View Single Post
Unread March 18th, 2006, 11:14 AM
ToddStark ToddStark is offline
Join Date: Jul 2004
Posts: 174
Arrow Reasonable limits to the modularity thesis

If I get the drift here, it sounds like Tom is suggesting that human moral reasoning is essentially the application of something like instincts (via evolved computational modules in this particular case). So for example, moral responsibility is nothing more than the pull of guilt wired into us for doing the wrong thing. This implies to me that there is no such thing as human wisdom, or no basis for its development, or that it is something inflexible and hardwired. None of these is in accord with my experience.

So I have to disagree with this. For example I feel that the pull of guilt is very important or neccessary to moral reasoning, but not sufficient. I could feel guilty for imagining doing the wrong thing, and therefore be guided into making a more moral decision, or I could feel guilty for missing the opportunity to exploit someone. I don't think individual "emotional modules" can themselves contain enough contextual information to be able to make the sorts of decisions that we consider good judgment.

The older instinct theory of behavior and the newer evolved module theory share the same conundrum which it seems to me we would need to address here. If our decisions are "tied to" various hardwired responses to information or stimuli in our environment, then how do we manage to choose which hardwired response gets activated? Is there some property of a task that triggers one module vs. another? The evidence doesn't seem to point that way, we often perform the same task differently in different situations.

For behaviorists, the problem was that the environment simply doesn't provide enough structured information to make this decision for us according to our goals, so there must be some fairly sophisticated information processing going on within the individual mind, which I think is a big part of the rationale for the general shift of most researchers of mind, brain, and behavior to cognitive neuroscience over behaviorism.

For modularists, the problem is similar, but more sophisticated.

Human decision making is obviously littered with all sorts of blind spots and judgmental biases that have been well studied by experimental psychologists and social psychologists. This does, to me, imply that there are a lot of "click-whir" responses in a human brain that support the model of underlying modules being activated under particular conditions.

The reason the modularity problem is more of a dilemma to me than the behaviorist problem is that the tools that underlie human reasoning do not seem to be only domain-specific. For example, under some conditions we do consistently make a concerted and somewhat successful effort to test our conclusions for logical consistency with evidence, and apply this skill in a way that is not specific to a particular computational domain.

So while I would agree that even this sort of skill is probably exploiting underlying specialized abilities, it is being applied across domains. And most importantly, it is being applied in a way that is not informationally encapsulated. This should be emphasized because that is the cornerstone principle of the mind modularity model, that the information in each module is opaque to the other modules. If you break that assumption, then many of the most common implications we want to be able to draw from the model are no longer supported by it.

For example, people can no longer be said to simply be reacting from the activity of a particular module in a particular situation, that would be an untestable story akin to Freud's unconscious motivations in psychoanalysis. Even if the module was very real, it would not be appriopriate to pinpoint it as a causal factor in a particular behavior without a lot of additional information, and it would probably not be the sole cause.

To me, the implication seems to be that the human brain possesses strategies for accessing information across domains, and performing tasks differently under different conditions, and this implies to me that it is not modules all the way up.

It seems to me to still make sense to look for modularity when we see click-whir responses, just as it made sense to look for non-obvious reasons for behavior in psychoanalysis, but it makes no sense to me to assume that cross-domain skills like reasoning and moral judgment are themselves modular.

kind regards,

Reply With Quote