Question? Leave a message!




Social Simulation and Social Structure

Social Simulation and Social Structure
Intelligent Control and Cognitive Systems Social Simulation and Social Structure Joanna J. Bryson University of Bath, United KingdomTinbergen’s Questions Evolutionary (ultimate) explanations • Function (adaptation) • Phylogeny (evolution) • Proximate explanations • Causation (proximate mechanisms) • Development (ontogeny) •Replication Science Where there is a controversy / surprising • result, the first thing to do is try to replicate it. Fail → doubt on original result (and/or • yourself) Succeed → lets you test extend • theory.Replication Science Leading role for simulation in science: show • whether a parsimonious model really could explain the data. Goals of replication: • 1. validate experiment 2. extend results i. explain more data, or ii. understand model betterCase StudyCognitive Minimalism Rhesus Macaques picture: Bernard Thierry Egalitarian species show bilateral aggression, humanlike reconciliation. Research Question: Is cognition necessary or incidental to their social strategyWhy model monkeys Much better quantitative data than for • humans. Complete interaction statistics. • Not significantly affected by observers. • Understand our own origins and inclinations. • Political instability leading cause of ill health. •Macaque Social Order Some (e.g. Rhesus) show strict • dominance hierarchy; violent but infrequent conflict: “despotic”. Some (e.g. Tonkeans) show more • tolerance e.g. bilateral aggression; more frequent but less violent conflicts: “egalitarian”. van Schaik (1989), • Thierry et al. (2004)Bilateral Aggression Reconciliation Tonkean Macaques, an Egalitarian Species (video: Bernard Thierry)Two Hypotheses of Macaque Social Order Less resources (e.g. food) ⟹ • more violence ⟹ selective pressure for social structure (Hemelrijk 2001, 2002+). New conflict resolution • behaviour ⟹ less violence ⟹ less pressure for social structure (de Waal 2001, Flack de Waal).Hemelrijk’s Model Simple, cognitivelyminimalist boidslike • model. (Reynolds 1987; Hogeweg 1988) Despotic (vs. egalitarian) attributed to • greater variety in dominance rank value, consequence of aggression level. Sideeffect: dominants in centre of troop, • subordinates outside – like real troops. Convergent evidence for model. •Challenges Most researchers think something more • cognitive is going on with primates. Only scientific justification for a more • complex model is better match to data. Research question: Is there room to improve • on the match to dataDescribing a Model Bryson, Ando Lehmann (2007, 2011) Environment • Agents’ State • Agents’ Behaviour • Results Analysis •Environment Very simple torus: no • food or shelter, only space. Big enough with respect • to troop that agents couldn’t get lost look around see each other “around the world”.Hemelrijk’s Agents’ State Individual: DomValue (initially determined • by gender, changes by Eq. 2); X,Y position. Experimental Condition: StepDom • (aggression)–determined by gender and species; 2002 only: attraction (boolean: ♂⇒♀). • Statics: field of view; near view; max view; • personal space.Hemelrijk’s AS Separation Cohesion No AlignmentInteraction Equations ⎡ Dom i 1 Random(0,1) Dom +Dom i j 1 ⎣ w = i 0 else " Dom i Dom =Dom + w − ∗StepDom i i i Dom +Dom i j 2 " Dom i Dom =Dom + w − ∗StepDom j j i Dom +Dom i j for bees Hogeweg Hesper 1988Hemelrijk 2002 “SelfOrganization and Natural Selection in the Evolution of Complex Despotic Societies”, Biological Bulletin, 202(3):283288 Difference between despotic egalitarian • only increase of aggression (StepDom). Increased tolerance of females during • tumescence due only to their attractiveness.Replication By Hagen Lehmann JingJing WangEgalitarian, Normal Dominance RankingsAttraction On Aggression High (Despotic)Dominance L–coefficient of variation; R–mean of ♂♀Number of Female Dominance Interactions (Lehmann et al. only)AnalysisAnalysis Methodology Understand (replicate) model. • Find assumptions (implicit or explicit). • Treat assumptions as predictions. • Test predictions against data. •1. If one agent defeats another that vastly outranks it in a dominance interaction, do the two agents immediately change ranks within the troop (Unexpected outcome results in dramatic effect, Equation 2.) 2. Does it take fewer interactions to advance rank in a ‘despotic’ species (StepDom in Equation 2.) 3. Within species, if a fight is more violent (e.g. if blood is drawn) does it have more impact on the dominance hierarchy (StepDom as ‘aggression’, Equation 2.)4. Are females more likely to engage in fights when they are in tumescent If not then this model cannot account for their increased dominance. 5. Do females only become dominant during their tumescence in despotic species 6. When an animal in an egalitarian species is clearly outranked by another animal, are those two animals’ interactions similar to two more nearly ranked animals in a less egalitarian species Science Requires Expertise longtailed macaques Teeth baring as a gesture of submission (Bernard Thierry)Checked Questions with Thierry Not enough data to check (because...) • Ranks almost never change. • 1. If one agent defeats another that vastly outranks it in a dominance interaction, do the two agents immediately change ranks within the troop (Unexpected outcome results in dramatic effect, Equation 2.) 2. Does it take fewer interactions to advance rank in a ‘despotic’ species (StepDom in Equation 2.) 3. Within species, if a fight is more violent (e.g. if blood is drawn) does it have more impact on dominance hiera rchy (StepDom as ‘aggression’, Equation 2.) Bryson, Ando Lehmann 2007, 20114. Are females more likely to engage in fights when they are in tumescent If No not then this model cannot account for their increased dominance. 5. Do females only become dominant Probably during their tumescence in despotic Not species 6. When an animal in an egalitarian species is clearly outranked by another Probably animal, are those two animals’ interactions similar to two more Not nearly ranked animals in a less egalitarian species Hemelrijk Replication Conclusions Problems with existing model: • Predicts too much dominance volatility. • Inverts observed female violence. • A different (more complex) model is • justified.From evolution lecture... Science as Evolution Evolution requires variation, reproduction • and selection. Variety of theories get taught. • Theories in new experiments bare some • resemblance to what got taught. Memory of scientists, peer review, • prediction success perform selection.Tinbergen’s Questions Evolutionary (ultimate) explanations • Function (adaptation) • Phylogeny (evolution) • Proximate explanations • Causation (proximate mechanisms) • Development (ontogeny) •Weirdly Common Mistake The simplest explanation “wins”. • But an ultimate explanation and a • proximate explanation explain different things. You can have–in fact you need at least two • “winners”.Two More Theories... 1. SocioEcological Theory (van Schaik 1989) Soc. structure responds to environment. • Key factor is interindividual distance. • Egalitarians further apart: better for foraging, worse for predation. 2. Phylogenetic Inertia Theory (Thierry ’04) Migration history and genetic drift. • Despotics observed to be further apart. •...Unified through Simulation Predation pressure does • select for despotism. More recently evolved • species do seem more able to shift to this structure. Individuals can be more • distant on average, but have a lower minimum distance. (Lehmann, PhD 2009)Two Hypotheses of Macaque Social Order Less resources (e.g. food) ⟹ • Winner more violence ⟹ selective pressure for social structure (Hemelrijk 2001, 2002+). New conflict resolution • behaviour ⟹ less violence ⟹ less pressure for social structure (de Waal 2001, Flack de Waal).What is Status For Less resources (e.g. food) ⟹ • Winner more violence ⟹ selective pressure for social structure (Hemelrijk 2001, 2002+). Status does not indicate • fitness, it’s only about conflict resolution (2008).Explanations of Dominance Ranks Reduction in severe fights Reduction in selective pressure 800 5.00 (Lehmann, PhD 2009) 3.75 600 2.50 400 1.25 200 0 0 1.25 0 2.5 5.0 7.5 10.0 0 2.5 5.0 7.5 10.0 Dominance difference willing to fight with Dominance difference willing to fight with 1. Dominance certainly reduces conflict by establishing priority. 2. Dominance can still also increase distribution of beneficial traits (genetic or memetic). Severe fights in 1200 cycles Mean quality after 1200 cyclesSummary AI simulation is a method of doing science. • Cannot be the sole method, must have data • about the real world. But can be a source of evidence, lead us to • better understand the plausibility and consequences of our theories.
Website URL
Comment