Morality and the Neurocomputational Model of Decision Making

A member of the studio audience sent me a paper in which Paul Thagard wrote that desires are not propositional attitudes.

Thagard, P. (forthcoming). Desires are not propositional attitudes. Dialogue: Canadian Philosophical Review.

That article raises objections to the view that desires are propositional attitudes. While making that argument, the author does draw a comparison between the belief-desire model of intentional action and the model that seems to be coming out of modern brain studies.

(1) desires + beliefs -> actions

(3) emotional brain areas + cognitive brain areas + interactive mechanisms -> actions.

This suggests that some sort of reduction is to be expected from belief-desire theory to what the author calls the neurocomputational account of decision making. He reports that the neurocomputational account has much greater explanatory potential.

Well, when it turns that potential into actual explanations, then we will see what can be done with it.

At the same time, he also wrote:

The everyday belief/desire explanation of action often suffices for ordinary social purposes, but it fails to give anything like the kind of deep mechanistic explanations that have contributed so much to physics, chemistry, and biology.

This raises what I consider to be an interesting issue in moral theory - the issue of simplicity.

Let us assume that there are moral facts - that can be reduced ultimately to relationships between states of affairs and desires (or states of affairs and emotional-brain-area facts, as the case may be). And that these facts are knowable. They are legitimate objects of scientific study just as the relationships between atoms in a molecule are legitimate objects of scientific study.

That does not guarantee that these facts will be easy to know. We can . . . in fact, we should . . . expect that the neurocomputational account of decision making might be particularly hard to understand - something where only a few people heavily focused on this one area of study can actually be said to understand it. And that these neurocomputational facts are an integral part of the relationships that exist between states of affairs and desires (or neurocomputational emotional-brain-area facts).

Does it make sense that the discovery of moral facts requires that the agent acquire the equivalent of a PhD in neurophysics? Would it do to have moral calculus be so complicated that figuring out the right thing to do requires set theory and multi-level differential equations?

Morality needs to be something that fits into the experience of the common individual. It is not een sufficient to argue that the average person can understand it - this still leaes a substantial population incapable of making moral decisions. To include them, morality must be something that a substantial portion of the adult population can understand.

One argument that people can give is that it is entirely unreasonable (and a substantial piece of luck to boot) to expect moral facts to fit that level of simplicity - not if moral facts are real.

This leads to the possibility that even if a neurocompuationoal set of facts about the remationships between emotional brain areas and cognitive brain areas are more complex than a belief-desire model, that it will still be the belief-desire model that is to be used in making and defending moral claims.

This is analogous to the difference between Einsteinian physics and Newtonian physics. Newtonian physics is the physics of our day-to-day world. When people talk to each other about day-to-day events, that language is almost entirely conducted in Neutonian physics, even though the propositions of Newtonian physics have been proved to yield false conclusions. Those false conclusions are still good enough.

Similarly, a neurocomputational account of decision making might well prove that many of the belief-desire accounts of decision making are false. It may actually be the case that it trumps the belief-desire model in computational power. Yet, in spite of this, the belief-desire model (and desire utilitarianism itself) would still be the model to use in day-to-day moral discussions.

The answers, though wrong, are close enough to the truth for all practical purposes.