# Is this a solution to the disjunction problem of causal representation?

As I understand it, the disjunction problem is how could a causal theory of inner representation account for mistaken identification of external objects or object types. For example, if I see a fox and mistakenly take it to be a dog (my internal representation of dog is activated) how is this possible? My internal representation of dog is created and henceforth activated by dogs, and a fox is not a dog. The conclusion seems to be that the inner representation is not of a dog but rather represents the disjunction fox-or-dog.

The suggested solution has two parts. First, an inner representation of a, say, dog, is not a unitary atom but is composed of a bunch of property representations, each a representation in its own right. In poor lighting condition, at a distance, when occluded (etc.) when I see a fox, it activates the property representations that are a subset of the property representations that collectively mean dog. So at this point what is activated is a subset of the property representations that mean either fox or dog (is a disjunction). But it is not a distinction between fox and dog. Fox and dog are represented by further – and between each other different - property representations that are not activated on this occasion (due to poor light, distance, occlusion etc).

Secondly, The activation of the dog representation may not be a causal consequence only of looking at the fox. What if I recently saw a film about dogs? What if have a pathological fear of dogs and tend to jump to conclusions about things that look like dogs but aren't? What I've just read a book about dogs? In other words, the causal antecedents of the activation of the dog representation might not be only current sensory input.

The errors of the disjunction problem being (1) inner representations of external object types are atomic when in fact they are actually a collection of property representations only a subset of which might be activated, and (2) the activation of an inner representation of an external object type may not be purely a result of a causal chain from the current environment to the brain.

• Could you explain how your solution is different from the standard ones, if it is. Commented Oct 16, 2018 at 20:27
• @Conifold I'd like to address the 4 SEP test cases. Starting with case III, SEP says: On hearing “What kind of animal is named ‘Fido’?” a person might token the syntactic item “X” but we do not want the question to be among the content-determining causes of “X”. Explanation: We need to distinguish causing the content of "X" from causing a tokening of the content of "X". Since in computers "X" will be a structure, this is the different between building a structure and accessing the structure once built. The question causes access of the structure & this is non-content-determining. Commented Oct 26, 2018 at 0:17
• @Conifold I'd like to address all 4 cases. I'll put all this in an answer. It's taking a little while to get happy with what I want to say. But I think I can say something useful about each case. Commented Oct 26, 2018 at 0:21
• I was looking for something like "this is like [asymmetric dependency/ best case/etc.] but such and such is different and it helps so and so". I am getting lost in your examples, unfortunately. Commented Oct 26, 2018 at 0:24
• You have to allow space not only for development in a concept, but also creativity. Words and categories shift over time. Commented Apr 17, 2021 at 3:51

## 2 Answers

To determine the function or purpose of a part of a machine, you have to look at the totality of how that part interacts with the rest of the machine. This means you need to look at what the part does in every circumstance in which the machine operates.

• How does the part influence the operation of the machine in each circumstance? i.e., how does the state of the part influence the state of the rest of the machine?
• How does the machine influence the operation of the part in each circumstance? i.e. how does the state of the rest of the machine influence the state of the part?

Now, the part of your brain that detects dogs is a part of a machine. It's not enough to look at one circumstance - this mistaken identification of a fox - and conclude the function of the part based on that. You can clearly see that the dog-detecting part is a dog-detecting part based on how it interacts with the rest of your brain, and how it works in other circumstances.

• If you look more closely at the fox, the dog-detecting part will turn off, as you realize it is not a dog. This alone is sufficient to distinguish the dog-detecting part from a fox-dog-detecting part.
• Your brain will draw conclusions based on the dog-detecting part, e.g. that the animal can bark or is likely friendly to humans. This also distinguishes it from a fox-dog-detecting part, because those conclusions are not warranted for foxes.

Regarding your:

First, an inner representation of a, say, dog, is not a unitary atom but is composed of a bunch of property representations, each a representation in its own right.

This was already common knowledge realized by various types pf physicalists when they were trying to causally explain mental states from neurobiological physical states, especially if you're a reductivist and accept the existence of mental states such as perceiving a dog mistakenly instead of a fox. Thus they found they cannot have a good causal theory to determine the necessarily disjunctive information stored for such a mental state.

Regarding your:

Secondly, The activation of the dog representation may not be a causal consequence only of looking at the fox. What if I recently saw a film about dogs? What if have a pathological fear of dogs and tend to jump to conclusions about things that look like dogs but aren't?

I agree with you on this intuition, however, the principle here for physicalists' causal theory is Causal closure, so this certainly doesn't sound a good solution for most of them. Also what you described can be interpreted as same disjunctive indeterminate states before so you're just most likely restating the same problem here. Please note for a eliminative materialist this disjunction problem is not a problem at all since under this radical view, most mental states actually don't exist at all, perceiving a dog mistakenly is just a natural epiphenomenon of the brain's certain complicated, parallel and distributed neurobiological processes.

Since evolution cannot solve the disjunction problem, the right conclusion for the materialist is to accept eliminativism by denying that neural states have as their informational content specific, particular, determinate statements which attribute non-disjunctive properties and relations to non-disjunctive subjects.