Agency and decision-making Flashcards
Simulating complex systems: definition
Accounting for individual behavior and interactions in models leads to unpredictable outcomes = complexity
Agent-based models
ABM =
- Population of agents
- Environment
- Within which they act and interact
You can model complex SESs
- need to use various methods from various disciplines
- Helpful for making decisions
Key components of models:
● Rate of agent creation
● Decision-making strategies
● Drivers of behaviour
● Agent types and characteristics
ABM: Agents
- computational representations of real-world actors
- Agents interact with one another & their environment based on a set of rules (which can include learning)
- have characteristics, usually goals and behaviors
- have constraints and labels
assumption about human behavior in ABMs
Utility maximisation generally used
Heuristics in ABM: definition
(Simple model of) decision making trees - how we act
IF-THEN-ELSE
determining agent rules and responses
Problems faced with in ABM
- How to define realistic goals?
- Humans : we dont always act rationally or maximise utility (i.e. tragedy of the commons)
- bounded rationality/not omniscient/don’t communicate with all other humans
-> especially problematic for future projections - Societal rules
- Trust, cooperation, commitments
Heterogeneity in ABMs
& examples of
- you can’t model the average person because there are plenty of different types of agents (i.e. not like in typical equation-based modelling)
- Behaviors of one type of agent may differ based on external conditions
- Numerous possibilities for classifying agents
- Adaptive or learning agents possible
Examples:
- different goals
- changes over time
What are human behaviors and decisions based on in ABMs?
- Surveys, interviews; quanitative
- Storylines; qualitative
- Past behavior eg. census data
Example of applications of ABM
- traffic regulation
- crowd management
- market and land-use development
- smart-grids
Emergent behavior in ABMs
- higher level complex behavior can result from simple rules defined at level of individual agents
- It does not depend on its individual parts, but on their relationships to one another the arrangement
–> The interaction of relatively similar agents could lead to segregation
Actor characteristics in ABM derived from Social surveys
From actors to agents in SES models by Rounsevell et al.
= agent attributes
● Can enable/restrain the actor, e.g. legal drinking age
● Can alter decision-making
● Interaction can be promoted or inhibited by attributes
Decision-making strategies in ABM derived from Social surveys
From actors to agents in SES models by Rounsevell et al.
Heuristics (if-then-statements)
● Utility maximization and bounded rationality (not all options are
always available to an individual, but from those options the choice is made considering the greatest utility for individual -> rational thinking, but limited)
–> Bounded rationality: creating a subset of all possible outcomes representing a reasonable amount of options which are then chosen rationally
● Learning and adaptation (evolutionary processes)
Agent typologies (through induction or deduction) about the following characteristics in ABM derived from Social surveys
From actors to agents in SES models by Rounsevell et al.
- Functional role of agents
- Preferences and interests
- Behavioural mechanisms
- Geographical location (resources, political situation)
ABM - Scaling and generalization for larger areas
From actors to agents in SES models by Rounsevell et al.
- Scaling out: Increasing input data
- Scaling up: Higher representational level -> aggregation of individuals
- Nesting: Higher level processes as aggregate models that influence agent behavior at lower scale levels -> using feedbacks and interactions
The process of agent-based modelling as an approach to scientific enquiry ?
From actors to agents in SES models by Rounsevell et al.
external findings, theory, literature, social survey
lead to
- formalize perspective of the system
- model as medium for discussion guide project design, and proposal, research questions, model design
- conceptual model structure (guide data collection and social survey)
- develop/implement model versions
- computational laboratory
- communicate results
