Minds, models and mechanisms: a new perspective on intentional psychology

Abstract

In this article, I argue that intentional psychology (i.e. the interpretation of human behaviour in terms of intentional states and propositional attitudes) plays an essential role in the sciences of the mind. However, this role is not one of identifying scientifically respectable states of the world. Rather, I argue that intentional psychology acts as a type of phenomenological model, as opposed to a mechanistic one. I demonstrate that, like other phenomenological models in science, intentional psychology is a methodological tool with its own benefits and insights that complements our mechanistic understanding of systems. As a result, intentional psychology's distinctive scientific benefit is its ability to model systems in unique, non-mechanistic, ways. This allows us to generate predictions that we cannot otherwise generate using the mechanistic models of neuroscience and cognitive psychology necessary for various scientific tasks.

Keywords:

• intentional psychology
• mechanistic model
• phenomenological model
• prediction
• explanation

Acknowledgements

A number of individuals were pivotal in the shaping of this study, and I would like to offer them my thanks. This includes: Chris Eliasmith, Micheal McEwan, Kurt Holukoff and Alexander Winthers. I am also very grateful for the feedback received on this article from the audience of the 2011 Epistemology of Modelling and Simulations Conference.

Notes

Note

1. Some might object to this suggestion of a family resemblance between intentional and statistical models. After all, while statistical models are based on the well-defined axioms of probability theory, the axioms of rationality (on which intentional attributions seem to be based) are not so apparent. It is questionable whether there even are any clearly definable axioms of rationality. When we predict the behaviour of others, we do so without any explicit understanding of what such axioms might be. In this respect, intentional models may be too different in kind from statistical models to warrant inclusion into the class of phenomenological models. The problem with this sort of objection is twofold. First, it is a mistake to think that all domains to which statistical models are applied are therefore axiomatised. That there are statistical axioms does not mean that any particular statistical model (e.g. of stock markets) is also axiomatised in the relevant respect. We do not, for instance, have the axioms of stock market behaviour just because we have a statistical model of the stock markets. You could, after all, have a statistical model of rationality. Second, and more importantly, this sort of objection is a red-herring. Even if we assume that there are no explicit axioms of rationality (which is still up for debate), it is hardly a necessary characteristic of phenomenological models that they be based on axioms. That was never the claim. Intentional models are, however, relevantly like statistical models in key respects: First, they are predictively valuable in scientific practice. Second, they make predictions without telling us structural or mechanistic details of the system. Third, they are often used to identify patterns and regularities in behaviour produced by mechanistic systems. Finally, they are used in conjunction with mechanistic models to provide more complete understandings of systems. These similarities provide us with compelling reasons to consider intentional characterisations as a species of phenomenological model, just as statistical models are.