Presentation is loading. Please wait.

Presentation is loading. Please wait.

Modeling Human Decisions in Coupled Human and Natural Systems: Review of Agent-Based Models Li An San Diego State University Mapping and Disentangling.

Similar presentations


Presentation on theme: "Modeling Human Decisions in Coupled Human and Natural Systems: Review of Agent-Based Models Li An San Diego State University Mapping and Disentangling."— Presentation transcript:

1 Modeling Human Decisions in Coupled Human and Natural Systems: Review of Agent-Based Models Li An San Diego State University Mapping and Disentangling Human Decisions in Complex Human-Nature Systems AAAS Symposium, Washington, D.C. February 18, 2011

2 Coupled Human and Natural Systems (CHANS) Introduction Methods ResultsConclusion  Heterogeneity  Nonlinearity and thresholds  Feedback/adaptation  Time legacy…

3 Agent-based Modeling Introduction Methods ResultsConclusion  What it is?  Individual-based  Mimic real world processes  Agents & environment  Strengths of ABM:  Modeling individual decision making  Incorporating social/ecological processes, structure, norms, and institutional factors  Incorporating multi-scale and multi-disciplinary data  Mobilize the simulated world

4  Many consequences  Economic  Environmental  Sustainability  Multi-disciplinary in nature  Psychology (e.g., cognitive maps)  Sociology (e.g., organization of agents)  Political sciences (e.g., game theory)…  Difficult to model at local levels Modeling Human Decisions Introduction Methods ResultsConclusion

5 1.What methods have been used to model human decision-making and behavior? 2.What are the potential strengths and caveats of these methods? 3.What improvements can be made to better model human decisions in CHANS? Objectives Introduction Methods ResultsConclusion

6  Paper search  Web of Science (key words attached in the paper)  Personal archives of agent-based modeling papers  Descriptive statistics  A total of 152 papers reviewed  Early models in 1994  Exponential increase over time Review Methods Introduction Methods ResultsConclusion

7  Humans make decisions to maximize revenues or returns ($)  Humans optimize a certain utility-like functions Microeconomic Models Introduction Methods ResultsConclusion Option 1 Option 2… Option n Option 2 > Option n > … > Option 1 Take Option 2! Rationality  bounded rationality Effects of non-monetary variables: how to account for? What function?

8 Space Theory-based Models Introduction Methods ResultsConclusion  Absolute space theory  Relative space theory Soil? What relationships? Linear? Weights of the space variables?

9  Cognitive maps or abilities (e.g., memory, learning, innovation)  Social norms, beliefs, perceptions, or intentions  Reputation of other agents… Cognitive Models Introduction Methods ResultsConclusion Fear FoeClose FoeFar + - Evasion + Gras R, Devaurs D, Wozniak A, Aspinall A.. 2009. Artificial life 15:423-63. Quantification of these abstract concepts Psychological theories for building their relationships (Gras et al. 2009)

10  Closely linked to the above cognitive models  Institution can explain why there are similarities across agents Institution-based Models Introduction Methods ResultsConclusion Economic returns, utility, cognitive measures Hard to code some institutions!

11  Effective real-world strategies  Can be articulated  Inductively derivable from observations  Variants: artificial intelligence, expert knowledge, and fuzzy logic… Experience- or Preference-based Models Introduction Methods ResultsConclusion There could be theories that explain such experiences or preferences Simple, straightforward, and self-evident; overuses make ABM less mechanistic

12  No theories or other guidelines  Black- or grey-box data-driven approach (e.g., neural network or decision tree )  Go through relatively complex data compiling, computation, and/or analysis.  Variants of this approach  Agent typology approach  Participatory modeling Empirical- or Heuristic Models Introduction Methods ResultsConclusion

13  Computational processes similar to natural selection  Agents carry a series of numbers, characters, programs, or strategies (chromosomes)  Multiple parental strategies compete and evolve to produce offspring strategies (copying, cross-breeding, and mutation)  higher fitness (intricate f(x) )  Calculate approximated f(x) through fitting the data Evolutionary Programming Introduction Methods ResultsConclusion A special type of empirical- or heuristic models Computationally intensive Consistent with findings from general econometric models

14  No data or theories exist  Adopt hypothetical rules (likely based on common knowledge or experience)  Calibration: let the outcomes of the model decide what rules are good Hypothetical and/or Calibration- based Models Introduction Methods ResultsConclusion Not all the possible candidates are available Multiple rules or values, if subject to calibration, could cancel out each other Use it very cautiously!

15  Not meant to be exclusive  Balance between simplicity and complexity when modeling human decisions in CHANS-related agent-based models  The KISS rule: “Keep it simple, stupid” (Axelrod 1997)  Develop mechanistic and/or process-based models (feedbacks, adaptation of decisions, and other complexities)  Develop protocols or architectures in modeling human decisions in CHANS:  By different types (e.g., agents, decisions, objectives…)  Hybrid  Advancements in other disciplines Conclusions & Discussion Introduction Methods ResultsConclusion

16  Sarah Wandersee  Ninghua Wang  Alex Zvoleff  Gabriel Sady  National Science Foundation PIRE Program  Visit the Space-Time Analysis of Complex Systems (STACS) Group at http://complexity.sdsu.edu/ Lan@mail.sdsu.edu Acknowledgements Questions?


Download ppt "Modeling Human Decisions in Coupled Human and Natural Systems: Review of Agent-Based Models Li An San Diego State University Mapping and Disentangling."

Similar presentations


Ads by Google