0427/2007 –2007/0247(COD)), A6-0321/2008, PE398.542v03-00, 22 July 2008. 39) Directive2007/65/EC of theEuropean Parliamentand of theCouncil of 11 December 2007 amendingCouncil Directive 89/552/EEC on thecoordination of certainprovisions laid downbylaw,regulation or administrativeactioninMember States concerningthe pursuit of television broadcastingactivities [2007] OJ L332/27, availableat http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2007:332:0027:0045:EN:PDF 40) Cf. recital 25: “All thecharacteristics of an audiovisual media service set out in itsdefinition andexplained in Recitals16 to 23 shouldbepresentatthe sametime.” 41) See,for instance,Chavannes,Remyand Castendyk, Oliver,in European Media Law,edited by Castendyk, Oliver,Dommering, Egbert, andScheuer,Alexander, Alphen a/d Rijnn, 2008. Chapter I. Definitions, 813-819. Valcke,Peggy andLievens, Eva. 2008. RethinkingEuropean broadcastingregulation -UnravellingEurope’spolicyfor thedigital media landscape: critical analysis of theAudiovisual Media Services Directive. In RethinkingEuropean Media andCommunicationsPolicies,edited by Pauwels, Caroline, andKalimo, Harri. Brussel: VUB Press (forthcoming). 42) Think of thewebsite of atravelagent or car manufacturer whereclips of holiday resorts, respectively thelatest car model areshown.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 72 SEARCHINGFOR AUDIOVISUAL CONTENT

formsofgamblingservices,aswell as on-linegames and search engines shouldbeexcludedfromthe scope of theDirective.

Since in most casessearch engines on theInternetwill notfulfil thecriterion of providing programmes as their principal purpose,they will notqualify as audiovisual media services,evenifthey were to occasionally show audiovisual material.

If thesearch engineweretobeameansofaccessingvideo clips –likeGoogleVideo– some commentatorsargue that it mightpotentially qualify as an on-demandaudiovisual media service, amountingasitdoes to acatalogue of programmes. 43 In their view, this will dependonwhether such an offer wouldbesufficiently “television-like” (cf. recital 17) to be “comparabletothe form and contentoftelevision broadcasting”, as Article1(b) of theDirective(containingthe definition of “programme”) putsit, andconsequently on how thesearch engineispresented to thepublic.They refer to aversion of YouTube (which clearly has search functions) made availablethrough an AppleTV as an examplethat mightvery well qualify.

In my view, however, theabsence of editorial responsibilityinthe sense of theDirectivewill prevent such an audiovisual search tool from beingconsidered an audiovisual media service.This notion of “editorial responsibility” has –atthe instigation of Parliamentand Council –becomeakey element of thedefinition of audiovisual media service. 44 It means“theexercise of effectivecontrol both over theselection of theprogrammes andovertheir organisation either in achronological schedule, in the case of television broadcasts, or in acatalogue,inthe case of on-demandaudiovisual media services” (but does notnecessarily imply anylegal liability undernational law forthe contentorthe services provided; Article1(c) AVMS Directive).

It is difficult to argue that themachine-poweredgeneration of search resultsbased on algorithms is similar to theeffectivecontrol overthe selection andorganisation of programmes in achronological scheduleorcatalogue by abroadcaster or video-on-demandprovider.Evenifsearch engines dedicated to video or video sites likeYouTube adopt apolicyofremovingvideos that havebeen flagged by the usersofthe community themselves as inappropriate,this interventionis ex post, actingupon complaints from users.45 Moreover, theeditorialcontrol is what VanHoboken calls“distributed”, i.e., shared with theinformation providersand Webusersthemselves tagging andrankingthe contentthey post andconsumeonline.46

What if theselection andpresentation of audiovisual materialwouldbeperformed by human- poweredorganisationswith editorial centres? In that case,wecomeclose to thenotion of electronic programmeguides,which arementioned in recital22: “Whilethe principal purpose of an audiovisual media service is theprovision of programmes,the definition of such aservice shouldalsocovertext- based contentwhich accompanies programmes,suchassubtitlingservices andelectronic programme guides. Stand-alone text-based servicesshouldnot fall within thescope of this Directive, which should notaffect Member States’ freedom to regulate such services at national levelinaccordance with the Treaty.”

In other words, electronic programmeguides arenot considered an audiovisual media service in themselves (since their principal purpose is notthe provision of programmes,but thelistingofthose programmes with additional information andfunctionalities likezapping, recording, etc.), but if they accompanyanaudiovisual media service,they nevertheless fall underthe scope of theDirective. The electronic programmeguideisthenconsidered part of theservice package of theaudiovisual media service provider and, as such, shouldfollow therules of theDirectiveaswell (for instance,those relatingtothe protection of minorsand of human dignity, or mutatis mutandis therules on audiovisual commercial communications).

43)Chavannes,Remyand Castendyk, Oliver.2008. op.cit.,814. 44) See Schulz, Wolfgangand Heilmann, Stefan. 2008. In IRISSpecial: Editorial Responsibility,edited by Nikoltchev,Susanne, Strasbourg, European Audiovisual Observatory. 45) See,for instance,YouTube Community Guidelines at http://www.youtube.com/t/community_guidelines 46) VanHoboken, Joris.2008. Legal Space forInnovativeOrdering–Theneed formoreclarity regardingthe liability of selection intermediaries forthirdparty information. Paper presented at theEuroCPR2008 conference,Seville, March 2008, at p. 4.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 73

Acontrario,ifthe electronic programmeguideisoffered as astand-aloneservice (i.e., notpart of an audiovisual media service),for instance by thecableorsatellite operator,the rules of theDirective no longerapply.

1.4.…in the E-CommerceDirective

TheE-Commerce Directive,47 adoptedin2000,intends to create acommon fundamental legalframework forelectronic commerce in theInternal Market andstimulate thedevelopmentofcross-borderonline services.Itapplies to theso-called “information society services”, defined as “any service normally provided forremuneration, at adistance,byelectronic meansand at theindividual request of a recipientofservices”. 48 This notion covers awidevariety of services provided online, rangingfrom onlinenewspapers, specialised news services (such as business or financial information) andonline sellingofvariousproductsand services (likebooks, DVDs,computer hardwareand software, travel services, pharmaceuticals, etc.)tothe onlineprovision of financial services (onlinebankingand onlineinvestment).

Recital 18 of theDirectivereads: “…information societyservices arenot solely restricted to services givingrise to on-linecontractingbut also, in so farasthey representaneconomic activity,extendto services which arenot remuneratedbythose whoreceivethem, such as those offeringon-line informationorcommercialcommunications, or those providingtoolsallowingfor search, access and retrievalofdata”.Inother words, onlinesearch engines are, in principle, includedinthe scope of application of theDirective.

Therelevant rules,however, remain limited. Firstly,because thecoordinated fields in theDirective andthe harmonisationachievedbyitremain limited in scope,and secondly,because therules that seem most relevant forsearch engines –notably theliability rules forintermediary service providers –donot apply to them (only to mere conduit, hostingand cachingactivities). We will notfurther address this issue,asitisdealt with in more detail in thecontribution of Joris vanHoboken.49

1.5.Interim Conclusion

Thepreviousparagraphshaveshownthat thecurrentEUlegal framework dealsonly in avery fragmented manner with audiovisual search tools. Limited provisionsexist –both in theElectronic Communications Directivesand in theAudiovisualMedia Services Directive–with regardtoelectronic programmeguides (which we interpret here as “editedaudiovisual search tools”, linked to digital broadcastingina walled-gardenenvironment). In contrast, onlinesearch tools–eventhough they areincludedinthe generalscope of application of theE-Commerce Directive–are notevencoveredbythe safe harbour provisionsinthe E-Commerce Directive. In other words, notmuchhas changedwith regardtothe legal position of search toolssince vanEijk’sanalysis in IRIS plus 2006-2 (“Seek andYeShall Find?”).

2. PreliminaryThoughts on Regulating the Audiovisual Search Function

2.1.DoweNeed to Regulate?

Afundamental question is whether we need to regulate at all. We need to havegood arguments in ordertoregulate functionsfor which we haveneverenacted sector specificregulationsinthe print andanalogue environment–think of TV listingguides,bookshops,newspaper kiosks, libraries; they

47) Directive2000/31/EC of theEuropean Parliamentand of theCouncil of 8June2000 on certain legal aspectsofinformation society services,inparticular electronic commerce,inthe Internal Market (Directiveonelectronic commerce), [2000] OJ L 178/1, availableat: http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2000:178:0001:0016:EN:PDF 48) Article1(a) Directive98/34/EC of theEuropean Parliamentand of theCouncil of 20 June1998 layingdownaprocedure forthe provision of information in thefieldoftechnical standards andregulation, [1998] OJ L204/37, as amendedby Directive98/48/EC of 20 July 1998, [1998] OJ L217/17. 49) See also vanHoboken, Joris.2008. Legal Space forInnovativeOrdering–The need formoreclarity regardingthe liability of selection intermediaries forthirdparty information. Paper presented at theEuroCPR2008 conference,Seville, March 2008. Seealsothe articlebyJoris vanHoboken in this publication.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 74 SEARCHINGFOR AUDIOVISUAL CONTENT

haveall been selectingand facilitatingaccess to contentfor manyyears, without anyone defending theneed to subject them to special neutrality or other rules.What is so differentabout theonline, digital environment–apart from thefactthat we are, at this stage,not as familiarwith it andthat theprovidersofthe search services can be located virtually anywhereonthe planet?50 In thecontext of onlinetext-based search engines,the (hidden) manipulation of search resultsand thedominance of oneprovider (Google) havebeen mentioned as possiblereasonsmotivatingthe introduction of regulatory intervention.51 But such issues havenot yetbeen broughttothe forefrontwith regardto audiovisual search tools.

2.2. Whom/what Do we Want to Regulate?

Thesecondquestion: whom/what do we want to regulate? Only audiovisual search toolsthat involvesomeform of editorial control (likethe prior selection of thechannelsorprogrammes that will be listed andbepresented in asearchableform) or also “automated” search services (with removalof certain references ex post,upon receipt of complaints from users, at themaximum)? Thereissomething to be said forthe first option, since these imply agreater control by theprovider.But, ironically,this may lead to perverse effects: themoreyou control, themoreyou will be regulated, so letuscontrol less…

2.3. WhydoWeWant to Regulate (and how)?

Thirdly, what arethe policygoalsthat wouldneed regulatoryintervention (andhow couldthey best be achieved)? Concernsovermarket powerand market foreclosuremighttrigger intervention, but the first toolsfor realisingfreeand fair competitionare thegeneral competition rules (in thefirst place, in our case,the prohibition of abuse of adominantposition in Article82ECTreaty).52 Only in sectors that were formerly characterised by monopolies,havesector specific instruments been introduced allowingthe imposition of ex ante remedies on undertakings with acertain position on themarket (likethe significantmarket powerregimeintelecommunications). Thedigital television sector has been apeculiar case,with access obligationsfor conditional access providersirrespectiveoftheir market position. ThecurrentAccess Directiveinherited this regimeand created comparable(though notsimilar) rules for“new”digital bottlenecks, likeapplication programmeinterfaces andelectronic programmeguides.Asdiscussed already,these rules relate to transmission aspectsofthose facilities only andsay nothingabout presentational or contentissues.

When it comes to cultural objectives (such as media pluralism), theneed forinterventioncouldarise when certain contentthat is considered of public importancewouldnolongerappear at avisibleplace in thesearch list. Therefore, thesuggestion made by Cullen in the2006 Study to specify that must- carry guarantees access to thetransmission network andtoall theassociated facilities –intheir view to be interpreted as all distributionfunctions, hencecovering, besides CAS, API andEPG, also DRMs, set-top-boxes, themultiplex,and search engines53 –seemsworth considering, but may be only afirst step. Indeed, in an eraofabundance,the important thingisnot merely to be listed, but to be listed in a prominent place.The German provisionson“Navigatoren” in the Satzungdigitaler Zugang, containingthe obligationfor a“balanced” listingofthe programmeoffer of both public and commercial channels, can be cited as an exampleofprovisionspursuingthat aim.54

50) Although if they really want to targetaspecific geographically defined audience,they arelikely to haveatthe medium to long term at least alocal branch or representativetostay in touchwith that audience’scultureand preferences,and establish close contactswith local advertisers. 51) VanEijk 2006, op. cit. at p. 6and vanEijk, Nico. 2007. Search Engines,the new bottleneck forcontentaccess.Paper presented at theInternational TelecommunicationsSociety 19th European Regional Conference,Istanbul, 2-5 September 2007, at p. 20 (availablefromhttp://www.ivir.nl/publications/vaneijk/Paper_SearchEngines_ITS_2007.pdf) 52) Mergercontrol rules couldalsocomeinto thepictureifpowerful search tool providersstarted aprocess of horizontal or vertical integration through alliances andtake-overs, or theprohibition of cartel agreements andconcerted practices may apply to (exclusive) vertical agreements between contentprovidersand search tool providers. 53) Cullen International, op. cit.,footnote 344, pp.120 and125. 54) Satzungüber die Zugangsfreiheit zu digitalen Diensten gemäß §53Abs.6Rundfunkstaatsvertrag,§13 (3): „ Aufdas öffentlich-rechtlicheund private Programmangebot muss gleichgewichtig hingewiesen werden.“ (adopted by the Direktorenkonferenz derLandesmedienanstalten (DLM) on 13 December 2005, availableat http://www.alm.de/fileadmin/Download/Gesetze/Satzung_gem.____53_-End-.pdf).

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 75

Consumer protection might be another relevant policygoal in theareaofaudiovisual search tools. Underthis heading, rules seekingtoenhance transparency of rankingmethods (and“sponsored” links) couldbeenvisaged, as well as rules dealingwith theremovalof(and/or thewarningfor) searchresults referringtoillegal (and/or harmful) content(likesettingthe conditionsand procedures forsuch removalorwarningnotices). As an exampleofthis,wecan refer to thecodeofconduct forsearch engines in Germany, which is described in thecontribution of WolfgangSchulz.55

Privacy protection couldbearising star in thefirmamentofpolicygoalsjustifying(more)regulatory intervention. Concernsabout people’sprivacy aregrowinginthe lightofemergingtechnologies designed to traceand recall past user activity onlineorondigital TV platformsinordertopersonalise services.Aswas demonstrated by Michael Zimmer, 56 dozensofproductsand services arebeinglinked via cookiesand user accounts andpeople areleaving behindanincreasingamountof“digital fingerprints”onthe Internet (likecontacts, blogs, books, stocks, appointments, e-mail, friends, computerfiles,discussion groups,URLs). Moreover, they lose control of what Zimmer calls“their likeness online”, since they mightneverknow what personaldata relatingtothem(including audiovisual material, likephotosorvideos) other peopletakeand place online(letalone consentto it). In this respect, we wouldliketodrawthe reader’sattention to therecently adopted Opinion 1/2008 on data protection issues related to search engines,inwhich theArticle29WorkingParty –urgedonbygrowingconcernsamong European usersofsearch engines about potential breaches of their privacy rights –scrutinises searchengineactivity anddata storageunder thecurrent EU directivesrelatingtothe protection of personal data anddata retention.57 This opinion clarifies that search engineproviders, dependingontheir exactactivities, will in manyinstances be considered as controllersofpersonal data andhavetocomply with theapplicablerules andprinciples.These include theduty to delete or makeanonymous personal data once they arenolongernecessary for thepurpose forwhich they were collected, as well as theneed forsearch engineprovidersthat specialise in thecreation of value addedoperations, such as profiles of natural persons–the so called “peoplesearch engines” –and facial recognition softwareonimagesnot only to havea legitimate groundfor processing(such as consent), but also to meet all other requirements of Data Protection Directive95/46/EC (such as theobligation to guarantee thequality of data andfairness of processing).

3FinalComments

At present, both electronic programmeguides andsearch engines fall largely outsidethe scope of thesector-specific media and(tele)communications regulatory frameworks. Nevertheless,weare of the opinion that most policygoalsabout which concernsare growinginthe context of search toolscan be fairly addressed by applyingexistingand/or generalprinciples or rules,suchascompetition rules (toaddress market powerconcerns), data protection legislation (tocater forprivacy concerns) or rules on unfair commercial trade practices andadvertising(to enhance transparency on the“manipulation” of results). It may be more amatter of ensuringconsistency with these rules,rather than havingto enact new rules.

We agree, though, with worried commentators 58 that developments shouldbemonitored closely,as thesearch functionindeed becomes crucial in an eraofabundance (perhaps evenmoreimportantthan theprovision of contentitself, now that we arewitnessinganunprecedented loweringofthe threshold forbecomingacontentprovider yourself).

55)See thearticlebyWolfgangSchulz in this publication. 56) See thearticlebyMichael Zimmer in this publication. 57) Article29WorkingParty,Opinion 1/2008 on data protection issues related to search engines,4April 2008, WP148. The Article29WorkingParty is aEuropean Union advisory body comprised of representatives of theMember States’ data protection agencies.Opinionsofthe Article29WorkingParty arenot legally binding, but often becomethe “EU standard”, andtherefore arehighly influential in shapingEUpolicyaswell as Member State implementation in theprivacy field. 58) See,for instance,the critical commentwith regardtothe lack of transparency in internet search engines andthe hidden manipulation of search results:Machill, Marcel, Neuberger, Christoph andSchindler,Friedemann. 2003. Transparency on the Net: functionsand deficiencies of Internet search engines. 5 Info 1, pp.52-74.

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 76 SEARCHINGFOR AUDIOVISUAL CONTENT

In theaudiovisual context, thesearch function is often bundled with other distribution functions (likethe provision of aplatform forhostingand sharingaudiovisual content), which makes us wonder whether potential regulatory attention shouldnot be broaderthan thesearch function itself. In some EU Member States59 andalsoinJapan60,legislatorshaveintroduced or areintroducingathree-layered regulatory approach forthe info-communicationssector,distinguishingbetween content, distribution andtransmission.61 Hence, theprovidersof“intermediate functions” (i.e., intermediate between the editingand provision of content, on theone hand, andthe provision of transmission networksand services overand via which that contenttravels, on theother hand) wouldfall underacommon regulatory framework. If this were to be thecase at EU levelaswell, this wouldallow theEuropean legislator to set downthe basic rules andprinciples that wouldapply (likeopenness andfairness), whileatthe sametimeleavingscope forfurther refinementofthese basic rules through –infirst instance –co- and/or self-regulatory instruments.After all, findingthe rightbalance between conflicting policygoalsand/or rights andfreedomsmay be alengthyprocess.For instance,where does thefreedom of expression –and hencethe freedom to adopt an activeselection policy–ofthe search provider endand wheredoes his obligation of non-discrimination start? How shouldwecorrelate the search provider’sliabilitywith theextentofactiveselection? Such questionswill undoubtedly occupy us formanymoreyearstocome.

Thereisarisk in regulatingtoo muchand toosoon (which may providethe wrongincentives and stifleinnovation); thepromises of new technologies shouldonly be tempered when thethreatsbecome apparentand serious(although we shouldnot strikethe Faustian bargain that has been referred to by Michael Zimmer).62 Perhaps therewill comeatimewhenwewill curse search tools(once they start to know us better than we know ourselves), but forthe timebeingIbelieve they can still be considered extremely usefultoolsfor thedetection of relevant information.Admittedly,there is alot of information that we do notfindwith thepopular search toolsbecause they rank other itemsfirst for commercial reasonsorbecause they filter certain things out, but wouldn’t therebe alot more information that we wouldnot find without them (andno, I’m notsponsored by oneofthemtowrite this)?

Certainly,providersofsearch services shouldbeaware of their increasingroleinthe structuringof thepublic sphereand their growingimpact on society.Therefore,they shouldbeprepared to assume their social responsibilityand,ifthey fail, public powers may havetointerveneto“bind” theprivate powers. In this light, it is our role as academics to foster andcontinue thepublic discourse on these issues,since it cannot be left to theprovidersthemselves alonetoset theappropriate standards; they arenot expected to think andact in thegeneral interest. Initiativesfor further analysis andresearch areextremely valuableinthis context. So, search tools, be warned: youare notonly watchingus, we arealsowatchingyou!

59) Forinstance Belgium (theFrenchCommunity introduced athree-layered approach in its2003 BroadcastingAct –aswe described in our previouscontribution “TheFutureofMust-carry”.2005. In IRISSpecial: To Haveornot to HaveMust-carry Rules, edited by Nikoltchev,S., Strasbourg, European Audiovisual Observatory –and theFlemish Community will adopt a similar three-layered approach in itsnew Media Act, implementingthe AVMS Directive); see also theFrenchCommunications Act(Loi n°86-1067 du 30 septembre1986 relativeàlaliberté de communication, “Loi Léotard”) that uses theconcept of “distributor” (in particular in thecontext of digital terrestrial broadcasting). 60) In itspresentation at the17th Biennial Conference of theInternational TelecommunicationsSociety in Montréal (24-27 June 2008), Prof.Minoru Sugaya(Institute forMedia andCommunications, Keio University) presented thecurrentpolicyreforms in Japan, explaininghow Japan intends to integrate itsexistingregulatory frameworksfor (tele)communicationsand broadcastinginto asinglecomprehensive“information andcommunicationslaw”, structured alongthe lines of “content”, “platforms” (as intermediate layer) and“transmission infrastructure” (http://www.canavents.com/its2008/plenary3/13.pdf ; formoreinformation about therecommendationsofthe Study Group on aComprehensiveLegal System forCommunications andBroadcastingtowhich thepresentation of Sugayarefers, see: http://www.soumu.go.jp/joho_tsusin/eng/Releases/ NewsLetter/Vol18/Vol18_21/Vol18_21.html). 61) Or actually four-layered, if we consider terminal equipmentasthe fourth layer to be regulated. 62) See thearticlebyMichael Zimmer in this publication.

©2008, European AudiovisualObservatory, Strasbourg (France)

SEARCHINGFOR AUDIOVISUAL CONTENT 79

Regulating Search Engines? On the Use of Self- and Co-regulation in the FieldofInternetSearch

WolfgangSchulz* Hans-Bredow-Institut

Introduction

When media managers from differentsectorscongregate nowadays, it is likely –atleast in Germany –that they will begin squabblingamong themselves about theeconomic challengesthey faceand the perceived cut-throat character of thecompetition. Yet, thereisone common topic which brings them all back on speakingtermsagain: theloomingfear of Googleeatingawayatthe advertisingmarket.

Hence, theregulation of search engines,their possibleinfluence on public opinion making, their ability to discriminate,and other regulatory issues haveclimbed up thepolicyagenda. But thereisno clear idea of how to deal with thephenomenon. This contribution outlines theconcept of information intermediaries, such as search engines,itidentifies which public policygoalscouldbeendangeredand, finally,itaskswhether conceptsofco- andself-regulation mightbeadequate to meet theregulatory challenges.

1. Concept of Information Intermediaries

Conceptsofmedia regulation implicitly or explicitly makeassumptionsconcerningthe functioning of individual andpublic communication. Thespecificprotection of mass media underArticle10ECHR or Article5Section 1 Grundgesetz (theGerman Constitution –GG) is based on thepresumption that mass media fulfil aspecific function when it comes to public andindividual communication in a democratic society. 1 TheGerman constitutional court has explicitly emphasizedthe role of mass media anddistinguishes between theprotection of speech, i.e., of theactual contentofcommunication regardless of whether it is uttered as individual communication or is published by mass media services, andthe protection of mass media services (broadcasting, press andfilm arementioned in the Constitution, but it can encompass new mass media services as well).2 From this assumption, the federal constitutional court deduces specific elements of constitutional protection against State interference,aswell as theconstitutionalobligationsfor thelawmaker to ensurethat thepowerof public opinion making–that in principleisexercised if onecontrolsamass medium –are controlled effectively.

*) Director of theHans-Bredow-Institut, lecturer at theLaw Faculty of theUniversity of Hamburg 1) Hoffmann-Riem/Schulz in: PolitischeKommunikation in der demokratischen Gesellschaft,Wiesbaden1998, 154 (159 f.) (in German only); Hoffmann-Riem, RegulatingMedia,New York 1996, 119. 2) Cf. Decisionsofthe German Constitutional Court (in German only): BVerfGE 73, 118 (158); 74, 297 (324); 83, 238 (298); 90, 60; see also thelatest decision: BVerfG, 1BvR 2270/05, 11 September 2007 (in German only) availableat: http://www.bverfg.de/entscheidungen/rs20070911_1bvr227005.html

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 80 SEARCHINGFOR AUDIOVISUAL CONTENT

Thespecific role of mass media communication is an assumption which is still in linewith actual findings in communicationsscience today.Yet, therehas been astructural change in public communications. Whilemass media arecharacterised by controllingpublication as such and, furthermore, by editorial pre-selection, more andmoreformsofcommunication haveemergedwhere other types of service providersare of structural importance.The latter meansthat State interference with those services notonly affectsasingleact of communication, but theoverall structureofpublic communication. Iwouldliketocall these other types of service providers“service intermediaries”. Communication science is still in theprocess of characterizing types of onlineservices which might form thebasisfor future legal debate.Certain typesofplatforms such as Internet forumsmight be seen as “information intermediaries”.

Theimportant role of search engines has been proveninseveralstudies. 3 Whether it is appropriate to say that, as aprecondition, to exist on theInternet it is necessary to be indexedbyasearch engine 4 or that this mightrather be an exaggeration, is anybody’sguess.However, it is evidentthat search engines fulfil abasic function formakinginformation on theInternet accessible. Even if usersare becomingless dependent on search engines as they becomeincreasingly adept at usingthe Internet, this function still remainsimportant.

At first glance onecouldsay that search engines,asspecial intermediaries in communication, are completely differenttoclassical media service providers. Thereisnojournalistic-editorialpre- selection. Thesearch engineonly shows what theuser has specified in his or herquery.Schönbach has emphasizedthis difference by callingtraditional media “display media”.Itisapre-selectionwhich shall surprise theuser (in thecase of classical media services youwantsomeoneelsetochoose the contentaccordingtoagiven professional standardand,therefore,tosurprise youbythe content). On theother handwefindthe so called “research-media”, of which search engines areone very prominent, as well as paradigmatic,example.5 However, on closer examination, this classification of search enginesdoes notcompletely match all relevantfactors, especially when takinginto consideration thefactthat someofthe algorithms used by search engines try to targetthe most up- to-date pagesorthat search engines useindicatorsofprominence which arerelevant forjournalistic- editorial choices as well. Still, themain function forwhich search engines arebeingused remainsthe detection of aspecific piece of information about which theuser already has someclue,but theexact location of which he does notyet know.

Thespecific role of information intermediaries in individual andpublic communication may lead to specific riskstriggered by these intermediaries. Nevertheless,the specific function fulfilled by search engines forindividualorpubliccommunication mightalsorequirespecificprotection.This mightcall forspecific safeguards against regulation, rather than suggest legislativeaction. Thechoice depends on thespecific construction of theregulatory framework, which cannot be elaborated upon here.In this context, thefundamental rights and, in particular theguarantee of freedom of speech, has to be kept in mind.6

3) Cf. VanEijk, Nico. Search engines: Seek andyeshall find? Theposition of search engines in law.IRIS plus 2006-2; Strasbourg, European Audiovisual Observatory; availableat: http://www.obs.coe.int/oea_publ/iris/iris_plus/iplus2_2006.pdf ;Philippi, Internet Suchmaschinen: Suchmethodik –Erlösmodelle–Preisstrategien, (2006); Nitsche, Suchmaschinen im Internet - Bedeutung, Trends, Chancen undEntwicklungen(2005); Machill/ Neuberger/ Schweiger/ Wirth, WegweiserimNetz -Qualität undNutzungvon Suchmaschinen (2003); Introna/Nissbaum, Shapingthe Web: Whythe Politics of Search Engines Matters, TheInformation Society 171 (2000); Maurer et al. ,Report on dangers andopportunities posed by large search engines, particularly Google(2008), availableat: http://www.google-watch.org/gpower.pdf 4) Cf. Introna/Nissbaum, Shapingthe Web: Whythe Politics of Search Engines Matters, TheInformation Society 171 (2000). 5) Cf. Schönbach, Das EigeneimFremden. ZuverlässigeÜberraschung–einewesentlicheMedienfunktion? [Theown in the foreign. Reliablesurprise: an importantfunction of themedia], Publizistik,2005 (50): 344-352. 6) Forsearch engines relatingtoGerman constitutional law see Schulz, Vonder Medienfreiheit zum Grundrechtsschutz für Intermediäre? CR 2008, S. 470 et seq.

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 81

2. Risks Triggered by Search Engines

This contribution shall focusonthe riskspresented by Internet search engines andthe possible regulatory approaches forfacingthese risks.

TherisksarisingfromInternet search engines aremanifold:

•Access to harmfulcontent •Exploiting protected works •Access to illegalcontent •Exploiting personal data

•Distortionofcompetition, •Discrimination of content including •Misleading consumers •Transfer of market power •Influence on opinion to othermarkets making (e.g.advertising)

•Fragmentation of the public sphere

As adownsideofthe importantrolethat search engines play as gateways (if notgatekeepers) to theInternet, each andevery legal issue on theInternet can sooner or later be foundtoaffect them.

Thefirst block of riskstriggered by search engines is that they makeharmful contentorillegal content(more easily) accessible. What is considered to be harmful or illegal varies from country to country. In most countriesatleast, childpornography 7 is illegal without exception andinmost countries thereare also specific restrictions with regardtothe access of youngpersonstocontent which mightimpair their development.8

Another set of risksmightbetriggered by thepossibility forsearch engineproviderstochoose which pagesthey will include in their index,toremove pagesfromtheir index andtoarrange their result lists. This mightleadtodiscrimination of contentand,inthe case of commercial communication, misleadconsumersaswell as,onthe whole, influence theprocess of public opinion making. During thewar in Iraq rumourscirculated that certain search engines hadremovedpictures of US soldiers torturingprisonersfromtheir indexes.9 Given thefactthat Googleprovides over80%ofthe Internet search resultsrequested in Germany, it is no exaggeration to state that thefirst listings on Google’s search resultsonaspecific topic arehighly relevant forpublic opinion making.10 Furthermore, this has to be seen against thebackgroundofashift from traditional media, such as broadcastingand press,

7) Cf. forexample, §4 JMStV (Germany); FederalLaw 18 U.S.C. §2252 (USA); forEurope also see http://www.saferinternet.org/ww/en/pub/insafe/safety_issues/faqs/pornography.htm;harmonisation by theCouncil of Europe by wayofthe Convention on Cybercrime(ETSNo. 185). 8) See forexample: Awareness andHarmful Media Contentinthe Digital Age,Ulla Carlsson (Ed.) (2006), availableat: http://unesdoc.unesco.org/images/0014/001469/146955E.pdf 9) In regardtoInternet Filteringcf. Deibert/ Palfrey/Rohozinski/Zittrain, Access Denied: ThePractice andPolicyofGlobal Internet Filtering(2008). 10) Cf. Schulz/Held/Laudien, Search Engines as GatekeepersofPublic Communication Analysis of theGerman Framework ApplicabletoInternet Search Engines IncludingMedia Law andAnti-trust Law,in: German Law Journal, 2005 (vol 6, no 10), pp. 1419-1433, availableat : http://www.germanlawjournal.com/pdf/Vol06No10/PDF_Vol_06_No_10_1419-1432_Developments_SchulzHeldLaudien.pdf

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 82 SEARCHINGFOR AUDIOVISUAL CONTENT

to onlinemedia, due to theinformational needs of specific user groups,whilethe traditional mass media services still remain importantfor public communication.11

Returningtothe differences between display-media andresearch-media mentioned above,somesay that thefragmentation of thepublic spherehas already ledtomoreand more special interest services (andvice versa) andhas been pushed forbyresearch media. To begin with, empirical studies show that thereisasignificantchange in theoverlappingofrelevant issues on an individual’sagendain comparisonwith those on theagendaofother citizens, if individualsinform themselves mainly through theuse of research media.

Furthermore, it cannot be denied that search engines provide(easy) access to protected worksand to specific services,likesearches forimagesoraudiovisual content. In thecase of images, theroleof thesearch engineprovidersismoreactivethan just makingthe linkstoworksavailabletothe public, seeingasthe search engines create thumbnails of these pictures.However, it is arguablethat copyright is notinfringed.12 Thesameislargely applicabletoaudiovisual content.

Search engines areamong themost frequently visited sites on theInternet and, furthermore, the queries submitted to them by usersmakeanexcellentbasis forpersonal profiling. Therefore, data protection issues arise from theuse of search engines as well.13 As farasportrait search is concerned, other aspectsofpersonalrights mightalsobethreatened.

Finally,since in severalcountries oneand thesamesearch engineprovider has gained an extremely highmarket sharefor bothsearch servicesand targeted onlineadvertising, theproblem of competition distortion has to be addressed as well. Given that companieslikeGoogleextendtheir activities into other fields,likeadvertisingservices, 14 thetransfer of market powertoother markets, likethe advertisingmarket, also increases in relevance.

3. RegulatoryApproaches

3.1. General Remarks

Based on thefactthat oneofthe above-mentioned risksisconsidered as beingregulated and assumingthat thecurrentregulation is notsufficienttoaddress theexistingproblems, theright regulatory choice has to be made.Onavery abstractlevelthere is aregulatory triangledescribing which regulatory conceptshavetobetaken into consideration:

Law

Code Social Norms

11)Cf. Ridder/Engel , Massenkommunikation 2005: Imagesund Funktionen derMassenmedien im Vergleich. In: Media Perspektiven,Nr. 9/2005, S. 422 et. seq. 12) Cf. Wimmers/Schulz, Wernutzt? -Zur Abgrenzungzwischen Werknutzer undtechnischem Vermittler im Urheberrecht ,CR2008, 170 ff.; LG HamburgCase No. 308 O449/03, MMR 2006, 697-699; LG Erfurt Case No.: 301108/05, MMR 2007, 393-394; Commented on by Berberich in CR 2007, 393-395 andRoggenkamp in K&R 2007, 328-330; OLG Jena, Judgment of 27 February 2008 -2U319/07. Forfurther information about legal aspectsofsearch engines,linkingand framingand German Copyright Law see http://www.linksandlaw.com by Stephan Ott. 13) See forthe debate on European levelthe Article29Group, Data Protection WorkingParty,availableat: http://ec.europa.eu/justice_home/fsj/privacy/docs/wpdocs/2008/wp148_en.pdf 14) In this context especially thedoubleclick-deal on Googleisofimportance,Cf. http://www.businessweek.com/technology/content/apr2007/tc20070414_675511.htm

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 83

Social norms, likeprofessional rules of journalism, govern certain kinds of social behaviour; there aresettings of normswhich specifically apply to theInternet (e.g., netiquette). Theremightbe explicit rules forcertaingroups,likecodes of conduct or,onalowerlevelofformality,but evenmore common, internal rules,which nevertheless can be powerful. Legal regulation can tie in with these social norms, but legal regulation can also be eroded by them or weakened in itsenforcement, especiallywheninterferingwith social norms. That may occur in particular if social normsare not taken into accountbythe legislator when draftinglaws.

On theInternet especially,the softwareand hardwarearchitecturealsoimposes constraints, which areknownascode.15 It is less common forlegal regulation, though, to takethis type of codeinto consideration when designingaregulatory framework, which mightproduce aweakness of regulatory conceptsfor Internet regulation in thefuture.

3.2.Co- andSelf-regulation

Thespecific regulatory approaches on which this articlefocuses areself- andco-regulation. Theuse of co-regulation is encouragedbythe EU; explicitly in Art. 3ofthe Audiovisual Media Services Directive (which does notapply to Internet search engines), andalsoingeneral termsaspart of theconcept of better lawmaking.16 Talkingabout co-regulation andself-regulation meanstalkingabout the interaction between State regulation andnon-State regulation andthe leewayeachsidehas in determiningthe regulatory concept. On theone hand, thereisthe ideal of “command-and-control- regulation“, entirely designed andenforced by theState,while, on theother hand, thereispureself- regulation (which in practice is usually surroundedbylaw,but is,for theissue in question, entirely in thehands of non-State control).

Onegeneral advantage of self-regulation lies in theexpertise of theactors, which is most effective if they can taketheir owndecisions. This is in addition to theefficiency,the problem-orientation, the flexibility andthe high acceptance forthis type of regulation. But self-regulation is generally criticized as lackingdemocratic andlegal legitimacy,circumventingmechanismstoprotect basic freedoms (enforced private censorship), lackingaccountability andtransparency,beingunfair andcausinglegal uncertainty,while, at thesametime, arousingfearsabout fragmentation of thelaw.

Co-regulation is amixtureofboth theabove-mentioned regulatory approaches.Fromthe Hans- Bredow-Institute’spointofview, co-regulation can be defined as follows:“Co-regulation means combiningnon-Stateregulation andState regulation in such away that anon-State regulatory system linksinwith State regulation”. 17

Thelatter definition specifies thefollowingcriteria which mustbemet by aregulatory concept so as to classifyitasbeingco-regulation:

Thenon-Stateelement of theregulatory system includes:

•Creation of specific organisations, rules or processes •inordertoinfluence decisions •performed by theaddressees themselves

15)Cf. Lessig, CodeV2, availableat: http://codev2.cc/download+remix/ 16) Cf. Interinstitutional AgreementonBetter Lawmaking2003/C 321/01. 17) Cf. Baldwin/Cave, Understandingregulation: Theory,Strategy,and Practice.(1999); Ayres/ Braithwaite, Responsive Regulation,(1992); Hans-Bredow-Institut, Study on co-regulation measures in themedia sector,availableat: http://ec.europa.eu/avpolicy/info_centre/library/studies/index_en.htm ;Held/Schulz, Regulated self-regulation, available at: http://www.hans-bredow-institut.de/publikationen/apapiere/8selfreg.pdf

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 84 SEARCHINGFOR AUDIOVISUAL CONTENT

Withregardtothe link between anon-State regulatory system andState regulation, onecan speak of co-regulation if thefollowingcriteria aremet:

•Achievement of public policygoals •Legal connection •Discretionary powerofthe non-State regulatory system •State uses regulatory resources18

Co-Regulation requires special conditions. First of all, acertain regulatory cultureisneeded. This is especially importantfor models of co-regulation that dependonindustry associationstodraft codes or enforce rules,for example. Moreover, incentives forthe co-operation andenforcementofrules are of essential importance.The motivation forthe industry playerstoco-operate amongthemselves has to be adequately high. In most cases theincentivelies in theavoidance of State regulation: TheState musttherefore reduce itsregulatory powersignificantly.However, it muststill appear likely to the industry that theState will implementits ownregulation, if afailureofnon-State regulation becomes apparent.

Besidethe incentives,afunctioningco-regulatory system still has to enforce theregulation by implementingadequate andproportional sanctions. Another condition foraco-regulatory system is that theState shouldretain resources that areused to influence theoutcomeofnon-State regulatory processes.These resources arenecessary,because theState remainsresponsiblefor guaranteeinga sufficientlevelofprotection forthe respectivepublicgoal to be achieved. In this regard, acertification procedurefor codes and/or non-Statebodies has proven efficient. Other resourcescouldbethe delegation of memberstonon-State bodies,financial incentives or publicity. Finally,aclear legal basis andtask-sharingare essential forco-regulation. Aclear division of tasksisastrongincentivefor the industry to participate,whereas an unclear division of competences mightcause lack of transparency, increase therisk of “capturing” theregulator andleadtosmokescreens.

3.4. Examples forCo-regulation of the InternetinGermany

When it comes to theprotection of minorsinthe media, in Germanynon-State bodies have traditionally played an importantrole: i.e., they havebeen, andstill are, responsiblefor age- classification. Thefederal Jugendschutzgesetz (Federal Actfor theProtection of Minors–JuSchG) distinguishes between differentlevels of content: contentthat is harmful to children (“jugendgefährdend”)isclassifiedbythe Bundesprüfstellefür jugendgefährdende Medien (Federal Departmentfor Media Harmful to YoungPersons–BPjM). Thetask of theBPjM is to protect children andadolescents in Germanyfromany media which mightcontain harmful or dangerous content. The BPjM monitors, amongothers, thefollowingmedia: videos,DVDs,computer games,audio records and CDs,printmedia andInternet sites.Objectsare considered harmful or dangerous to minorsifthey tend to endangerthe process of developingasocially responsibleand self-reliantpersonality.Ingeneral, this applies to objectsthat contain indecent, extremely violent, crime-inducing, anti-Semitic or other racist material.Ifanobject is dangerous to youngpersons(“ jugendgefährdend “),the titleofthis object is placed on the“list of youth-endangeringmedia” (“Liste jugendgefährdender Medien“ -“index“). Distributorsofthat medium arethenforbiddenfromselling, rentingout or evenpresentingthis object in public or broadcastingit. Thecontentthat is harmful to minorsmustnot be shownatplaces to which children haveaccess anditmustnot be provided to children. In thefieldofmovies andvideo games,contentthat is notharmful to children, but capableofimpairingthe developmentofchildren (“entwicklungsbeeinträchtigend”), is rated by the Oberste Landesjugendbehörde (State Authorities Responsiblefor theProtection of Children).

Thisage classification is carried out in co-operation with non-State bodies: the Freiwillige Selbstkontrolle Filmwirtschaft (Voluntary Self-regulation of theFilm Industry–FSK)and the

18) Cf. Schulz/Held , Together they areStrong? In: Regulation, Awareness andHarmful Media Contentinthe Digital Age, Ulla Carlsson (Ed.) (2006), pages49 et seq. (pages51+52), availableat: http://unesdoc.unesco.org/images/0014/001469/146955E.pdf

©2008, European AudiovisualObservatory, Strasbourg (France) SEARCHINGFOR AUDIOVISUAL CONTENT 85

UnterhaltungssoftwareSelbstkontrolle (Self-regulation of EntertainmentSoftware–USK). Thereisa close link between theState andthe non-State bodies. 19

Followingthis tradition, search engines haveestablished acompletely voluntary self-monitoring system. Theincreasingpopularity of search engines andtheir continuous developmentleads to an increase in their significance forthe protection of youth in themedia as well. With this factinmind, manywell-knownsearch engineprovidersmet underthe umbrella of theabove-mentioned self- regulatory body,FSM, andfoundedthe Self-MonitoringofSearch EngineProvidersinFebruary 2005.20 This unionisbased on ajointCodeofConduct. Theaim of this Codeistoimprove consumer protection, as well as theprotection of children andyoungpersonsusingsearch engines in Germany. Underthis Code, search engineprovidersagree to explain thesearch engine’smodeofoperation to theuser.They also agreetostructuretheir searchresult pagestransparently.Furthermore, thesignatories to theCode shall endeavour,within their means, to enabletechnical precautions, which aresuitablefor promoting theprotection of children andyoungpersonsfromcontentharmful to them. Moreover, search engine providersintegrate theso-called “BPjM Module” provided by theBPjM into their search. Thespreading of pornographic andother harmful media contentvia theInternet is acriminal offence underGerman penal law.Pornographic contentonthe Internet is legal only if technicalmeasures prohibit minors from gainingaccess to theobject (Age Verification System or Adult-Check-System –AVS). Content which is harmful to minors(“ jugendgefährdend “) is placed on theabove mentioned “index”. The“BPjM Module” is alist of foreignInternet services that contain harmful contentkept by theBPjM andshared with theFSM. TheFSM can usethis list fortechnical precautionary measures (filtering) andthe providers, as mentionedabove,haveagreed to remove and/or nottoshow respectively anyURLs which areplaced on thelist by theBPjM, insofarasthe providershaveaccess to theURL andthe commercial expense is reasonable. In case of anyviolation of this CodeofConduct, sanctionsmay be imposed on theviolator.Search engines haveagreed nottoshow respectiveresultsontheir .deservices.

4. Conclusion: Regulation in the Field of Search Engines

Returningtothe list of riskstriggered by search engines andapplyingthe criteria that determine in which cases co- andself-regulation mightbeuseful, thereare someareas remainingwhere acloser examination mightbecrucial.

Hence, theGerman exampleshows that, in thefieldofprotection of minors, co- andself-regulation couldfunction.The samegoes forthe problemsofdiscrimination of contentand misleadingconsumers. Again, thevoluntary self-regulationofsearch engineprovidersinGermanyalready addresses someof theissues that concern thetransparency of theselection process (not of thealgorithmassuch). Regardingthe risk that search engines mightplay aroleinexploitingprotected (audiovisual) works or personal data, therealsoseemstobeatleast someleewayfor co-regulation.

However, when it comes to public policygoals, likecontrollingthe influence of public opinion making, andthe fragmentation of thepublic sphere, which mightbeaggravated by search engines, thereisnoincentivefor search engine providerstoco-operate.Moreover, thedistortion of competition andthe transfer of market powers is obviously notafieldinwhich it couldbeexpected that service providerswouldoffer their cooperation voluntarily.Inthese fields,ifany regulation is called for, it wouldbetraditional State regulation that wouldseem to be necessary.

However, when contemplatingdifferentregulatory approaches,one shouldkeep in mindthat search engine providersare notcompletely awareoftheir role in public communications. Theprovidersof Internet search engines havebeen focusingonelaboratingthe search technology.They operate on the

19) Cf. Schulz/Held, Together they areStrong? In: Regulation, Awareness andHarmful Media Contentinthe Digital Age ,Ulla Carlsson (Ed.) (2006), pages49 et seq.(57 et seq.), availableat: http://unesdoc.unesco.org/images/0014/001469/146955E.pdf ;onthe issue of self-censorship Held/Schulz, Der Index auf demIndex?Selbstzensur undZensur bei Suchmaschinen, in Machill/Beiler, Die Machtder Suchmaschinen –The Powerof Search Engines (2007). 20) Cf. http://www.fsm.de/en/Search_Engines

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 86 SEARCHINGFOR AUDIOVISUAL CONTENT

assumptionthat they arejustcreatingtoolsfor theusers’ hands andthat they havealimited role in structuringpublic communication. Theconcept of information intermediaries with special functions forpublic communication is relatively new eventothe academic debate.Thus, it shouldbeassumed that thesameapplies to personsworkinginthis field. Anyway,the new developments havetobeseen as an advantage,because correspondingprofessional ethics can still be established andlegal regulation mightalsoplay avital role in establishingthem.

Furthermore, all service providersare interested in externalisingresponsibility when it comes to makingchoices regardingthe content. Accordingly –touse thesameexampleagain –the list of sites containingharmful contentwhich is provided by theGerman BPjM has been warmly welcomed by the industry,for thesimplereason that it relives theindustry playersfromhavingtodecidewhich sites they havetoblock.

Finally,for providersoperatinginternationally thereislimited scope forrespondingtonational regulatory initiatives.Self- andco-regulation areabletoprovideflexibletoolsfor international cooperation, while, at thesametime, allowingfor global solutionsand tyingthe process in with national regulatory developmentpaths. Therefore, it remainsvital that thecodeallows forlocalisation (search engines providingfor variationsofalgorithm).

©2008, European AudiovisualObservatory, Strasbourg (France)

SEARCHINGFOR AUDIOVISUAL CONTENT 89

Searching for Audiovisual Content

Participants to the Workshop held on 12 April2008 in Amsterdam

C h r i s t i n a A N G E L O P O U L O S Institute forInformation Law (IViR), NL (rapporteur)

M ó n i c a A R I Ñ O Ofcom, UK

P i e r r e - J e a n B E N G H O Z I CRC, Ecolepolytechnique, FR

M a c h i e l B O L H U I S European PolicyCouncil, Google, EU

F r a n c i s c o C A B R E R A European Audiovisual Observatory,CoE

R a m ó n C O M P A Ñ Ó IPTS, ES

A n j a D O B B E L S T E E N Institute forInformation Law (IViR), NL (logistics)

N i c o v a n E I J K Institute forInformation Law (IViR), NL (moderator)

U r s G A S S E R University of St Gallen, CH

J a n e t G R E C O Broadcastprojects, UK

N a t a l i H E L B E R G E R Institute forInformation Law (IViR), NL

J o r i s v a n H O B O K E N Institute forInformation Law (IViR), NL (rapporteur)

E r i k J O S E F S S O N Electronic Frontier Foundation, US

E d d a n K A T Z Electronic Frontier Foundation, US

M i h a K R I S E L J APEK, SI

T a r l a c h M C G O N A G L E Institute forInformation Law (IViR), NL

M i l t o n M U E L L E R Syracus University,USA,Technical University,NL

S u s a n n e N I K O L T C H E V European Audiovisual Observatory,CoE

M i c h a e l R O T E R T EuroISPA, BE

T h o m a s R O U K E N S Telenet, BE

A l e x a n d e r S C H E U E R Institut forEuropean Media Law (EMR), DE

W o l f g a n g S C H U L Z Hans-Bredow-Institut, DE

©2008, EuropeanAudiovisual Observatory, Strasbourg (France) 90 SEARCHINGFOR AUDIOVISUAL CONTENT

D a v i d S T E V E N S ICRI, University Leuven, BE

R a l p h T R A P H O E N E R Empolis,DE

P e g g y V A L C K E ICRI, University Leuven, BE

E r i k V A L G A E R E N Stibbe,BE

A d a m W A T S O N - B R O W N DG Infosoc,European Commission, UE

M i c h a e l Z I M M E R University of Wisconsin-Milwaukee,US

©2008, European AudiovisualObservatory, Strasbourg (France) IRIS Special: Editorial Responsibility

European Audiovisual Observatory, Strasbourg 2008 ISBN: 978-92-871-6476-6 EUR 75

Director of the Publication: Wolfgang Closs, Executive Director of the European Audiovisual Observatory E-mail: [email protected]

Editor and Coordinator: Dr. Susanne Nikoltchev, LL.M. (Florence/Italy, Ann Arbor/MI) Head of the Department for Legal Information E-mail: [email protected]

IRIS Special: Editorial Responsibility Study carried out by: Dr. Wolfgang Schulz and Stefan Heilmann

Editorial Assistant: Michelle Ganter

Marketing: Markus Booms E-mail: [email protected]

Translation/Proof-reading: Caroline Bletterer, Michael Finn, Britta Probol, Nathalie Sturlèse, Candelaria Van Strien-Reney

Typesetting/Print: Pointillés, Hoenheim (France)

Publisher: European Audiovisual Observatory 76 Allée de la Robertsau F-67000 Strasbourg Tel.: +33 (0)3 88 14 44 00 Fax: +33 (0)3 88 14 44 19 E-mail: [email protected] http://www.obs.coe.int

Please quote this publication as: Susanne Nikoltchev, Ed., IRIS Special: Editorial Responsibility (Strasbourg, European Audiovisual Observatory, 2008)

© European Audiovisual Observatory, 2008. Opinions expressed in this publication are personal and do not necessarily represent the views of the Observatory, its members or the Council of Europe. Available in English, Special Serie French and German

What can you expect from IRIS Special What is the source of the IRIS Special in terms of content? expertise? Published by the European IRIS Special is a series of publications from the European Every edition of IRIS Special is produced by the European Audiovisual Observatory that provides you comprehensive Audiovisual Observatory’s legal information department in Audiovisual Observatory factual information coupled with in-depth analysis. The themes cooperation with its partner organisations and an extensive chosen for IRIS Special are all topical issues in media law, which network of experts in media law. we explore for you from a legal perspective. IRIS Special’s The themes are either discussed at invitation-only workshops approach to its content is tri-dimensional, with overlap in or tackled by selected guest authors. Workshop participants some cases, depending on the theme. It offers: and authors are chosen in order to represent a wide range of 1. a detailed survey of relevant national legislation to facilitate professional, academic, national and cultural backgrounds. comparison of the legal position in different countries, for example IRIS Special: Broadcasters’ Obligations to Invest in IRIS Special - unique added value Cinematographic Production describes the rules applied by IRIS Special publications explore selected legal themes in a way that 34 European states; makes them accessible not just to lawyers. Every edition combines 2. identification and analysis of highly relevant issues, a high level of practical relevance with academic rigour. covering legal developments and trends as well as suggested While IRIS Special concentrates on issues and interactions solutions: for example IRIS Special, Audiovisual Media within Europe, it takes a broader geographical scope when the Services without Frontiers – Implementing the Rules offers theme so requires. a forward-looking analysis that will continue to be relevant long after the adoption of the EC Directive; Who needs to read IRIS Special? 3. an outline of the European or international legal If you hold a position of responsibility in the media industry, context influencing the national legislation, for example particularly cinema, broadcasting or new media, if you work for IRIS Special: To Have or Not to Have – Must-carry Rules a specialised law firm, for a regulatory authority or in media explains the European model and compares it with the and cultural policy making, or if media studies is your academic American approach. discipline, then the IRIS Special series is a must for you.

So far published in the series: n Editorial Responsibility n Tomorrow’s Delivery of Audiovisual Services SearchingSearching forfor ISBN 978-92-871-6476-6 - 50 pages - © 2008 – EUR 75 ISBN 978-92-871-5707-2 - 56 pages - © 2005 - EUR 35 http://www.obs.coe.int/editresp http://www.obs.coe.int/tv-future n Legal Aspects of Video on Demand n Political Debate and the Role of the Media Audiovisual ISBN 978-92-871- 6391-2 - 91 pages - © 2008 - EUR 89 Audiovisual ISBN 978-92-871-5675-4 - 125 pages - © 2004 - EUR 44 http://www.obs.coe.int/vod-legal http://www.obs.coe.int/political-debate n The Public Service Broadcasting Culture n Regulating Access to Digital Television ISBN 978-92-871-6188-8 - 184 pages - © 2007 - EUR 62,50 ContentContent http://www.obs.coe.int/psb ISBN 978-92-871-5401-9 - 127 Pages + 28 pages Glossary - © 2004 - EUR 44 n Audiovisual Media Services http://www.obs.coe.int/digital-tv-access without Frontiers n ISBN 978-92-871-6115-4 - 76 pages - © 2006 - EUR 58,50 Co-Regulation of the Media in Europe http://www.obs.coe.int/av-media-services ISBN 978-92-871-5141-4 - 140 pages - © 2003 - EUR 27 http://www.obs.coe.int/co-regulation n Broadcasters’ Obligations to Invest in Cinematographic Production n Jurisdiction over Broadcasters in Europe ISBN 978-92-871-5971-7 - 127 pages - © 2006 - EUR 62,50 ISBN 978-92-871-4854-4 - 64 pages - © 2002 - EUR 27 http://www.obs.coe.int/tv-invest-film http://www.obs.coe.int/jurisdiction

n To Have or Not to Have - Must-carry Rules n Television and Media Concentration ISBN 978-92-871-5897-0 - 52 pages - © 2005 - EUR 58,50 ISBN 978-92-871-4595-6 - 89 pages - © 2001 - EUR 27 http://www.obs.coe.int/must-carry http://www.obs.coe.int/media-concentration

European Audiovisual Observatory: tel. +33 (0) 3 88 14 44 00 - fax +33 (0) 3 88 14 44 19 - E-mail [email protected]

IRIS_Special_Searching for Audiovisual Content_Cover [6mm].indd 1 2/12/08 16:35:39