February 04, 2004

Postrel and Pinker push Hayek -- a bit too far

In article entitled "Friedrich the Great" (in the Boston Globe on 1/11/2004), Virginia Postrel writes:

Hayek's 1952 book, "The Sensory Order," often considered his most difficult work, foreshadowed theories of cognitive science developed decades later. "Hayek posited spontaneous order in the brain arising out of distributed networks of simple units (neurons) exchanging local signals," says Harvard psychologist Steven Pinker. "Hayek was way ahead of his time in pushing this idea. It became popular in cognitive science, beginning in the mid-1980s, under the names `connectionism' and `parallel distributed processing.' Remarkably, Hayek is never cited."

This paragraph makes one assertion which is wildly exaggerated (that "Hayek is never cited"), and an implication that is false (that Hayek's 1952 book suggested the basic idea of connectionism 30 years ahead of anyone else).

Asking google about "hayek cognitive science" turns up this page as the highest ranked result, which provides dozens of citations for books and articles that (are said to) discuss Hayek's contributions to cognitive science, including works by Edelman, Minsky, and Rosenblatt. Looking around a bit on the hayekcenter.org website, or asking google about "quote hayek neuroscience" or "quote hayek cognitive science", turns up this page, which includes many actual quotes, for instance:

"[Hayek] made a quite fruitful suggestion, made contemporaneously by the psychologist Donald Hebb, that whatever kind of encounter the sensory system has with the world, a corresponding event between a particular cell in the brain and some other cell carrying the information from the outside word must result in reinforcement of the connection between those cells. These day, this is known as a Hebbian synapse, but von Hayek quite independently came upon the idea. I think the essence of his analysis still remains with us . . ". (Gerald Edelman, Neural Darwinism, 1987, p. 25).

"Most theoretical work since the proposals of Hebb (1949) and Hayek (1952) has relied upon particular forms of dependent synaptic rules in which either pre- or postsynaptic change is contingent upon closely occurring events in both neurons taking part in the synapse." (Gerald Edelman, Neural Darwinism, 1987, p. 181).

"The first proponent of cortical memory networks on a major scale was neither a neuroscientist nor a computer scientist but .. a Viennese economist: Friedrich von Hayek (1899-1992). A man of exceptionally broad knowledge and profound insight into the operation of complex systems, Hayek applied such insight with remarkable succes to economics (Nobel Prize, 1974), sociology, political science, jurisprudence, evolutionary theory, psychology, and brain science (Hayek, 1952)." (Joaquin Fuster, Memory in the Cerebral Cortex: An Empirical Approach to Neural Networks in the Human and Nonhuman Primate. Cambridge: MIT Press, 1995, p. 87)

These quotes simultaneously suggest that Hayek's 1952 work has not been completely ignored by neuroscientists and cognitive scientists until today, and also that it had many of the same themes as the (much better known) work of Donald Hebb, most famously published in 1949.

So what's the truth of the matter? Well, it seems to be that Hayek's important 1952 work is not nearly as widely known or as widely cited as it should be -- for instance, Pinker's own widely-read popular books on cognitive science, like his 1997 How the Mind Works, may fail to cite Hayek. (I'm traveling and don't have a copy at hand, so I can't be sure, but I don't remember any discussion of Hayek there. [Mark is absolutely right: there is no reference in the bibliography to anything by Hayek and no appearance of his name in the index. --Geoff Pullum, 02/04/04, 2:10pm EST.]) And it's clear that Hayek came independently to some very important basic ideas about emergent organization in collections of neurons, shortly after psychologist Donald Hebb and others came up with similar notions.

Summing it up: Hayek deserves plenty more cogsci kudos than Pinker gave him in 1997, but not as much as Postrel quotes Pinker as giving him in 2004.

This is another example of attributional abduction -- should we really hold Steven Pinker responsible for the irresponsibly misleading content of the quote as deployed in the paragraph cited above? or was it Virginia Postrel's fault? or perhaps an emergent property of their conversation? an artefact of some subsequent editing process?

By the way, Donald Hebb also deserves to be better remembered, even though "Hebb's rule" and "Hebbian synapses" are commonplace terms. So here is an interesting "personal recollection" by Stevan Harnad.

[Postrel link via Mark Seidenberg]

[Update: Cosma Shalizi emailed to say

I think the back-story to the "Hayek invented connectionism" meme is that he claimed to have come up with the basic idea in the 1920s, after reading Ernest Mach, but didn't publish (at least in English?) until _The Sensory Order_; this at least is the story he tells in the preface to that book. My memory of this is a bit hazy, because my copy was destroyed by the post office; I'll have to track this down in the library.

Also, Mark Seidenberg emailed the comment that "Hayek gets no citations in MITECS (there is one for Jurgen Habermas, however) or in Talking Nets, the very interesting collection of oral histories of the founders of modern neural network research." ]

Posted by Mark Liberman at February 4, 2004 07:04 AM