Complicating the Belief-Desire Model of Intentional Action

A member of the studio audience directed me to a podcast on National Public Radio called "Choosing Wrong".

The motivation behind this recommendation is grounded on the fact that much of what I write assumes a belief-desire model for intentional actions. On that model, a person performs that act that would fulfill the most and strongest of one's desires in a universe in which their beliefs are true and complete.

Some of the points made in that broadcast are meant to show that this may not be true.

Actually, the belief-desire model is not entirely accurate. I can mention a couple of areas where the model fails.

The belief-desire model is an empirical theory, which means that it is supposed to explain and predict a range of observations. In this case, it is meant to explain the actions of intentional agents. This means that we can demonstrate that the theory falls short if we can provide relevant observations that the theory cannot handle. This means that we either need to throw out the theory or modify it.

Here is a couple sets of observations that the belief-desire model cannot handle.

Habit

Take a standard keyboard that a person has been using for years and change two letter. For example, switch the keys for the letters 'a' and 's'. Now, tell the experimental subject to type something. You will discover that there will be repeated cases where the subject typed the letter 'a' where he meant to type 's' and visa versa. The agent may, for example, type 'thia' instead of 'this', or 'thst' in place of 'that'.

Typing is an intentional action. Indeed, typing a specific word such as 'this' or 'that' is an intentional action.

We cannot explain this behavior of typing the wrong letter in terms of change in desire. We are assuming that the agent wanted to type the words 'this' and 'that' - but the agent failed to realize these states of affairs.

We cannot explain the failure in terms of a false belief. The agent knows that the letters have been switched. We may even have the agent switch the letters himself. In many cases, the agent will realize the instant he types the wrong letter that he has typed the wrong letter - because he knows where the letters are in fact.

In order to explain this behavior, we have to add a new element to our belief-desire model. This new element is 'habit'. Habits exist. Habits also influence intentional actions. We do not actually have a belief-desire model for intentional action. We actually have a belief-desire-habit model.

Remembering/Forgetting

An agent resolves to pick up some toilet paper on the way home because she is almost out. On the way home, she drives past the store, pulls up at her house, walks in, and suddenly remembers the toilet paper.

We cannot explain the failure to get the toilet paper in terms of a change in desire. Nor can we honestly say that the agent's beliefs suddenly changed such that, for a few moments, she believed that she had toilet paper at home. If this was a genuine belief it would require some evidence that would convince the agent that she was in error - but no evidence is required.

Instead, we explain the failure to pick up the paper in terms of "forgetting'. Intentional action not only requires certain beliefs - it requires that the beliefs that actually control an action be, effectively, in working memory. Other beliefs may exist, but if they are not in working memory then they are forgotten, for the time beling.

Now, we have a belief-desire-habit-remembering/forgetting theory of intentional action.

Avoiding Complications

In most of my writings, I leave out habits and remembering/forgetting because, even though they influence intentional action, they do not create reasons for intentional action. Desires are the only mental states that actually give a person a reason to choose one option over another.

In fact, we can categorize a habit as a good habit or a bad habit according to whether it serves the agent's desires. Nothing becomes good or bad in virtue of its ability to serve a habit.

Similarly, a good memory is a memory that places into working memory those beliefs that are useful and keeps out those that are irrelevant. In other words, a good memory, like a good habit, is one that puts relevant beliefs in working memory so that they can guide action to fulfill current desires.

Neither habits nor remembering/forgetting generates end-reasons for intentional action. Since they do not provide end-reasons for an agent to do things, they are not relevant to the question of value. Consequently, I set them aside and use the belief-desire model for the sake of simplicity.

Interim Conclusion

There are likely other complications when it comes to the behavior of intentional agents. For example, we will need to handle the set of observations whereby people make good choices when they consider something in the future (e.g., buying a membership in a gym so that they can exercise) and yet fail to act properly in the short term (e.g., never actually goes to the gym),

In the light of this, it may be interesting to look at the claims made in the podcast "Choosing Wrong".

I will look at the claims made in that podcast in my next post.