2

When studying AI, computationalism was always referred to as a theory, a theory of mind, the theory that the mind is an executing computation. But is it really a theory? How could it be disproved or disconfirmed? If a certain computation (program, algorithm) turns out to be unintelligent, that just means the right computation has yet to be discovered. The idea that the mind is a computation isn't challenged. What experiment could disconfirm the computational mind?

The theory of phlogiston says combustible material contains a substance, phlogiston, that leaves during burning, hence the lesser weight of the ash, etc. This idea could be disconfirmed – and was: magnesium ash weighed more than the unburned metal. What empirical test could disconfirm computationalism? If there's no such experiment, can computationalism properly be called a scientific theory? If not, can the research project based on it – AI – be a scientific project?

|improve this question
  • It's not a theory strictly speaking but a hypothesis... – Mozibur Ullah Dec 16 '17 at 7:07
  • OK. But what I was wondering is what would disconfirm the hypothesis? What would show that it is false? If there's no way to falsify it, is it really an hypothesis or is it just a belief masquerading as a claim that could be proved or disproved (when in fact it can't be)? – Roddus Dec 16 '17 at 7:15
  • 1
    Church's thesis gives us a definition of computability, if the mind is a computer it obeys that definition of computability. If an experiment is done that shows that the mind doesn't obey that definition, it is not a computer and computationalism is wrong. Of course everything is a lot more subtle than that but I don't see how that general argument isn't apparent by the definition of computationalism. "The mind is a computer, which we have a definition of. If you can show the mind doesn't fall under that definition, then it is not a computer." – Not_Here Dec 16 '17 at 8:32
  • 1
    The computational theory of mind is the theory that the mind acts as a computer, i.e. obeying the laws of computation, which are perfectly falsifiable. If we discover an algorithm tomorrow that solves the halting problem, we will have falsified Church's thesis. In a similar way, if we discover that the brain can do something a Turing complete computer cannot, then we have falsified computationalism. Again, I'm not seeing where the disconnect is in your view of it being a theory. – Not_Here Dec 16 '17 at 9:08
  • 2
    As the answer points out, it's not that easy to show that a physical computation violates the Church-Turing thesis @Not_Here. You might be interested in this post on Falsifiability and Gandy's Variant of the Church-Turing thesis. – Artem Kaznatcheev Dec 17 '17 at 9:57
3

I read your question as Is computationalism falsifiable?

In the comments, Not_Here argues that it is, because we would just need to demonstrate that minds can solve non-computable problems, e.g. the halting problem. However, to experimentally demonstrate that a mind solves the halting problem, we need to show that it correctly determines halting for all possible inputs. We immediately run into two problems: First, we have no general way of checking the answers - after all, the halting problem is non-computable. Second, any experiment can only deal with a finite number of inputs, and we would need to test them all. Essentially, the claim that a device (which always answers) solves the Halting problem is itself only falsifiable, but not decidable.

We could find out that a given model of minds is able to solve the Halting problem, but that only tells us that this model and computationalism are inconsistent. It is not helpful in determining which one to let go.

If you wish to avoid using theory for non-falsifiable stuff, I would recommend the word paradigm.

|improve this answer
  • 1
    "Paradigm" seems pretty good. The Church-Turing thesis was called a thesis because it is partly intuition, so maybe the computational thesis of mind? – Roddus Dec 17 '17 at 23:52
  • 1
    My main interest in this question of falsifiability of computationalism comes from the idea that computationalism might be false but nevertheless digital computers will think. In other words, that computers can operate as designed without computing on their inputs. And further, that these non-computational operations can be sufficient for thought. But this idea seems to raise difficult conceptual issues - What is computation? for instance. If there is some practical way to test whether computationalism is true, then maybe some headway could be made. – Roddus Dec 18 '17 at 1:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.