<<

The Social Web: Global Village or Private ?

Abstract Rhetorics of Web 2.0 emphasize the sharing of user Shyong (Tony) K. Lam generated content. But how much content is actually GroupLens Research openly shared? Is the Web really an open arena for University of Minnesota content, or more suited to sharing in small groups? Will Minneapolis, MN 55455 USA sharing change as more people become aware of [email protected] potential legal and social pitfalls? Will media services increasingly be used for personal archive? As designers, Elizabeth Churchill we need to understand how individuals are making Yahoo! Research decisions about what to share, and with whom to Santa Clara, CA 95051 USA share. We need to be cognizant of possible differences [email protected] in social and cultural norms among different populations. We present a sketch of sharing defaults on a number of well-known social sites, and of user practices in a photo-sharing, social networking site, Flickr. Our project is in its first phase, intended to scope a broader study of sharing practices and drive application design ideation.

Keywords Privacy, access control, user generated content, sharing, copyright, Web 2.0, visualization, social norms

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee Project/problem statement provided that copies are not made or distributed for profit or The world of Web 2.0 is quintessentially about user commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, generated content and social sharing [5]. If one is to to post on servers or to redistribute to lists, requires prior believe the rhetoric, the web is a vast, open digital specific permission and/or a fee. Copyright © 2007 AIGA | The professional association for design. space of grassroots creativity, personal publishing and

1 ISBN: 978-1-60558-308-2 content exchange. However, recent news reports have precipitating event occurs which may spur them to be emphasized the impact of corporate, legal, political and more cautious [6]: behaviorally this will manifest in personal views on the sharing and publication of user people “taking their content down” or “flicking the generated content1. Countries differ in laws and norms switch to stop sharing”. Previous interviews on access around legal/ethical issues and responsibility for control suggest people take their lead from what (they distribution of problematic content (e.g. copyrighted, believe) others are doing. Sensationalism (FOAF and obscene, or offensive)2. At the personal level, user newspaper horror stories) and social emulation generated content can lead to many levels of (imitating others for personal/social gain) are key vulnerability – from legal action to stalking to social factors in people’s decision making, as are minimal embarrassment3. cognitive effort (stick with the defaults) and lack of awareness of the risks involved [7]. What does this mean for the user experience? And what do we need to do as designers? What are the The business implications of this revolve around trust. business implications? While users’ decisions are personal and may be local and changing, people often assume some responsibility Decisions around user/personal content sharing are on the part of the site operators, and understandably enacted on an individual level as content flagging (e.g. are annoyed when they feel they have not been fully mature or offensive content), visibility (private/friends- informed of the risks as well as the benefits of sharing only/public) and licensing (e.g., Creative Commons, their content. Legality and rationality of judgment are public domain). Example dialogue boxes from the not useful arguments in assessing these reactions; if photo sharing site, Flickr that illustrate these choices we are not demonstrably addressing the issues, there are shown in Figure 0 and Figure 0 in the appendix. could be a projection of negligence. Thus, our longer- term project aims to deliver a deeper understanding of Studies imply people are often uncertain about the the factors involved in user content sharing decisions, consequences of their choices with regard to these and to elaborate design prototypes for new sharing options, and make decisions based on unclear or paradigms that will better scaffold user decision volatile mental models [1][4]. People can also be making. Our primary audiences are designers and dynamic in their choices: anecdotally we know people business managers of social sharing sites. tend to err on the side of trust and openness until a Specifically, our project goals are to address: (1) What 1http://www.cbsnews.com/stories/2007/05/14/tech/ are default options that are routinely offered and how main2798132.shtml are they explained to users? (2) What are the patterns 2http://www.iht.com/articles/ap/2007/06/15/business/ of content access setting that can be observed across NA-FIN-US-Flickr-Censorship. different sites to which we have access in our analyses? 3http://www.mnheadhunter.com/mh/2006/11/ (3) Are there cultural patterns that can be discerned minnesota_teen_.html that suggest alternative understandings and selections

2 ISBN: 978-1-60558-308-2

that we, as designers, need to be sensitive to and personal security, concerns for social embarrassment, design for; and (4) Are there visualizations that can be and/or content control for business purposes. presented to users to help them get a sense of what others are doing, what cultural norms are, and what It is difficult to infer users’ individual thought processes their legal responsibilities are [1]? For example, would and intentions given snapshots of usage or system people feel differently about their choices if they know data. Thus, a major challenge lies in interpreting most content on a social networking site is in fact system data and mapping it to user behaviors, intents, privately shared between friends and not openly public? and their evolution over time. Thus, only through the combination of detailed qualitative analyses and Background broader quantitative data analysis can we develop an Our team consists of a computer scientist and a social understanding of patterns and requirements to better scientist; both engaged in experience support people’s wishes and intents. research. We combine bottom-up activity log analysis and heuristic evaluation of social sharing sites. In For the first phase in the project, we have focused on future we will couple these analyses with interview and broad-scale analyses of online behavior that may offer survey data to address user intent and understanding insights into different cultures, age groups, or other and complement activity and site analysis. demographic groups – these broad brush sketches of user behavior and user demographics are derived to The project began in May 2007, and is ongoing. The help us with ideation of interface alternatives for first phase, activity data analysis and activity different subgroups in different local and national visualization, will be completed in August 2007, after cultural settings. In deriving our initial sketch of which the interview and survey studies will begin. We practice, our greatest challenge has been to obtain intend to iterate and evaluate visualizations of social sufficiently rich usage data that contain the information sharing practices to see whether these affect people’s we desire to analyze, but that do not violate users’ understanding of location-based norms for content rights to privacy. sharing, and aid in decision making processes. While gathering publicly visible information is, at worst, Challenge a technical exercise in web crawling and screen- Our challenge is to understand what factors affect scraping, obtaining usage information or non-visible people’s content flagging, access control and copyright information involves working through myriad political, choices, and how we can provide better support for ethical, and legal issues, and ensuring that users and understanding and selection according to the different their private content are not improperly disclosed. As needs people have. Certainly people have different such, data are often ill-suited to answer detailed views on public versus small group sharing, and research questions of the kind we have been posing. different assessments of the consequences of open Therefore, to honor user anonymity but address the content sharing. Sharing assessments involve fear for research issues, we have proposed a data triangulation

3

method: site default analysis to assess the role of Flickr Photo Sharing Distribution Content Default Site options on user behavior; analysis of online activity Type Visibility % 70% .8 data to provide a shadow of people’s behaviors; and 9 Flickr Photos Public 5 upcoming survey data analysis to address intent. 60% s r

Youtube Videos Public e 50% s U

Social Public to f 40% o

Solution t n 30% Network institution e % c .2 r 5 e 20% % 1 Social P .0 A. Process 8 % % % MySpace Public 4 % % % % % % 3 .1 10% . .4 .2 .2 .2 .4 .8 . 4 Network To gain a better understanding of the various privacy 2 1 1 1 1 1 1 2 0% del.icio.us Bookmarks Public options offered to users, we began by selecting a set of e ll n % % % % % % % % % % A o 0 0 0 0 0 0 0 0 0 0 Research N -1 -2 -3 -4 -5 -6 -7 -8 -9 0 web sites that primarily consist of user generated 0 0 0 0 0 0 0 0 0 -1 Connotea Public 1 2 3 4 5 6 7 8 0 References content (photos, videos, bookmarks, articles, 9 Percent of Photos Publicly Shared By User Weblogs Public references, social networks, etc.). For each site, we News identified the default privacy and visibility settings for Digg Public Figure 0. User distribution of public photo sharing. Articles contributed or published content and the available Yahoo! Group settings for content. by default. Most sites offer the ability to selectively Public Groups Discussion restrict access to a smaller group of people, although Evite Event Info Restricted Following this, we analyzed data from Flickr, a site that the granularity of control varies among sites. Even sites that fundamentally offer the same functionality Upcoming Event Info Public allows users to store, organize, and share photographs online. Colloquially known as a Web 2.0 pioneer, there can have very different sets of privacy controls. An Table 1. Content types and default is an underlying expectation that many Flickr users example of this can be seen in comparing the visibility settings for popular social openly share their photos with the world, and indeed, providers Vox (vox.com) and Blogger (blogger.com). media sites. any photo that is published to Flickr is, by default, Vox allows access control settings to be defined per visible to everybody. We divided the set of Flickr users post, but Blogger only offers access control for the blog into groups (demographic, regional, etc.) and examined as a whole, thus making it an all-or-nothing choice. differences in their aggregate sharing behavior as measured by the average percentage of photos publicly Considering Flickr specifically, it is possible to restrict shared by each user. Our analyses do not entail access access to a photo to the owner’s friends, family, or just to photo content nor to detailed user information. Only the owner, as well as flag a public photo such that it is users with at least 25 photos in their collection are less easily discoverable. It is also possible to indicate included in our analysis. one’s willingness to share a photo even more openly by assigning a Creative Commons license to it. Such a B. Solution, Preliminary Results license grants others permission to copy and adapt the Table 1 shows a subset of the sites and features that photo to their own needs. Figure 0 and Figure 0 show we looked for. Unsurprisingly, the vast majority of the interfaces for making such decisions for a photo. current sites make contributed content publicly visible

4

Flickr Photo Sharing By Age Flickr Photo Sharing By Connectedness d d e e 95% 100% r r a a h h S 90% S

95% y y l l c c i i l 85% l 90% b b u u P P

s 80% s 85% o o t t o o h 75% h 80% P P

f f o o

70% 75% % %

g g v 65% v 70% A A 10 20 30 40 50 60 70 0 1 2 4 8 16 32 64 128 256 512 Age Number of Contacts

Figure 0. Public photo sharing on Flickr by age. The Figure 0. Public photo sharing on Flickr by number of yellow line at 79% represents the aggregate sharing contacts. behavior of users who do not report their age.

On Flickr, about 20% of all photos uploaded are non- openness, which suggests a higher propensity to share public; that is, access to these photos has been photos with the world. There were no meaningful restricted to some subset of people. Figure 0 shows gender differences observed. On the other hand, when the distribution of Flickr users with respect to how looking at differences by age, there is a decline in much of their photo collection is publicly shared. As one public sharing observed as age increases. This trend might expect, there is a heavy skew toward the default can be seen in Figure 0. We hypothesize that this is setting (all photos shared) – nearly 60% of Flickr users due to less Web exposure and more conservative risk simply share their entire collection of stored photos. assessments among older users. However, note that 40% of users do adjust their photo visibility settings to a wide variety of extents. The left- Like many other sites centered around user-generated most data point on the chart indicates roughly one in content, Flickr also provides social networking twelve users does not share any photos with the public. functionality wherein each user can specify a list of other users as contacts. Subsets of the contact list can Users who provided any demographic information be designated as also being friends and/or family, (gender, age, or location) shared more of their photo which are used in conjunction with the photo access collections with the public compared to users who control settings to determine who can view a photo. withheld such information. The yellow line in Figure 0 indicates that people who did not report their age to Since Flickr contacts are inherently used as an access Flickr are among the least willing to share their content control mechanism, one might hypothesize that people publicly. This is not surprising, as the act of providing who maintain contact lists would also be more likely to personal information is, in and of itself, a sign of have some non-public photos. Indeed, as shown in

5

Figure 0, people who have a small number of contacts time, to ascertain what the level of volatility is in the do have a lower percentage of public pictures than data – that is, are people changing the settings? Do users with no contacts. However, the trend is reversed people briefly allow pictures to be seen and then take as the number of contacts rises; having many contacts them down? Are there patterns (seasonal, news seems to indicate a greater willingness to share. related) that can be discerned as to how people make their decisions. Our sharing maps will be used to drive We speculate that people with fewer contacts may be a series of visualizations to complement existing access part of small, closely knit groups using Flickr as a control options intended to give people a sense of private sharing space. Certainly many forms of social sharing practices in the large. sharing can be observed on Flickr – archetypes range from artists using Flickr as a hosting promotion site, to References peacocks promoting their skill or themselves, to special [1] Ahern, S., Eckles. D., Good, N., King. S., Naaman, interest groups, and to provisions of lightweight social M. and Nair, R. Over-Exposed? Privacy Patterns awareness in distributed loosely affiliated social groups. and Considerations in Online and Mobile Photo Sharing. Proceedings of CHI ’07, New York: ACM One person may belong to several of these categories; Press. in informal discussions, people report having a number of Flickr identities, each used for different purposes. [2] DiGioia, P and Dourish, P. Social Navigation as a Model for Usable Security. Proceedings of SOUPS

’05, New York: ACM Press. Finally, to identify broad cultural differences, we [3] Good, N. and Krekelberg, A. Usability and privacy: produced the maps shown in Figure 0 and Figure 0 by a study of kazaa p2p file-sharing. Proceedings of dividing the users into groups based on their self- CHI ’03, New York: ACM Press. reported country or state of residence. A distinct [4] Gross, R. and Acquisti, A. Information Revelation difference in user behavior between the United States and Privacy in Online Social Networks. and Europe can be observed, with those in the United Proceedings of WPES ’05, New York: ACM Press. States generally sharing a larger proportion of their [5] O’Reilly, T. What Is Web 2.0? Design Patterns and photos publicly. We speculate that Europe’s greater Business Models for the Next Generation of privacy awareness as exemplified in more stringent , http://www.oreillynet.com/pub/a/ regulations contributes to more conservative selections oreilly/tim/news/2005/09/30/ at the individual level. what-is-web-20.html, last accessed 7.10.07. [6] Palen, L. and Dourish, P. Unpacking “Privacy” For a C. Next Steps Networked World. Proceedings of CHI ’03, New For the purposes of this preliminary analysis, we have York: ACM Press. presented a picture of the norm for openness in [7] Whalen, T., Smetters, D. K. and Churchill, E. F. defaults on social sharing sites, and analysis of sharing User experiences with sharing and access control. data from a single moment in time on Flickr. Our next In CHI 2006; NY: ACM; 2006; 1517-1522. steps will be to compare these data from a different

6 Figure 0. Flickr public photo sharing by country or US state of residence. Bright green points indicate a higher percentage of publicly shared photos, while bright red points indicate a lower percentage. Countries or states without points have insufficient user data.

Figure 0. Display of metadata, Figure 0. Several of the visibility and sharing Figure 0. Close-up views of as well as categorization settings available in Flickr. Europe and the United States. settings for a photo on Flickr.

7