The Atomic Components of Thought

by John R. Anderson

This book achieves a goal that was set 25 years ago when the HAM theory of human memory was published. This theory reflected one of a number of then-current efforts to create a theory of human cognition that met the twin goals of precision and complexity. Up until then the standard for precision had been the mathematical theories of the 1950s and 1960s. These theories took the form of precise models of specific experiments along with some informal, verbally-stated understanding of how they could be extended to new experiments. They seemed to fall far short of capturing the breadth and power of human cognition that was being demonstrated by the new experimental work in human cognition. The next 10 years saw two major efforts to address the problems of scope. In 1976, the ACT theory was first described and included a production rule system of procedural memory to complement HAM's declarative memory. This provided a computationally adequate system which was indeed capable of accounting for all sorts of cognition. In 1993, a new version of ACT--ACT-R--was published. This was an effort to summarize the theoretical progress made on skill acquisition in the intervening 10 years and to tune the subsymbolic level of ACT-R with the insights of the rational analysis of cognition. Although the appearance of generally-available, full-function code set off a series of events which was hardly planned, it resulted in this book. The catalyst for this was the emergence of a user community. Lebiere insisted that assembling a critical mass of users was essential to the ultimate success of the theory and that a physical gathering was the only way to achieve that goal. This resulted in the First Annual ACT-R Summer School and Workshop, held in 1994. In writing the book, the authors became seized by an aspiration that went beyond just describing the theory correctly. They decided to try to display what the theory could do by collecting together and describing some of its in-house applications. This book reflects decades of work in ACT-R accumulated by many researchers. The chapters are authored by the people that did that particular work. No doubt the reader will be impressed by the scope of the research and the quality of the individual work. Less apparent, but no less important, was the effort that everyone put into achieving the overall consistency and technical integrity of the book. This is the first work in cognitive science to precisely model such a wide range of phenomena with a single theory.

724 pages, Kindle Edition

First published January 1, 1998

About the author

There is more than one author with this name in the Goodreads' database

Reviews

[Greg · Rating 3 out of 5]

Anderson has a very clear idea of what he believes cognition to be, and even in a text as dry as this one, subtle hints of language betray his (and his colleagues') near-dogmatic certainty (for instance, comparing ACT-R to the theory of relativity). There are thus two reasons for which you will not enjoy this book: 1) you feel Anderson is bashing you over the head; 2) you do not accept his basic assumptions--or the basic assumptions of cognitive science--on which ACT-R, cognitive science, and this book rest.

That said, I do believe this book is worth reading for those interested in computational/mathematical models of cognition, even if one is not an ACT-R modeler. First, ACT-R is a widely used cognitive architecture and so those working in the field, or with a scholarly interest in cognitive modeling, should have at least a working knowledge of the theory, particularly as it has been very successful in modeling human performance across a wide range of phenomena. Second, it is an excellent example of a production system model of cognition (of which there are several), and so serves as a useful introduction to production systems in general. Third, it shows the level of sophistication and specificity to which one should aspire when investigating cognition. This last point cannot be overemphasized, yet it seems to go unnoticed by many who approach not only ACT-R, but all scientific inquiry into the mind. Many people believe the mind is the province only of philosophy or religion. This book--and over fifty years of successful research--refutes that belief. For that reason, the reader of "The Atomic Components of Thought" should not necessarily agree with its claims, but at least appreciate their rigor.

[Michael Venzke · Rating 1 out of 5]

I have major fundamental issues with this book and ACT-R.

my main issue is that I really hate it.

maybe I gave up on it too early, but the whole premise seems to be fudging an exact functional replication of how the brain operates, based on anecdotal tests and stupid assumptions.

it is a journey down a wrong path.

it is an ad-hoc temporary solution, accepting that we don't have enough information or understanding to do an actual representation of human thought, but i just had major problems with it, and maybe an ACT-R simulation can tell me why (after the programmers decide what the possible reasons should be & then try it on 100 people).

im not against cognitive science as a whole, and there are other books and methods I highly enjoy, but act-r just really pissed me off. it didn't help that the lisp version was so clunky.