Type of Credit: Required
Credit(s)
Number of Students
In this lecture, I will talk to you about my personal exploration on the path of complex science and hence an authentic approach to thinking macroeconomic phenomena from the bottom up. In the late 20th century, a new approach to modeling emerged in the field of complex systems, known as agent-based modeling (abbreviated as ABM). In these fifteen lectures, we will introduce ABM not just as a unique tool in economic and financial simulation, but also with its embodied interdisciplinary concepts, which in turn are accompanied by the backdrop of the history of economic thoughts and modern economy (Week One). Alternatively put, ABM, in this class, will be introduced to you through an historical and interdisciplinary perspective.
From a very practical viewpoint, one distinctive use of agent-based modeling is on the policy and mechanism (market) design, especially when data support is poor and the degree of uncertainty (unknown unknowns) is high. A concrete example will be given in Week Twelve.
Throughout the lecture, two canonical kinds of ABM will be introduced to you, namely, cellular automata (CA) and evolutionary computation (EC). These models are canonical in the sense that they offer prototypes of (social) interactions. Using either kind of these two, the students can learn how to effectively represent the interactions of individuals at a micro level in his/her own model. From a retrospective viewpoint, one can say that the literature of ABM begins with the former kind, abbreviated as CA-ABM, but thrives with the second kind, shortened to EC-ABM. More details on the historical development of these two canonical kinds of ABM will be detailed in Week Seven, in which the idea of social networks is formally introduced.
The two kinds of ABM differ in their original scientific pursuit. CA-ABM originates from the quest for how to simulate life, as if we try to fabricate an android which can demonstrate life-like behaviors. This line of research was pioneered by John von Neumann (1903-1957) in his self-reproducing automata (Von Neumann, 1966). We shall get to this kind of models in Weeks Two and Three. The EC-ABM originates from the quest for how to simulate evolution, not just biological evolution, but broadly the evolution of the entire universe. We shall get to this kind of models in Weeks Twelve and Thirteen.
CA-ABM can also simulate evolution, but mainly at the macro level; EC-ABM not only aims to simulate the evolution of one level but also aspires to simulate the co-evolution among multiple levels; generally speaking, EC-ABM is Darwinian, relying on the mechanism of selection and variation. The idea of multi-level or hierarchical evolution, an essential characteristic of Neo-Darwinism, has been widely absorbed in the interplay between complexity theory and evolution theory (Simon, 1962; op Akkerhuis, 2016) as well as their joint influence on shaping the modern evolutional economics (Foster and Metcalfe, 2001). While the appearance of the idea of the co-evolution among multiple levels or the micro-macro level can be traced back to the origin of evolutional economics at the beginning of the last century, without tools like ABM modeling of this highly complex process is daunting if not impossible.
The two canonical kinds of ABM serve as fundamental tools to interpret and address macroeconomic dynamic processes as social epidemics. This idea was recently popularized by Robert Shiller (Shiller, 2019) (Weeks Six and Eight). Once this idea is introduced, we will revisit it repeatedly throughout the rest of this lecture (Weeks Twelve and beyond).
Rules (codes, genes, neurons, and programs), instead of just agents, are fundamental elements of ABM (Weeks One to Four). Rules for agents can be ascribed or acquired; for the latter, agents need to have the capability to learn and adapt, which is normally considered as part of the literature on bounded rationality. At this juncture, we observe the convergence of biology, psychology, computer science, machine learning, and artificial intelligence, making it much broader in scope than the conventional literature on bounded rationality (Weeks Six, Nine, Thirteen, and beyond).
References
Foster, John, and J. Stanley Metcalfe, eds. Frontiers of evolutionary economics: competition, self-organization, and innovation policy. Edward Elgar Publishing, 2001.
op Akkerhuis, Gerard AJM Jagers, ed. Evolution and transitions in complexity: the science of hierarchical organization in nature. Springer, 2016.
Shiller, Robert J. Narrative economics: How stories go viral and drive major economic events. Princeton University Press, 2019.
Simon, Herbert. "The Architecture of Complexity." Proceedings of the American Philosophical Society 106, no. 6 (1962): 467-482.
Von Neumann, John. Theory of self-reproducing automata. Urbana: University of Illinois Press, 1966.
能力項目說明
This class will cultivate students’ capacity for the so-called 'decentralized mindset,' in contrast to the conventional 'centralized mindset,' enabling you to think and model macro phenomena from the bottom up. For example, in Weeks Ten and Twelve, we illustrate the agent-based thinking of markets.
This class will enable students to discern the differences between agent-based modeling and equation-based modeling (Week Eleven), understanding that they complement each other rather than being in a competitive relationship. Hence, the students can combine the two in a more creative research framework to tackle with various macroeconomic issues (Weeks Fourteen, Fifteen, and Sixteen). As a continuation of this empowerment, this class also enables students to incorporate their skills in econometrics, exposure to big data, and experimental economics (laboratory with human subjects) into the aforementioned integrating framework (Weeks Four, Nine, and Sixteen).
This class will enable students to have an interdisciplinary view of economics.
| 教學週次Course Week | 彈性補充教學週次Flexible Supplemental Instruction Week | 彈性補充教學類別Flexible Supplemental Instruction Type |
|---|---|---|
General Specification:
(a) The student is expected to spend 12 hours per week on this course, which means 9-hour preparation and review work plus 3-hour class attendance.
(b) The assignment (the reading and the homework) will be given at the end of each ppt of the lecture.
(c) In this class, there will be a total of 15 lectures, and the remaining time will be reserved for the final exam (Week 18, January 8, 2023).
Weekly Progress
Week One (Lectured on September 11, 2023)
The Overview of the Class
Week Two (Lectured on September 18, 2023)
Canonical Agent-Based Model (I): Cellular Automata, Part 1
(a) Precursor I: John von Neumann (1903-1957) and his work
(b) Precursor II: John Conway (1937-2020) and Game of Life
(c) Precursor III: Stephen Wolfram and Elementary Cellular Automata
Week Three (Lectured on September 25, 2023)
Canonical Agent-Based Model (I): Cellular Automata, Part 2
(a) Thomas Schelling and the Segregation Model
(b) Variants of the Schelling Model
(c) James Sakoda and the Checkerboard Model of Social Interactions
(d) Robert Axelrod and the Model of Cultural Dissemination
Week Four (Lectured on October 2, 2023)
Agent-Based Modeling and Big Data
(a) Biology, Entomology, and Big Data
(b) Rule as the Fundamental Unit of ABM
(c) ABM and Experiments: ABM as a Laboratory
(d) NetLogo
Week Five (Lectured on October 9, 2023)
National Holidays (no class)
Week Six (Lectured on October 16, 2023)
Models of Social Epidemics (I)
(a) Social Interactions: Social Influences and Social Epidemics
(b) Social Epidemics and Narrative Economics: From Kenneth Boulding to Robert Shiller
(c) Market Sentiment: Model of Social Epidemics
(d) Learning and Adaptation: Kalman Filter Learning and Belief Formation
Week Seven (Lectured on October 23, 2023)
Canonical Agent-Based Model (I): Social Networks, Part 4
(a) Network-Based Agent-Based Models
(b) Spatial Games
(c) The Salient Break (the middle and late 1990s): Causes for Missing Networks
(d) The Second Generation (after the late 1990s)
(e) Small-World Networks and Market Efficiency
(f) Network Topologies and Cooperative Behavior
Week Eight (Lectured on October 30, 2023)
Social Epidemics (II): The Vriend Model of the Effect of Big Data
(a) Was Hayek an ACE?
(b) Rules: Choice-Making, Classifier System and Heuristics
(c) Wisdom of Crowds and Stupidities of Herds
(d) Ecology of Rules and Ecological Rationality
Week Nine (Lectured on November 6, 2023)
Reinforcement Learning
(a) Reinforcement Learning: Origins and Background
(b) The Multi-Armed Bandit Problem
(c) Roth-Erev Reinforcement Learning
(d) Reinforcement Learning in Auction Experiments
Week Ten (Lectured on November 13, 2023)
Agent-Based Idea and Modeling of Market
(a) A Spatial Agent Model of Prediction Markets
(b) What could agent-based markets mean?
(c) Zero-Intelligence Agents (Entropy Maximization)
(d) Stylized Facts Accounted: Favorite-Longshot Bias
(e) The Economics and Psychology of Personality Traits
Week Eleven (Lectured on November 20, 2023)
Agentization: From EBM to ABM
(a) Illustration 1: Lotka-Volterra Equations
(b) Illustration 2: Kermack-Mckendrick (SIR) models
Week Twelve (Lectured on November 27 2023)
Autonomous Agents: Agent-Based Lottery Markets
(a) Why Autonomous Agents?
(b) Market Design
(c) Agents
(d) Autonomous Agents
(e) Optimal Lottery Tax Rate
Week Thirteen (Lectured on December 4, 2023)
Canonical Agent-Based Model (II): Evolutionary Computation
(a) From Rule Given to Rule Discovery
(b) Three Intellectual Roots of Genetic Algorithms
(c) Representation
(d) GAs as a Model of Social Interactions
(e) Variants of Genetic Algorithms
Week Fourteen (Lectured on December 11, 2023)
Agentization of Macroeconomic Models (I)
Week Fifteen (Lectured on December 18, 2023)
Agentization of Macroeconomic Models (II)
Week Sixteen (Lectured on December 25, 2023)
Agentization of Macroeconomic Models (III)
Week Seventeen (Lectured on January 1, 2024)
National Holidays. No Class.
Week Eighteen (Lectured on January 8, 2024)
Final Exam
The course will be taught in English. The course will proceed in lectures. All lectures are prepared in power points, and the students can get these power points before or after the classes. Students are encouraged to use skype to interact with the instructor outside the classes. The students have to read the materials in advance. The lecture will only highlight the ppt. During the class, the students are invited to ask questions based on their readings of the preparatory materials and are also required to answer questions posed to them by the instructor. 40% of the score will be based on the in-class interacting performance of the student. The evaluation of the student performance will be based on the final exam (60%), and in-class interactions (40%).
The final exam will be open-book, and the student will be given two weeks for the final to work on it. So, for example, if the final exam is on January 8, 2023, at 18:00 pm sharp, then the student only needs to submit his/her answer before January 22, 2023,18:00 pm. Late submission will not be accepted.
Aoki M, Yoshikawa H (2006), Reconstructing Macroeconomics: A Perspective from Statistical Physics and Combinatorial Stochastic Processes. Cambridge University Press.
Bookstaber, R. (2017). The end of theory: Financial crises, the failure of economics, and the sweep of human interaction. Princeton University Press.
King, J. E. (2012). The microfoundations delusion: metaphor and dogma in the history of macroeconomics. Edward Elgar Publishing.
King, M., & Kay, J. (2020). Radical uncertainty: Decision-making for an unknowable future. Hachette UK.
Mackay, C. (1852/2012). Extraordinary popular delusions and the madness of crowds. Simon and Schuster.
Miller, J. H., & Page, S. E. (2009). Complex adaptive systems: an introduction to computational models of social life. Princeton university press.
Schelling, T. C. (1978). Micromotives and macrobehavior. WW Norton & Company.
Shiller, R. J. (2019). Narrative economics: How stories go viral and drive major economic events. Princeton University Press.
Wilensky, U., & Rand, W. (2015). An introduction to agent-based modeling: modeling natural, social, and engineered complex systems with NetLogo. MIT Press.