COGS 122: Agent-Based Modeling University of California, Merced Spring 2017

Transcription

1 Instructor Paul E. Smaldino, PhD COGS 122: Agent-Based Modeling University of California, Merced Spring 2017 Prerequisite: COGS 001 or permission of instructor Course Overview In this class we will cover the use of computational models to understand complex adaptive systems, with an emphasis on social and evolutionary phenomena. Such phenomena often involve feedback between individuals and their social environment, which can be difficult to parse. Building and analyzing simplified models of how social patterns emerge from individual behaviors can be an important step toward understanding. In this course, you will learn about agent-based modeling as a useful tool for anyone interested in social behavior. We will first cover how to appreciate and interpret models created by others, learning about important models in the history of complex systems. You will then learn how to build your own models, and how to analyze such models to draw concrete insights. What are Agent-Based Models? Agent-based modeling is a computer simulation strategy aimed at modeling the behavior and interactions among individuals in an environment. Agents can be any entities that behave and interact with each other. Typical agents are animals or humans but they can be any entity that interacts socially, from genes and neurons to firms and nations. Agentbased models (ABMs) are built by specifying the properties of agents and the rules by which they interact. Thus, they can easily include factors like individual differences and spatial (or network) structure. ABMs have particular value in explaining how populationlevel phenomena emerge from individual behaviors. Examples of such emergent phenomena are wide ranging and include cellular development, neural functions, insect foraging, fish schooling, bird flocking, traffic jams, ethnic segregation, and the evolution of social complexity. ABMs are increasingly used in for studying systems in many fields, including psychology, economics, political science, evolutionary biology, ecology, and sociology. Course Goals and Outcomes Course Goals: The goal of the course is to teach the conceptual foundations and methodology of agent-based modeling as it applies to understanding social behavior, and to provide you with hands-on experience in building and analyzing such agentbased models. Course Learning Outcomes: By the end of this course, you should be able to: (1) think coherently about causation in a complex system, and to communicate those thoughts 1

2 in writing and speech, (2) create an original agent-based model in NetLogo, and (3) analyze an agent-based model to draw clear conclusions. Relation to Cognitive Science Program Learning Outcomes. The entire course focuses on formal and computational approaches, and thus supports PLO 3 (Interpret and appreciate formal and computational approaches in cognitive science). Reading and reporting on seminal modeling papers, as well as directly implementing and extending some of them computationally, supports PLO 1 (Explain and apply knowledge of landmark findings and theories in cognitive science). Weekly writing assignments and a major research report supports PLO 4 (Argue for or against theoretical positions in cognitive science). Developing hands-on computational skills that are highly valued by many disciplines and industries supports PLO 5 (Use a cognitive science education outside the undergraduate classroom, particularly in the service of careers). Relation to Guiding Principles at UC Merced. COGS 122 supports the following principles for general education at UC Merced. (1) It supports Scientific Literacy by evaluating scientific theories and implementing cutting edge quantitative techniques. (2) It supports Decision Making by describing how social phenomena respond to different socio-ecological and individual choices. (3) It supports Communication as students learn to communicate complex ideas in writing. (4) It supports Self and Society by describing the complex interconnectedness of societies and how behaviors can feedback and cascade in social networks. (5) It supports Aesthetic Understanding and Creativity because creating a formal model of a social phenomenon is an inherently creative act, involving interpretation of the features of the world to tell a coherent narrative. (6) It supports the Development of Personal Potential by providing opportunities for individuals to implement their worldviews and examine the logical consequences of their assumptions. Software In this course we will be using NetLogo, a free, open source, and very user-friendly environment for developing and running agent-based models. It was specifically built to facilitate rapid learning, but is powerful enough to run research-quality simulations. You should download and install NetLogo as soon as possible, available from: There are also other software platforms for building ABMs, such as MASON and Repast, both of which use the Java programming language. You can also build an ABM from scratch using any programming language. In this course we will focus on models built in NetLogo. Technical requirements There are no specific technical prerequisites for this course. Although some models can be quite technically sophisticated mathematically and/or computationally, we will restrict our study to models that are relatively simple. In terms of mathematics, facility with algebra and basic probability theory is sufficient. In terms of programming, no prior experience is strictly necessary, though any such experience will help. Some familiarity with conditional (if-then) statements and for-loops will come in handy. 2

3 Readings The primary source for readings in this course will be: Wilensky, Uri & Rand, William (2015) An Introduction to Agent-Based Modeling: Modeling Natural, Social, and Engineered Complex Systems with NetLogo. MIT Press. This book (abbreviated IABM in the schedule) is available in the campus bookstore and will cover most of the topical and technical aspects of the course. However, the book is not a rigorous guide to using NetLogo. To supplement, students should also download the NetLogo User Manual, which is available in both PDF and HTML formats at (and abbreviated NLUM in the schedule). Some weeks, additional readings in the form of journal articles or book chapters will be assigned. Such readings will be made available online, through the course website. In addition, there are many fine popular books on the topics covered in this course. Here are some books I think are good. None of these are required for this course, but reading any of these during the course will give you a broader perspective on the subject, and perhaps spark an idea for a final project. Most of these books are broad in scope. If you are interested in reading more on a specific topic, ask me. I may have suggestions. Schelling, Thomas (1978) Micromotives and Macrobehavior. W. W. Norton & Co. Resnick, Mitchel (1994) Turtles, Termites, and Traffic Jams: Explorations in Massively Parallel Microworlds. MIT Press. Ball, Phillip (2006) Critical Mass: How One Thing Leads to Another. Farrar, Straus and Giroux. Buchanan, Mark (2007) The Social Atom: Why the Rich Get Richer, Cheaters Get Caught, and Your Neighbor Usually Looks Like You. Bloomsbury. Expectations What I expect from you Attendance. You are expected to attend the lectures, labs, and discussions. Skipping is not recommended for several reasons. This is a technical course, and you will almost certainly benefit from my instruction and assistance, and from watching your classmates struggle with and resolve their own problems. Material will be presented in class that is either not mentioned or discussed only fleetingly in the text. I also hope that classes will be fun and interesting, so that you will regret missing any. That said, you will not be directly penalized for being absent, and you do not need my permission to skip class or my forgiveness for having done so. Punctuality. Assignments should be turned in on time. This is in large part so that the TA and I can effectively schedule sufficient time to grade and provide feedback on assignments. Extenuating circumstances do occur. In such cases, please inform me and the TA as soon as possible. 3

4 Courtesy. Please attempt to arrive to class on time and plan to remain for the duration of the lecture to minimize disrupting others. Students should not engage in conversations unrelated to course materials, especially during lectures. Students should also refrain from using smartphones or other devices in ways unrelated to the class activity. You may think you are being discreet, but we all know what s up. . s to the instructor or TA should be written in a professional manner. Students are expected to check their regularly for any announcements. What you can expect from me I will answer all questions to the best of my ability, as will the TA. Please don't be shy about approaching us. If you feel offended or slighted, please tell me (in private) as soon as possible. s to the instructor or TA will be responded to within 24 hours, excluding weekends. Questions that are fully answered by the syllabus may be ignored. Evaluation Grades will be determined by performance on weekly homework assignments and a final project. Homework assignments (60%). For any technical skill, hands-on practice is crucial. There will be weekly homework assignments. These may vary quite a bit from week to week. They may involve writing, critical thinking, interpretation of assigned reading, mathematical analyses, and coding exercises. All homework should be turned in digitally, ed to both the instructor and the TA. Final project (40%). Beginning in week 10 of the course, you will begin to propose ideas for a final project. This will be an original modeling assignment. You will: (1) Select a topic related to social behavior. (2) Identify a problem or set of problems that can be addressed through modeling. (3) Perform a review of the literature on this topic, and identify previous model analyses related to your problem(s). (4) Create an agent-based model in NetLogo. This could be an extension of an existing model, including (but not limited to) one discussed in class, or it can be a wholly original model. (5) Analyze the model behavior. This involves presenting statistics, tables, and/or graphs that quantify the model behavior. You should identify key parameters of the model, and describe how the model behaves differently under different values of those parameters. Most importantly, don t forget to relate your results back to your original question(s)! (6) Write a paper detailing your project. This should include: a. An introduction where your topic and specific problem(s) are clearly identified. b. A literature review that describes previous attempts to address this or related problems. Notes that not all attempts need be agent-based models. 4

5 Verbal theories and empirical data (such as from experiments) can also provide supporting evidence. c. A non-technical description of the model. What is the idea here? What is the environment, and how do the agents behave and interact therein? What are the expectations or the theory you are presenting? d. A technical description of the model. Describe the algorithms used to make your model work. Pseudocode may be used, though it is not required. e. Describe your results. You encouraged to describe the model behavior in words. This description should be supplemented by data in the form of numbers, tables, and/or graphs. f. Discuss the implications and limitations of your work. This is your opportunity to put your individual stamp on the paper. What do you think about what you did? Did you learn anything? Was everything just as you expected? How might empirical data help to validate or invalidate the theory implicit in your model? What are some interesting questions still unanswered. I really want this section to include what you think about the project, so don t hesitate to get speculative. g. An acknowledgment of any assistance you received while working on the project. Science is highly collaborative, and we all do better work when we get help from others. It is important to acknowledge the role of that help in shaping your work. h. A list of references. Any citation style is fine (e.g. APA, MPA, etc.), as long as you stick to one and use it consistently. You will turn in your paper, as well as the NetLogo code for your model. There will be ample opportunity to get help on all aspects of your project throughout the course, and we will be devoting the final two weeks of class to work on this. Please take it seriously, and you may be surprised at how fun it is. A note on collaboration. The final project is an individual assignment. Every student will turn in his or her own project, with a different model and set of questions from anyone else. Nevertheless, communication and feedback are fuel for the engine of inquiry. You are highly encouraged to work together in pairs or groups, ask each other questions, show each other your work, and give respectful critical feedback on what your classmates are doing. If a pair or group becomes interested in a related set of problems, it is absolutely fine to work together and help each other, including sharing references and bits of code. Such help should be clearly acknowledged in a final Acknowledgments section at the end of your paper. That said, in the end each student will be wholly responsible for his or her own project, and each individual must ultimately focus on a problem that is distinct from that of the others. A note on programming languages. Throughout the course we will use NetLogo for programming agent-based models. If you have significant experience with other programming languages and wish to code your model in one of them for your final project, please see the instructor. I am very likely to approve your request. 5

6 Schedule Week 1: Introduction to ABM Reading: IABM, preface and chapter 0 Week 2: Why Model? Reading: Epstein, J (2008) Why model? Journal of Artificial Societies and Social Simulation 11(4), 12. Smaldino, P (in press) Models are stupid, and we need more of them. In: RR Vallacher, A Nowak, SJ Read (eds.), Computational Models in Social Psychology. Psychology Press. IABM, chapter 1 Week 3: Introduction to NetLogo Reading: NLUM, pp Week 4: Creating simple models Reading: IABM, chapter 2 Week 5: Analyzing models with NetLogo Reading: NLUM, pp Week 5: Exploring model behavior Reading: IABM, chapter 3 Week 6: Exploring model behavior (cont d) Reading: Schelling TC (1978) Sorting and mixing (ch 4). In Micromotives and Macrobehavior. WW Norton & Co. Week 7: Creating agent-based models Reading: IABM, chapter 4 Week 8: The parts of models Reading: IABM, chapter 5 Week 9: Model analysis Reading: IABM, chapter 6 Week 10: Biological patterns Reading: Bonabeau E (1997) From classical models of morphogenesis to agent-based models of pattern formation. Artificial Life 3, Sumpter DJT (2006) Principles of collective animal behaviour. Philosophical Transactions of the Royal Society B 361, Week 11: Finding mates and establishing dominance 6

7 Reading: Kalick SM, Hamilton TE (1986) The matching hypothesis reexamined. Journal of Personality and Social Psychology 51, Bryson JJ, Ando Y, Lehmann H (2007) Agent-based modelling as scientific method: a case study analysing primate social behaviour. Philosophical Transactions of the Royal Society B 362, Week 12: Opinions and polarization Reading: Deffuant G, Neau D, Amblard F, Weisbuch G (2001) Mixing beliefs among interacting agents. Advances in Complex Systems 3, Flache A, Macy MW (2011) Small worlds and cultural polarization. Journal of Mathematical Sociology 35, Week 13: The evolution of cooperation Reading: Axelrod R, Hamilton WD (1981) The evolution of cooperation. Science 211, Aktipis CA (2004) Know when to walk away: contingent movement and the evolution of cooperation. Journal of Theoretical Biology 231, Week 14: Work on final projects Reading: TBD Week 15: Work on final projects Reading: TBD 7