Goals What is an agentbased model Why construct agentbased models What are the parts of an agentbased model How does one construct agentbased models What is an agentbased model What is an agentbased model ID: 1025743
Download Presentation The PPT/PDF document "Agent-Based Modeling for Behavioral Scie..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. Agent-Based Modeling for Behavioral Science
2. GoalsWhat is an agent-based model?Why construct agent-based models?What are the parts of an agent-based model?How does one construct agent-based models?
3. What is an agent-based model?
4. What is an agent-based model?Computer simulationAgentAgentAgentEnvironment
5. What is an agent-based model?Predator-prey modelsGrass
6. Why construct agent-based models?
7. Why construct agent-based models?What kinds of questions are agent-based models good for? How do they help with those questions?
8. What questions are agent-based models good for?Wilensky and Rand (2015):Integrative questionsDifferential questions
9. Integrative questionsHow does some psychology extrapolate out at group level??Racial BiasRacial InequalityConformityEmergence of norms
10. Differential questionsWhat psychologies can explain some observed phenomenon??Neighbor preferencesRacial Segregation
11. Mate choice exampleWhat is known:1. People have mate preferences2. People select mates3. People’s preferences do not strongly predict their choicesWhat could explain this?1. People do not act on their mate preferences2. People attempt to fulfill their preferences, but are constrained
12. Differential questionsWhat psychologies can explain some observed phenomenon??Constraints on mate choiceWeak preference fulfillmentIneffectual preferences
13. How do agent-based models help?An agent-based model is a way of:Expressing a hypothesisDeriving predictions
14. How do agent-based models help?Usually:Hypotheses: WordsPredictions: ReasoningHypothesis 1: “If people do not act on their preferences, then mates will not fulfill preferences”Hypothesis 2:“If people are constrained, then mates will not fulfill preferences”Agent-based model:Hypotheses: Computer codePredictions: ObservationHypothesis 1:Create world where agents do not act on preferencesObserveHypothesis 2:Create world where agents act on preferences but are constrainedObserveThen: compare predictions to data
15. Human reasoning is fallibleSometimes we reason incorrectlyTrivers, 1972Dawkins, 1976Coleman and Gross, 1991
16. Human reasoning is fallibleSome hypotheses are too complex to reason throughMate choice constraints:Availability of preference-fulfilling matesPresence of competitorsRequirement of mutual attractionHow much preference fulfillment should we see under constrained mate choice?
17.
18. Models facilitate communicationOur assumptions are often implicitComputer code forces assumptions to be explicitComputer code can be shared with othersVerbal arguments can be misleadingComputer code speaks for itselfComputer code can be manipulated
19.
20. What are the parts of an agent-based model?
21. What are the parts of an agent-based model?Key components: Agents (and Environments)Life cycle
22. AgentsWhat is an agent?Autonomous entity:PeoplePsychological mechanismsGroups of peopleEtc.Autonomous entity
23. AgentsWhat are agents like?Decision rules:If hungry, search foodIf find food, then eatIf not food, then dieIf alive, reproduceEtc.Autonomous entityAutonomous entityDecision rules
24. AgentsWhat are agents like?Features:Identities (Sheep vs. wolf)States (Hunger: 0 = Hungry; 1= full)Traits (Movement Speed: 0-10 steps/tick)Autonomous entityDecision rulesAutonomous entityDecision rulesFeatures
25. AgentsWhat are agents like?Environment:Other agentsSpaceOpportunities (food, shelter)Challenges (barriers, traps)Autonomous entityDecision rulesFeaturesEnvironment
26. What are the parts of an agent-based model?Autonomous entityDecision rulesFeaturesEnvironmentKey components: AgentsLife cycleLife CycleKey components: AgentsLife cycle
27. Life cyclesRepeated series of decisions and behaviorsBroken into discrete stagesEach stage:Agents behave/Environments updateTypically do so:In response to previous stageIn preparation for next stageAll agents pass through each stage in a specified order
28. Life cyclesForage: Move around and check for foodConsume: If you find food, eat itDie: Die if you didn’t find foodReproduce: Reproduce if you are still aliveForageConsumeDieReproducePredator/Prey
29. Life cyclesForageConsumeDieReproducePredator/Preyx100Data Collectionx100
30. How does one create agent-based models?
31. How does one create agent-based models?1. Develop a conceptual model2. Implement that conceptual modelLanguages and software
32. Conceptual ModelsA thorough but abstract description of your agent-based modelShould be complete enough for another person to write your model for youShould include:PurposeAgents and EnvironmentLife cycleAnalysis
33. Conceptual modelsPurpose1-3 sentences longWhat psychological process do we think extrapolates into a complex downstream phenomenon?What downstream phenomenon do we think is explained by some psychological process?What, roughly, will the model do?
34. Conceptual modelsPurposeTwo parts:1. One-sentence description of the purpose for writing this model2. Brief, high-level description of what will happen in the modelMate choice:“Determine whether human mate choice is better explained by preference-random mate choice or preference-driven but constrained mate choice. Agents will select mates from a simulated mating market incorporating realistic constraints; mate choices will be either random or based on simulated mate preferences”
35. Conceptual modelsPurposeTwo parts:1. One-sentence description of the purpose for writing this model2. Brief, high-level description of what will happen in the modelLife-Dinner Principle:“Attempt to model the life-dinner principle. Agents representing foxes will attempt to prey on agents representing rabbits; whether rabbits escape encounters with foxes will be determined by their speed relative to foxes. Rabbits will survive and reproduce only if they escape foxes; foxes will survive and reproduce in proportion to the number of rabbits they eat.”
36. Conceptual modelsAgents and Environment:A complete description of the agents and their environmentEverything necessary to generate your agentsShould include:Population size and characteristicsAgent featuresEnvironment features
37. Conceptual modelsAgents and Environment:A complete description of the agents and their environmentEverything necessary to generate your agentsPredator/prey model:“A population of 100 agents will be generated; 90 of these agents will be “sheep”, 10 will be “wolves.” Agents can be hungry or sated; all agents will start off hungry. Each agent will have a unique location in a 2-D field of grass, specified by X and Y coordinates. Agents will also have a food check variable that indicates whether there is food at their current location. For sheep, only grass counts as food; for wolves, only sheep count as food.”
38. Conceptual modelsLife cycles:A description, stage-by-stage, of what will happen throughout the life cycleMate choice:“1. Compute attraction 2. Select mates 3. Reproduce 4. Die #Computing attraction#Each agent will compute how attracted they are to all opposite-sex agents. Attraction will be calculated as the summed product of agent preferences and potential mate traits”
39. Conceptual modelsAnalysisA description of what model criteria will be analyzed and howPredator/prey:“At the end of each generation, the model will save (1) the amount of grass remaining uneaten, (2) the population size of sheep, and (3) the population size of wolves. After all generations are complete, the model will plot changes in these values across generations.”
40. Conceptual ModelsPurposeAgents and EnvironmentLife cycleAnalysis
41. Ease of useUtility/EfficiencyNetlogo
42.
43.