A Simple Model For A Complex World?
Review of “Unlearnable Games and ‘Satisficing’ Decisions: A Simple Model for a Complex World”
Today we look at “Unlearnable Games and ‘Satisficing’ Decisions: A Simple Model for a Complex World,” by Jérôme Garnier-Brun, Michael Benzaquen, and Jean-Philippe Bouchaud, presents a compelling critique of the rational expectation framework in economics. We also keep a keen eye out for any work by Jean-Philippe Bouchaud. We believe Bouchaud is one of the finest Quant minds on the planet, see our list: Top 10 Quants On Wall Street 2023.
Bouchaud, Garnier-Brun & Bensazquen’s work introduces the “SK-game,” a binary choice model inspired by mean-field spin-glasses, to explore the feasibility of agents learning optimal strategies in complex economic environments.
Key Findings?
- Learning in Static Environments: The authors argue that even in a static environment, agents cannot learn collectively-optimal strategies. This outcome is attributed to the learning process getting trapped in sub-optimal fixed points or evolving endlessly without convergence.
- Impact of Memory and Reaction: The study reveals that long-term memory of past rewards aids learning, while over-reacting to recent past leads to detrimental cycles or chaos.
- Role of Competition: Increased competition destabilizes fixed points, initially causing chaos and, in extreme cases, leading to quasi-cycles.
- Randomness in Learning: Intriguingly, introducing some randomness in the learning process appears to improve collective decision-making.
- Emergence of Non-Stationary Behaviors: The model demonstrates spontaneous “aging” behavior in a substantial parameter space, despite the static nature of the world.
- Satisficing Solutions: Additionally, the paper underscores the idea of ‘satisficing’ solutions (as proposed by Herbert Simon). Where agents settle for satisfactory but sub-optimal rewards. This concept challenges the traditional view of rational agents making optimal decisions.
Analysis!
The paper effectively illustrates the limitations of the rational expectation paradigm in economics, especially in the context of complex situations. The authors’ use of the SK-game model serves as a robust tool to simulate the learning behaviors of economic agents. Their findings suggest that complexity in economic systems often renders optimal decision-making unfeasible, necessitating a shift towards satisficing solutions.
The exploration of factors like memory, reaction to past events, and the role of randomness provides deep insights into how agents might behave in real-world economic systems. These aspects highlight the intricate dynamics that govern decision-making processes, challenging the oversimplified assumptions of rationality in classical economics.
Implications?
This paper holds significant implications for economists and policymakers. It suggests a need to reconsider certain economic models that rely heavily on the assumption of rationality and perfect learning. By emphasizing the concept of satisficing solutions, the authors advocate for more realistic and adaptable economic models.
“Unlearnable Games and ‘Satisficing’ Decisions” is a thought-provoking study that challenges conventional economic theories with its innovative approach. It provides valuable insights into the complexities of learning and decision-making in economic systems, making a substantial contribution to the field of econophysics and behavioral economics. The paper is a commendable effort to bridge theoretical models with the nuanced realities of economic behaviors.
Read the full paper:
2312.12252.pdf (arxiv.org)
More Jean-Philippe Bouchaud:
- Why Monetary Policy Is Ineffective?
- What is the random matrix theory?
- Inflation Epicycles
- How does the Fama French model work? Does it really work?
- A Defense of the Nobel Prize Winners in Economics
