Explanation and Cognition

“Those books don’t go out,” our psychiatric librarian said when I told him I had been asked to review another book on cognitive science, this time focused on the burning issue of explanation: its development, its kinds, its domains, its whys, and its causes. So as I read this book, I sought an explanation for his observation that our psychiatrists and researchers, who seem to read everything, don’t read much cognitive science. First the humble theory: it’s too hard. There are formulas in this book, and few physicians love math or symbolic logic, but only two of a baker’s dozen chapters have them, and they are not hard math, of the sort of

which addresses the question, Does i cause e, when a may cause e and j may interact with i to cause e? Perhaps we physicians do not ask ourselves questions in this way, unlike Patricia Cheng (“Causality in the Mind: Estimating Contextual and Conjunctive Power”). Our medical thinking has often been unicausal. We are spoiled by our experience that one thing is wrong with the patient at the moment and one intervention restores to good order an immensely complex but powerfully self-correcting homeostatic biosystem. I guess we doubt we need more logical orders.

Do we need philosopher Paul Thagard’s chapter to tell us that disease explanation is “causal network instantiation” and not deductive, statistical, or unicausal? Perhaps, when we recall that psychiatrists have argued that schizophrenia, autism, depression, homosexuality, or intelligence are caused by nature alone or experience alone and that psychotherapeutic effects could be ignored in favor of pharmacotherapeutic effects. Or maybe we would be edified by Robert McCarley’s delightful musings about the naturalness of religion and the unnaturalness of science, which will never threaten the persistence of religion and is always itself in danger of disappearing (as it once did) because religion and human thinking generally prefer causal agents (in the case of religion, what McCarley terms “culturally postulated superhuman agents”) and science restricts agent causality. We did once prefer the agency of schizophrenogenic or autistogenic mothers. Although the conveying of conceptual logic to children is much a part of this book, it does not tread into the misconveying thereof.

Explanation runs through all of human life, the editors contend in their chapter. Explanatory theory is weaker than the “abyss” of theory, as they put it. But Keil’s work has shown that even preschoolers know how pieces of explanatory knowledge are clustered in the minds of others. Welcome to what’s called “theory theory.” Philosopher Robert Cummins critiques five paradigms in contemporary psychology, including connectionism and neuroscience. Clark Glymour discusses Bayes nets as psychological models and proposes that they are descriptive of the acquisition of causal knowledge. He clearly explains the causal Markov condition for linearly related causal features, isomorphic to Bayes nets, by examples involving a toy train, a clap-on, clap-off light, and television. These statistical combinations in causal chains have been used widely in computer science and in scientific and engineering data analysis. So why not as a psychological model, he argues, since humans are the source for probabilistic expert systems?

These very non-Aristotelian and fuzzy systems are akin to brain and neural network models. The most vivid chapter is by computer scientist Herbert Simon, who argues that “the weakest link in the chain today is the general absence of bridging theories between the EIP [elementary information processes] and neurological levels” (p. 33).

There is much lore in this collection for the persistent reader, and persistence is required at times when philosophical cogitation approaches what oft was thought, yet ne’er so long expressed. Maybe it’s patience for long-drawn-out theory that psychiatrists lack in our age of atheoretical DSMs.

Experimental work is included at the end of the book. Christine Johnson and Frank Keil describe their experiments on conceptual combination involving imaginary noun links such as “hospital bicycle” and “kindergarten bird.” They found that, rather than superficial concept structures like feature typicality, it is the explanatory relations reflecting causal or intentional design that are the contingencies organizing a concept. Gregory Murphy explores explanatory concepts for his “kez” and “dax” nonsense categories, about which subjects are asked to form concepts from exemplars given to them on cards. It turns out that people’s ability to explain the features of a category has dramatic effects on their learning, which is faster when features fit together better than when they are arbitrarily related, but categories with clashing features are no harder to learn than those which are simply unrelated.

Alison Gopnik, in the most gonzo—and a touch overinclusive—chapter titled “Explanation as Orgasm and the Drive for Causal Knowledge: The Function, Evolution, and Phenomenology of the Theory Formation System,” argues that explanation, or uncovering causes, is to theory formation as orgasm is to reproduction; or perhaps as uncovering the spatial character of moving objects is central to the visual system. She argues that explanation is a drive (is this a return of the repressed old libido theory of Freud?); that is, our genes provide orgasm, and explanation, because reproduction, and theory (a veridical causal map of the world) are good for us in an evolutionary sense. Just as there is a phenomenology of emotions, Gopnik postulates a distinctive phenomenology of explanation, which she reduces to the “hmm” and “aha” experiences corresponding to “why,” the search for explanation, and “because,” the recognition that explanation has been reached. These are close, but not identical, to curiosity and interest, and Gopnik postulates distinctive facial expressions for each. She places the drive for causal knowledge in a child developmental context. For example, the terrible twos are a child’s experiments with differences in desires, even though they may elicit a mother’s rage; this is a drive going beyond mere cognition. Gopnik extrapolates that “in Swann’s Way, Swann compulsively tests Odette in search of her secret life, in spite of the emotional and practical pain this will cause him, a rather advanced case of ‘the terrible twos’ ” (p. 312).

Surely this tome explains something to everyone, and it will fly off our librarian’s shelf.