Why Animals Swarm for Swarms

Home , Swarming (honey bee), Swarm (simulation)

Until recently, following the crowd was not seen as a desirable goal in life. These days, however, everyone is talking about swarm intelligence. But are swarms really smarter than individuals? And what rules, if any, do they follow? With the help of new computational techniques, Iain Couzin from the Max Planck Institute for Ornithology in Radolfzell imposes order on the seeming chaos of swarms.

TEXT KLAUS WILHELM

aboon troops are strictly hier- tested ruler again, and the other troop archical with an alpha male members must make do with his left- ruling the roost. However, overs. “Even though individuals are signs of democracy have re- self-interested, it appears that democrat- cently been observed – for ex- ic principles still apply,” says Couzin. ample,B when the animals are hunting for food. “Even the uninformed animals FROM CHILDHOOD CURIOSITY in the troop participate in decisions TO CUTTING-EDGE RESEARCHER about possible food locations and the route the troop will take to find them. As a researcher, Couzin’s heart beats This may be advantageous to the dom- faster when he tells stories like this. He inant male, as it means that he can ben- comes across like a young boy in the efit from the decisions of others when process of discovering the world, but locating good food sources,” explains the biologist has been carrying out cut- Couzin, whose department conducts re- ting-edge research for two decades and search at the University of Konstanz. regularly makes astonishing discover- However, upon reaching the food ies. His research passion: the swarm,

source, the alpha male is the uncon- that most beguiling form of collective. Photo: Axel Griesch

62 MaxPlanckResearch 2 | 17 Swarm for Swarms for Swarm Animals Why

Photo: Axel Griesch have shown that the animals are driven by the fear of being eaten. being of fear the by driven are animals the that shown have analyses His swarms. such of behavior the determines what discover to wants Couzin Iain area. one in live animals many too when individuals billion a over comprising swarms form locusts migrating nature, In ENVIRONMENT &CLIMATE_Collective Behavior 2 | MaxPlanckResearch 17MaxPlanckResearch 63 ENVIRONMENT & CLIMATE_Collective Behavior

A school of fish is a self-organizing system. Decisions are based » on the movements of the individual animals

“The beauty of swarms fascinated me nally have some in the laboratory imals in the Sahara some time ago, he even as a child,” says the 42-year-old again!” Couzin’s delight at this is evi- almost died of starvation. “I hallucinat- Scot. “I always wanted to know why and dent. He has had to get by without the ed,” he recalls, “and I thought I would how animals gather in large groups.” insects for a long time – too long for die.” In the end, all of the insects were Today, swarm has become a byword his taste. blown away by a sand storm, and Cou- for wisdom. Just 50 years ago, serious zin left Africa with no data. scientists fancied that telepathic forces HUNTING FOR LOCUSTS Because of this experience, he fo- were at work when, for example, thou- IN THE SAHARA cused on researching the insects in the sands of fish move and turn together as laboratory. With his team, he built a if by magic. Even when a swarm spon- Couzin has many scientific passions, circuit on which the locusts could taneously changes direction, order is but he is particularly drawn to the be- move as they pleased. Every morning maintained and there are seldom colli- havior of insects – perhaps because lo- the scientists released up to 120 ani- sions. The animals coordinate their custs have helped him reach some very mals into the perfectly secured circuit. movements considerably better than exciting research findings. By evening, however, it was found that humans driving in traffic. However, he almost paid a very high some locusts had disappeared. “This Birds and insects have similar skills. price for this: when he spent weeks went on for days and I started to doubt Consider locusts, for instance. “We fi- looking for the legendary swarming an- my own sanity, or at least my ability to count.” That was until he was able to examine video recordings of the go- ings-on in the arena in detail and made a startling discovery: the animals eat one another. After all, locusts are cannibals. Experts had considered them to be vegetarians and cooperative, for example when they suddenly form gi- ant swarms and, as one of the great Biblical plagues, strip entire swaths of land bare. “But their behavior has nothing to do with cooperation. Rather, they are driven by the fear of being cannibal- ized by others within the swarm,” explains Couzin. When many insects come together, they can run out of es- sential nutrients, and that’s when they begin to turn on one another. It was

Couzin (right) and his PhD student Jake Graving observe the swarm behavior of locusts in the laboratory. To this end, they built a circular

track from which the animals can’t escape. Photo: Axel Griesch

64 MaxPlanckResearch 2 | 17 on the movements of the individual animalsthe of movements the on based Decisionsare self-organizing system. a is fish of school A

Photo: Axel Griesch

Photo: Axel Griesch neighboring particlestochange their orientation ofone“particle” cancause magnets. Changesintheposition and align with each other much like small “flowing magneticfield.”Their bodies ticle physics.Theanimalsare akintoa that thelocustsfollowlawsofpar- tion, Couzin’s team later discovered form aswarm. locusts immediately lost their ability to longer feltthebitesfrombehind.The the insects’abdomenssothattheyno in byseveringthenerves ing histheory further experimentalevidencesupport- circles. Couzin and his colleagues found ing shortofgroundbreakinginexpert a “forced march”. sult isamobilecannibalistichordeon being eatenbythosebehind.There- eat the onein front of it, andto avoid found thateachindividualattemptsto swarms. fish of behavior the guides what predominantly study keep to shiners researchers golden and The shop. sunbleaks pet a in like aquariums with lined is Konstanz of University the at cellar fish The With theaidofacomputersimula- wasconsiderednoth- This discovery each individual in a swarm and recon - puter. These computer models trace construct animalswarmson thecom- Only by working together can they re- ologists tospeakthesame language.” to teach the computer experts and bi- mathematicians onhisteam.“We have computer scientists,physicistsand sights. To do this, he needs biologists, obtaining morecomprehensivein andinvirtualrealities,thus laboratory animals’ behaviorinthewild, – holistic,ifyouwill.Heanalyzesthe conventional approachtohisresearch This example illustrates Couzin’s un- LAB ANDON THE COMPUTER ANALYSES IN THE WILD,IN THE eral kilometers. in this way, even over distancesof sev- a swarmsynchronizetheirmovements position, too. The individual animals in - quently optsfortheleft. and fivetotheright,swarm fre- a majority. Ifsix fish swim to theleft tend toadoptthedirectionselected by group. Furthermore,individual fish influences themovementof theentire stimulus andthusdisproportionately school is often the first to react to a with them;afishontheedgeof of conspecificsbutwithoutcolliding schooling: theyseekouttheproximity dark. The fish follow clear rules when but theyremainmotionlesswhenitis ters. Indaylight,theyswiminschools, with alengthofaroundsevencentime tive ofNorthAmerica–asjuveniles, he studiesthegoldenshinerfish–ana group oforganisms:fish.Forinstance, fascinated byanotherswarm-forming the rulesthatgovernswarm. this way, thescientistsaimtodecode struct the animals’ fields of vision. In In addition to locusts, Couzin is also 2 | MaxPlanckResearch 17MaxPlanckResearch 65 > - -

66 id decision-makingandallows many to acollectivebrain.Thisenables rap- than eachofitsindividualmembers. therefore more effective as acollective “knowledge” withothers.The schoolis and allowsindividualstosharetheir quickly more, theprocessunfoldsvery where theschoolwillswim.Further tion changesultimatelydetermines act. Thesumofthepositionanddirec- more slowly, theirneighborsusuallyre- change direction,orsuddenlyswim explains Couzin.Ifindividualanimals movements oftheindividualanimals,” system. Decisionsarebasedonthe “So, aschooloffishisself-organizing likelyfollow.tion, thegroupwillvery their simulation slightly in onedirec- searchers steervirtualindividuals in the individuals.Forexample,ifre- attraction, repulsion and alignment of control thebehaviorofafishschool: Couzin identifiedthreefactorsthat With thehelpofhiscomputermodels, Sunbleaks glide through huge basins. Cameras record their every movement. every their record Cameras basins. huge through glide Sunbleaks MaxPlanckResearch 2 MaxPlanckResearch The swarmformssomething akin | 17 - ronmental cues,” says Couzin. ronmental cues,”saysCouzin. dent decisionsbasedonthe same envi- member ofthegroupmakes indepen cisions. “Accordingtothis view, each view thatbiggergroupsmake betterde- conventional teaching, which takes the prudent decisions. This contradicts the more make small groups big. Sometimes swarm can diminish if it becomes too cent studiessuggests,thewisdomofa However, as oneofCouzin’s mostre BETTER DECISIONS SMALLER GROUPS MAY MAKE plains Couzin. ly effective collective decisions,” ex- neighbors, individualscanmakehigh- responding tothemovementsoftheir to explicitlysignaloneanother. By viduals. Theanimalsdon’t evenhave simple interactionsbetweentheindi- plex swarmbehaviorcanarisefrom meld intoasingleentity. “Evencom- thousands oforganismstoliterally - - have access to similar sensory informa- have accesstosimilarsensory close proximitywithinagrouptendto sible tofulfillinreality. Individualsin But suchconditionsarealmostimpos- strong dominancehierarchy? Howdo the above-described baboons, exhibit a also applytoanimalgroups that,like itarian fishandbirdswarms, ordothey summarizing theirfindings. cisions itseemstomake,”saysCouzin, bigger agroupgets,thepoorerde- five and25membersworkbest.“Sothe medium-sized groupswithbetween most cases, they found that small to hold upwhenthisrealismisadded.In view ofthewisdomcrowdsdoesnot to demonstratethattheconventional and hiscolleagueAlbertKaowereable ulations ofdecision-making,Couzin making decisions.Usingcomputersim- modalities,when tiple cues,orsensory Furthermore, individualsmay use mul not be independent, but correlated. tion, sothecuestheyexperiencewill But do these laws apply only to egal- -

Photo: Axel Griesch Photo: Axel Griesch

Photos: Iain Couzin imals moveawayfromthe group, the are onthemove.“Whenindividual an decide onaparticularroute when they them todeterminehowthe baboons to optimizethesoftware, allowing studies, theresearchers werethenable Kenya. Usingtheresultsoftheirfield theanimalsonsitein team observed movements inreallife.Inaddition,the gram that accounted for thebaboons’ scientists created a new software pro GPS datatheyhadaccumulated,the from thewildbaboons. gan to testhismodels using the data versity to Lake Constance, Couzin be- his research groupfromPrincetonUni- puter. Immediatelyaftermovingwith was thenabletoevaluateonthecom- ond-by-second basis,whichWikelski abouts oftheprimatesonasec corded information about the where- Kenya withGPScollars.Thedevicesre- had previouslyfitted33baboonsin a distance.Forexample,thebiologist ments ofawidevarietyspeciesfrom ters withwhichhecantrackthemove- problem. Wikelski developstransmit- was abletoofferhimasolutionthis his futurecolleagueMartinWikelski Planck InstituteinRadolfzell in 2012, searchers tokeepup. rough terrainissimplytooswiftforre- a dauntingtask.Theirmovementover animals whentheyareonthemoveis for years.However, keepingtrackofthe a question that hasinterested Couzin Thisis wander throughtheirterritory? the animalsagreeonadirectionasthey different individuals. different indicate colors The swarm. a in shiners golden of movements Bottom Swimming vision. of fields the between overlap greatest the with positions in stay to prefer fish the that shows image The are. they brighter the vision, of fields the between overlap the greater The highlighted. are fish individual the of vision of fields The lab. the in shiners Top Camera image of a swarm of golden golden of swarm a of image Camera When CouzinvisitedtheMax To be able to evaluate the copious - - - dominant animalswasfound nottoin ity,” explainsCouzin. of thegrouptendtofollow the major- ferent directions, undecided members route theyprefer. Iftheyheadoffindif express theiropinionaboutwhich them ornot.Inthisway, theseanimals others havetodecidewhetherfollow Moreover, theroutechosenby ENVIRONMENT &CLIMATE_Collective Behavior - - accelerate theprocessoffinding acol- uninformed individuals facilitate and computational models,we findthat larly importantrolehere: “Inour ed groupmembersmayhave aparticu- plies withit.Theignorantand undecid the decisionofhisentourageandcom- process. Eventhealphamaleaccepts fluence thecollectivedecision-making 2 | MaxPlanckResearch 17MaxPlanckResearch 67 - ENVIRONMENT & CLIMATE_Collective Behavior

lective solution. If, in contrast, many animals have their own ideas about the PEOPLE AS SENSORS right route to take, it takes longer for the group to be able to agree on the direction they will all take.” Of all the species Couzin has studied to date, the movements of groups of people have turned out to be the most predictable. His team developed comput- CHOOSING THE RIGHT ROUTE er vision software for tracking the movements of individual people in large crowds. They record information about where the individuals look and how But what happens when a stalemate oc- they behave with each other. curs and the different views can’t be The software is intended to help avoid disasters like the Duisburg Love reconciled? When the same number of Parade in 2010. The scientists can use it, for instance, to understand and sim- animals want to go left and right? Then ulate how crowds of people move in a street. In this way, potential sources of it depends on the angle between the danger during the construction of new buildings and neighborhoods can be two preferred options: “If it is smaller avoided from the outset. than about 90 degrees, the animals fol- In one study, Couzin hired actors and set them the task of behaving in a lowing behind opt for the middle path, conspicuous manner on a shopping street and in a large train station. He and in the case of larger angles, they wanted to find out how passers-by would react to unusual behavior exhibit- exhibit a strong tendency to select the ed by individuals. His analyses show that important information can be direction preferred by the bigger sub- gleaned from the way people look. People do detect suspicious behavior, per- group. “That is exactly what we pre- haps because they consider it to be a potential source of danger. However, dicted with the model we created for they don’t usually do anything about it. fish,” says Couzin. Hence, like the fish, Couzin found that observing the gaze of pedestrians allows individuals the baboons also follow a majority- to be employed as sensors. Furthermore, he found that individuals in rules principle in their collective deci- crowds follow the gaze of nearby pedestrians. Using a computer program, sion-making, providing evidence that he and his team were able to analyze these gaze responses, allowing them this is a fundamental principle of swarm to detect suspicious behaviors. Such technology could provide useful early- behavior across species. stage indicators of potential hazards even before the people themselves More-recent studies by Couzin and raise the alarm. his team have now shown that the baboons also consider their physical environment when deciding which route to take. Using a drone, the researchers took a large number of aerial images of the animals’ habitat and, from these, created a high-resolution three-dimen- sional reconstruction of their habitat. This allowed them to relate baboon movement decisions not only to social factors, but also to the environmental structure. They found that the animals employ roads built by humans, and sometimes also paths beaten by other animals. “This enables them to move faster and more easily between the locations where they feed and sleep,” says Ariana Strandburg-Peshkin, a former doctoral student in Couzin’s depart- The lines of sight of individuals in a crowd of people can be recorded and evaluated using ment. Vegetation, for example, is a con- image analysis software. This enables the early identification of conspicuous behavior that may be indicative of potential danger. siderable obstacle to the animals’ movements. Groups slow down in densely

vegetated areas, and it is more difficult Photo and graphic: Iain Couzin

68 MaxPlanckResearch 2 | 17 Baboons live in hierarchical groups. However, important decisions are not dictated by the group leaders, but are made democratically. If the group members can’t agree on the route to be taken, the undecided apes follow the majority – irrespective of the direction chosen by the dominant animals.

able for research on collective behavior. The scientists working there will be able to observe animal swarms in virtual ho- lographic 3-D environments and mea- sure their movements precisely. Special- ly developed transmitters, sensors and image processing will enable the simul- taneous tracking of thousands of individuals in real time. Five smaller virtual environments where the researchers present real fish schools with a virtual world are already in operation today. Although the fish swim in empty pools, they see virtual rocks, water plants and predators. It will soon be clear whether the fish be- for the baboons to adapt to each other that the Center for Visual Computing have naturally in this environment or in the directions they take. of Collectives, a major new project he not. “It will be unique. Only in Germa- A third discovery was that, instead initiated in Konstanz, will uncover ny would it be possible to build such a of reacting to the positions of others, far-reaching new insights. research center,” says Couzin. But he baboon motion is more strongly influ- The Center is due to commence op- still wants to work with locusts again – enced by the routes taken by other ba- eration in three years’ time, and will be this time in a more relaxed setting, in boons within the previous five-minute one of the most modern facilities avail- the tranquility of Konstanz. period – baboons are effectively fol-

lowing in the footsteps of others. “And the greater the number of baboons that cross a certain point in this peri- TO THE POINT

od, the more attractive it becomes,” l The individual animals in a swarm often follow relatively simple rules when re- explains Strandburg-Peshkin. In this, sponding to others, such as a tendency to be attracted toward, and align direction the baboons somewhat resemble ants, of travel with, others while maintaining personal space and avoiding collisions. which recognize olfactory traces left l The principle of majority rule applies in many animal groups: members normally by members of their species along follow the direction in which the majority of their neighbors are moving. their routes. It isn’t yet entirely clear l Uninformed or unbiased individuals play a profound role in collective decision- whether the baboons also perceive making. They allow groups to make faster, and in some cases smarter, decisions. such scents or whether they simply observe and remember the paths taken by their “predecessors.” GLOSSARY Locusts, fish, baboons, ants – Cou- Swarm intelligence: When individuals contribute their skills to a collective, the group zin has examined a huge range of or- can develop characteristics that none of the individuals themselves have. This improves ganisms in detail and has repeatedly the performance of the group as a whole and can enable it to develop into a kind of superorganism. Examples of swarm intelligence include ant colonies and the internet. encountered similar laws and rules

Photo and graphic: Iain Couzin Photo: mhgallery/ 123RF governing collective life. He is hopeful

2 | 17 MaxPlanckResearch 69