lecture1-s - computer science · – design systems to achieve desired outcomes • economic...
TRANSCRIPT
CS286r: Topics at the Interface between Computer Science and Economics
Fall 2008
Social Computing
Coursewebsite:h9p://www.eecs.harvard.edu/cs286r/
Computer Science
• is the study and the science of the theore@calfounda@ons of informa@on and computa@on andtheir implementa@on and applica@on in computersystems.[Wikipedia,Sep.2008]
– Buildingsystems– F(x)=?– Howfastcanwegettheanswer?– Focusoncomputa@onalandinforma@onalconstrains.
9/15/2008 cs286rFall'08 2
Economics
• isthesocialsciencethatstudiestheproduc@on,distribu@on,andconsump@onofgoodsandservices.[Wikipedia,Sep.2008]
– Economies(systems)
– Manyself‐interestedagents– Agents’preferences/u@li@esoveroutcomes
– Agents’informa@onandbeliefs
– Agents’decisionmaking– Game‐theore@cinterac@onsofagents
9/15/2008 cs286rFall'08 3
The Interface • Computersystemsareincreasinglybeingdevelopedand
usedbymul@plepar@eswithdifferentpreferences– Predictsystemoutcomes– Designsystemstoachievedesiredoutcomes
• Economicproblemssome@mesare(hard)computa@onalproblem– Resourcealloca@on– Pricediscovery
Theories, algorithms, and systems that satisfy both economic and computational constraints.
9/15/2008 cs286rFall'08 4
Lots of Compelling Applications
9/15/2008 cs286rFall'08 5
• InternetMone@za@on:
Google,Yahoo!,MSNareusingauc@onstosellads
• SocialnetworkandSocialTagging:Facebook,MySpace,LinkedIn,Flickr,LibraryThing
…
This Course
• Rota@ngtopiccourse– Fall2008.SocialCompu@ng
• Previous– Spring2008.Computa@onalFinance
– Spring2007.Computa@onalMechanismDesign– Spring2006.Mul@‐agentLearningandImplementa@on
– Spring2004.Itera@veCombinatorialExchanges…
9/15/2008 cs286rFall'08 6
Social Computing
• hastodowithsuppor@ng“computa@ons”thatarecarriedoutbygroupsofpeople,anideathathasbeenpopularizedinJamesSurowiecki'sbook,TheWisdomofCrowds.[Wikipedia,Sep.2008]
• Relatedterms:Collec@veintelligence,Web2.0,Folkonomy,…
Undertherightcondi:ons, crowdsaresmarterthanexperts.
9/15/2008 cs286rFall'08 7
Crowds Are Smarter…
• Whowantstobeamillionaire?– Fiky‐FikyCorrect50%ofthe@me
– Phone‐A‐Friend
– AsktheAudience
9/15/2008 cs286rFall'08 8
Crowds Are Smarter…
• Jelly‐Beans‐in‐the‐JarExperiment– ProfessorJackTreynorrantheexperimentinhisclass– withajarthatheld850beans– thegroupes@matewas871– onlyoneofthe56peopleintheclassmadeabe9erguess
9/15/2008 cs286rFall'08 9
James Surowiecki
9/15/2008 cs286rFall'08 10
VideoSource:youtube.com
Are Crowds Smarter?
• Noalways– Badcommi9eedecisions– Endlessgroupmee@ngs
• Inthiscourse,wefocusontheanalysisanddesignofInternet‐basedsystemsthatsupportcrowdstoachievesome“computa@onal”goals.
9/15/2008 cs286rFall'08 11
Structure of the Course
• Introduc@ontoGameTheory(4lectures)– Normalformgame,extensiveformgamewithperfectinforma@on,extensiveformgamewithimperfectinforma@on,andBayesiangames
• CurrentResearchPapers– Predic@onmarkets (Aggregateinforma@on)– SocialChoice (Aggregatepreference)– PeerProduc@on (Buildknowledgesources)– HumanComputa@on (Makehumandothework)– Reputa@onSystems (Determinera@ngs)– ScriptSystems (Facilitatetransac@ons)
9/15/2008 cs286rFall'08 12
Enrollment & Prerequisites • Enrollmentislimitedtoabout20students.Complete
Surveyatendofclass!• Prerequisites– Mathbackgroundisimportant!Atleastabasiccourseinlinearalgebra(suchasM21b,AM21b,orequivalent)
– Basicunderstandingofprobabili@es– Atleastanalgorithmcourse(CS124,orequivalent)– [email protected]/gametheorycourse.
Advancedcourseinalgorithms,microeconomics,gametheory,orlinearprogrammingarehelpfulbutnotrequired.
9/15/2008 cs286rFall'08 13
Grading
Problemsets 25% 2‐3homeworkproblemsets
Par@cipa@on 20% Readingpapers,submiqngshortsummariesandques@onsbeforeclass,andpar@cipa@oninclassdiscussion.(Note:Absentstudentsrarelycontributetodiscussions.)
Presenta@onofoneortworesearchpapers
15% [email protected].
Project 40% Projectproposal,classpresenta@on,andfinalreport.
9/15/2008 cs286rFall'08 14
Project
• Goal:developadeepunderstandingofaspecificresearchareaandtotheextendpossibletoworkonanopenresearchproblem.
• Canbetheore@calorexperimental.• Canbealiteraturereview,butneedsnovelty!• Alistofhigh‐levelprojecttopicswillbeprovided.Youareencouragedtoproposeyourowntopicforapproval!
• ProposalsduearoundThanksgiving.Presenta@onsduringreadingweek.Projectreportsattheendofreadingweek.
9/15/2008 cs286rFall'08 15
Logistics
• TFs– ShailiJai– DimitriosAntos
• OfficeHours– Yiling:Wed11‐12,Thu3‐4,MD339– Shaili:Tue2‐4,MD2ndfloorlounge– Dimitrios:TBD
MissedcoursematerialsfromtheTFs
9/15/2008 cs286rFall'08 16
Overview of Course Topics
• Predic@onmarkets
• SocialChoice • PeerProduc@on • HumanComputa@on
• Reputa@onSystems • ScriptSystems
9/15/2008 cs286rFall'08 17
CNBC Features Intrade, 8/25/08
9/15/2008 cs286rFall'08 18
Bet = Credible Opinion • Q:WillObamawinthePresiden@alelec@on?
• Beqngintermediaries– LasVegas,WallStreet,Bewair,Intrade,...
9/15/2008 cs286rFall'08 19
McCainwillwintheelec:on
Info
Ibet$1000Obamawillwintheelec:on.
Info
Prediction Markets • Apredic@onmarketisafuturesmarket(beqngintermediary)thatisdesignedforinforma@onaggrega@onandpredic@on.
• Payoffsofthetradeditemisassociatedwithoutcomesoffutureevents.
9/15/2008 cs286rFall'08 20
$1ifObamaWins
$0Otherwise
$1×Percentage of Vote Share That
Obama Wins
$1 if Patriots win
$0 Otherwise $f(x)
Prediction Markets
9/15/2008 cs286rFall'08 21
PM Problems to Discuss
• Socialinfluenceoninforma@onaggrega@on
• Newmarketmechanisms
• Manipula@onofmarketprices
• Combinatorialbeqng
• Distributedcomputa@on
9/15/2008 cs286rFall'08 22
A Combinatorial Betting Example
• 251outcomes,2251combina@ons
• Allowpar@cipantstobetonlogicalformulas– Createcontractsonthefly:$1ifOhioANDFloridaORNewYork,$0otherwise
– Specifybuypriceandquan@ty• Thematchingproblem:howtoclearthemarket• Computa@onallyhard!(co‐NP‐complete)
9/15/2008 cs286rFall'08 23
Voting
9/15/2008 cs286rFall'08 24
> >
> >
SocialChoiceMechanism
> >
Social Choice Problems to Discuss
• Hardnesstomanipulate– Almostallreasonablevo@ngrulesaresubjecttomanipula@on
– Computa@onalcomplexitybecomesagoodthing
• Fakediden@ty– New7WondersoftheWorld:“Jordan,withapopula@onof7million,managedtocastover14millionvotesfortheirarcheologicalsitePetra.”
• Rankingsystems
9/15/2008 cs286rFall'08 25
Manipulate Voting
9/15/2008 cs286rFall'08 26
> >
> >
SocialChoiceMechanism
> >
?
PageRank as A Social Choice Problem • PageRankisattheheartofsearchengines
• Nodesasagents;linksasvotes;PageRank=>findasocialranking
9/15/2008 cs286rFall'08 27
€
PR(u)∝ PR(v)L(v)v∈Bu
∑
Peer Production • Alargenumberofpeoplecoordinatedintocomple@ngmeaningfulprojects.
• Crowdsourcing
• Whatareregulari@esofpeerproduc@onsystems?• Howtokeepgoodstuffinandbadstuffout?
9/15/2008 cs286rFall'08 28
Human Computation • CoinedbyLuisvonAhn.• U@lizingthecomputa@onalabili@esofhumanbeings• Anexample:ESPGame–LabelingtheWeb
9/15/2008 cs286rFall'08 29
CAPTCHA: Completely Automated Public Turing test to tell Computers and Humans Apart
9/15/2008 cs286rFall'08 30
Reputation Systems
• FacilitatetrustinInternetinterac@ons
• Examples– eBay– Epinions– BizRate
9/15/2008 cs286rFall'08 31
Reputation Systems Problems to Discuss
• Valueofreputa@on
• Whitewashing
• Elicithonestfeedback
• Sybila9ack– PageRankcanbeconsideredasareputa@onsystem
9/15/2008 cs286rFall'08 32
Script Systems
• Ar@ficialcurrencysystemsthatcanbeusedingroupdecisionmaking,peer‐to‐peersystems,andetc.toachievebe9erefficiency.
• Howdoscriptsystemschangetheefficiencyandbehavioroftheoriginalsystems?
9/15/2008 cs286rFall'08 33
Next Class
Gametheory!
9/15/2008 cs286rFall'08 34