How not to build a hybrid: Simulation vs. fact-finding

Abstract

In accounting for the way we explain and predict behavior, two major positions are the theory-theory and the simulation theory. Recently, several authors have advocated a hybrid position, where elements of both theory and simulation are part of the account. One popular strategy for incorporating simulation is to note that we sometimes assign mental states to others by performing cognitive operations in ourselves that mirror what has occurred in the target. In this article, I argue that this way of thinking about simulation is misguided. The confusion stems from a failure to appreciate how the application of any internal theory, including folk psychology, requires the employment of other cognitive sub-systems and mechanisms. Just as we need to use our visual system to see what another person is seeing, so too we sometimes need to use our own internal sub-systems to ascertain facts about another's mental states. In many such cases, our cognitive mechanisms are used more as “fact-finders” rather than as simulators. After spelling all this out, I offer two ways to demarcate cognitive processes that are real simulation from those that are simply used in the application of a theory.

Keywords:

• Fact-Finder
• Inference System
• Mindreading
• Simulation
• Theory Application
• Theory-Theory

Acknowledgements

Versions of this article have been presented to a various audiences who have offered many helpful comments. I am especially grateful to Jacob Beck, David Beisecker, Neil Delaney Jr., Ian Dove, Greg Frost-Arnold, Karen Frost-Arnold, Maurice Finocchiaro, Christopher Gauker, Todd Jones, Karsten Stuber, Ron Wilburn, James Woodbridge and two anonymous reviewers for their feedback.

Notes

[1] With regard to whatever mechanism generates beliefs about what a target perceives on the basis of the target's direction of gaze and what is taking place in the environment, Nichols and Stich say “we don’t think there is any way to explain these belief-forming processes that relies on mechanisms akin to the simulation prototypes” (2003, p. 135–136).

[2] Presumably, an inference mechanism would not perform computations over types of propositional attitudes at all. In this sense, an inference mechanism is very different from a practical reasoning system.

[3] While this may seem far-fetched, something very close to it occasionally occurs with certain lesions. See, for example, Reverberi, Shallice, D’Agostini, Skrap, and Bonatti (2009).

[4] Heal might claim that in the case of the textbook, I am indirectly simulating the reasoning of the book's author, treating the book as simply an expression of the author's flow of thoughts. But it is easy to tweak the example so that the author is no longer an expert. Suppose I know the text was compiled by a team of undergraduates who do not really understand the theory, but are very, very good at regurgitating ideas found on the internet in a manner that masks their ignorance (a skill many undergraduates unfortunately possess). I would still predict the proximity of the theoretical principles in essentially the same manner as before.

[5] A similar point is made by Currie and Ravenscroft, who insist that a critical dimension of simulation is “variation in causal role” (1997, p. 164).

[6] As an anonymous reviewer has noted, this point primarily applies to prediction. If we are trying to explain how the target's mental states are formed, then, of course, we need to care a lot about the nature of those mechanisms responsible.

[7] Consider the now-famous fictional amnesiac, Otto, whose “memory” involves writing everything down in a notebook (Clark & Chalmers, 1998). Even though a person could know that the way she recalls information is quite different from Otto, she could nevertheless use her own memory to recall something that she knows they both learned and thereby successfully predict Otto's verbal expression of his “recollection.”