Non-Statistical Models
- If you want to say something about your data, I don't think there is
anything better than statistics. That's not AI, but it's got a strong
mathematical basis, and you can actually prove things about it. Psychology
as a science is just a bunch of theories, and they can not be proven. All
you can do is provide more support for your theory. The theory of gravity
has pretty sound support, but it's just a theory.
- You can also use AI mechanisms.
- For instance, you can use a large bit of data and classify it
manually.
- For instance, (and it's a simplification), person A has autism, B doesn't,
C does and so forth. This can all be derived from, say, a test they
take.
- Now you can build a classifier, say a support vector machine, that
classifies this data perfectly.
- You could then apply it to new data and say person X has autism.
(Again it's not proven correct. For that matter, the initial human
classification may not be perfect. For that matter, the idea of autism
may not be correct.)
- That can be useful, particularly if you don't have experts in classifying
people with autism.
