These agents aren't rational

Machine Dreams
January 23, 2003

Philip Mirowski has written an original and immensely scholarly book on the ways in which many economists have come to think about markets over the period from the 1940s to the end of the 20th century.

He starts with the dissonance between the theory presented in introductory courses in economics and what have become the principal concerns of mainstream economics. Much of the economics that is taught is about how fully rational agents allocate scarce resources optimally to given ends.

But the journals are full of different modes of agent behaviour, involving concepts such as Nash equilibrium, evolutionary games and network externalities. Noting this difference, the book offers a case study in the sociology of science - a close documentation of the reluctance of economists to abandon the concept of the economic rationality of agents even though the clear implication of their theoretical advances is that it is of little or no value.

The common theme linking many of the developments in economic theory in the second half of the 20th century is a concern about the capacity of agents - firms, individuals - to deal with market situations in a cognitive sense.

Mirowski argues that this has transformed economics from the study of the behaviour of completely rational agents to that of the economic agent as a processor of information. Economists need to accept this transformation. A particular implication is that the concept of the autonomous agent with fixed tastes and preferences makes little sense. Agents can only exist in, and be influenced by, society.

Mirowski believes that the emphasis on information processing has turned economics into a "cyborg science". His attempts to establish this concept require, in his own words, "prodigious documentation and explanation". The former is true, for the book contains well over 1,000 references as well as extensive material from more than 20 historical archives. "Cyborg" is a nice conceit, but it is one that is perhaps not established quite as clearly as the author hopes.

No matter, for there is a great deal of fascinating material on the development of economic thought. The fertile decades around the middle of the 20th century and the debates involving John von Neumann, John Nash, RAND and the Cowles Commission are documented in fine detail. Given game theory's importance today, a close examination of this period, during which the key concepts were developed, is very interesting. Mirowski's great hero is von Neumann who, he argues, wanted to abandon the concept of the rational, autonomous agent. A villain, in contrast, is Nash, whose concept of equilibrium now permeates economics. A neat story is von Neumann's immediate reaction when first shown Nash's work: "That's trivial, you know, it's just a fixed point theorem."

Mirowski's book is full of anecdotes such as this, which help to illuminate its themes. An amusing one is the opposition mounted to a 1953 proposal that candidate fellows of the Econometric Society should actually have done some empirical econometric work.

Of particular importance is the early testing done at RAND on Nash equilibrium in games, which pre-dates many of the 1990s findings of experimental economics by decades. For example, RAND secretaries were offered a choice: accept either a fixed cash prize or a large sum that had to be shared with another. The Nash solution was almost never found.

Instead, the secretaries appealed to the concept of fairness in dividing the spoils. The prisoner's dilemma was tested in 1950, with repetitions being played by a leading economist and a key mathematician. The prescribed Nash equilibrium emerged in less than a quarter of the plays.

The book contains a vast amount of material, which is reflected in the dense style. The sheer volume of information presented does not always allow the key themes to be followed as clearly as they might be. Further, the nature of the material, involving advanced concepts in economic theory, makes it hard for non-economists to follow. But, despite the problems of presentation, this is an important book that all economists who are genuinely interested in the future of their discipline should read.

Paul Ormerod is director, Volterra Consulting.

Machine Dreams: Economics Becomes a Cyborg Science

Author - Philip Mirowski
ISBN - 0 521 77283 4 and 77526 4
Publisher - Cambridge University Press
Price - £70.00 and £26.95
Pages - 655

Please login or register to read this article

Register to continue

Get a month's unlimited access to THE content online. Just register and complete your career summary.

Registration is free and only takes a moment. Once registered you can read a total of 3 articles each month, plus:

  • Sign up for the editor's highlights
  • Receive World University Rankings news first
  • Get job alerts, shortlist jobs and save job searches
  • Participate in reader discussions and post comments
Register

Have your say

Log in or register to post comments