BotsimprovehumancoordinationinnetworkexperimentsYongrenShiHumanNatureLabYaleUniversity
2
ShiradoandChristakis.2017.Locallynoisyautonomousagentsimproveglobalhumancoordinationinnetworkexperiments.3
Socialnetworkexperimentwithonlinetools
ColorcoordinationgamesonsocialnetworksGoal:allplayersselectadifferentcolorfromlocalneighborsEachplayerisassignedanodetoselectacolor.Playerschangetheirnodecoloranytimetheywantastheyseetheirneighborscolors.Thegameisfinishedwhentherearenoconflictsinnetwork.Evaluation:timerequiredtosolvethecolorconflictsKearnsM,SuriS,MontfortN(2006)AnExperimentalStudyoftheColoringProblemonHumanSubjectNetworks.Science313:824–827.
Player’sviewinthegame
CommentsfromhumanplayersSomeoneisauselessidiot,anditsnotme.Thisgameisbullshit.Thanksforwastingmytime.Imgoingtocry.Gotscrewedonthisgame.Foolmeshameonme....youcantgetfooledagain.Iverymuchenjoyedthisgame.Itwasstrange,becauseIfeltpersonallyattachedtotheotherplayersthatIdontevenknow,letaloneknowwhethertheyexistornot.Ifeltdependent,becauseIwasrelyingalotonmyteammembers.Ialsofeltindependent,becauseIwasdoingmyownthing.(snip)IfeltlikeIwasintheMatrixandIwastryingtocrackacodeorsomething.Idprobablyplaythisgameforfree,itwassoenjoyable.
Howcanbotsacceleratethecoordinationprocess?
Everyplayerchoosesthebestcolorlocally,
butthegameisnotalwayssolved.105sec.“Unresolvable”conflict
“Noise”mightimprovesocialcoordination.
Designofbots’behaviorEvery1.5secRandomPrandomPickacolorrandomlyIsthereabettercolortoreducelocalconflicts?ChangetothebettercolorMaintainthecurrentcoloryesyesnonoThreelevelofbehavioralnoisePrandom=0%,10%,and30%
Behavioralnoise (0%,10%,30%)Geodesiclocation (random,center,periphery)Howcanbotsacceleratethecoordinatio