Luan Hassett

Probability is the backbone of evidence based medicine, but it doesn’t describe reality with the depth that explanatory theories provide.

Let’s say that I am studying physics and I calculate the probability that earth will be struck by an asteroid in the next five years. I don’t know what a typical probability calculation would come to, but imagine it was 1 in 50,000. Now imagine that someone in the physics department comes to me and says they have finally replaced the lens on the best telescope in the building; I look through it and see an asteroid hurtling through the galaxy towards us. The probability then obviously changes. To take another example, anyone who has been involved in match fixing knows that the odds you get at the bookies are not an objective statement about reality.

So probability has assumptions. But could you not say that all theories contain assumptions? Yes, but the point is that while they may use probability instrumentally, good theories do not speak fundamentally in terms of what will probably happen, but instead speak of what can and cannot happen, the conditions under which things happen, and why they do happen. In other words good theories provide explanations. Physics can explain why you cannot travel faster than the speed of light. Microbiologists can tell you the conditions under which antimicrobial resistance happens at a level that causes problems for doctors and patients.

As it stands, psychiatry contains no explanations for why any major mental illness happens. We do not know the underlying cause of Depression, OCD, Bipolar Disorder, Schizophrenia, or anything else. This is despite (more accurately: because of) the copious use of probability in research. We know something about the genes and environmental exposures that are relatively tightly associated with developing those conditions. But there is no explanatory pathway for how any given gene or experience actually causes any given thought, be it an obsession, negative cognition or delusional belief.

Now let’s talk about p values. When you do a study comparing an antidepressant to placebo, p values describe the probability that the effect that follows administration of the drug would have happened if the antidepressant had no intrinsic therapeutic capacity. A low p value refutes the null hypothesis stating that the drug has no effect above placebo. Some people would say that p values “falsify” the null hypothesis. But when you falsify a hypothesis using probability, you don’t get explanatory resolution.

A ‘statistically significant’ result is saying that there is a systematic effect. It’s saying that antidepressants are doing something. That’s basically it. It doesn’t tell you what the systematic effect is due to. That’s left up to us psychiatrists to conjecture about. When it is left to us to conjecture, without a way of experimenting to adjudicate between the conjectures that each of us brings to the debate, then the knowledge category we’re working in is primarily philosophy rather than science. We can use probability statistics to assist in our criticisms of each other’s conjectures, and there is benefit in doing this, we just need to be clear what epistemic domain we are working in.

It’s important to understand that we currently lack the experimental and statistical mechanisms to grow our knowledge quickly. What I have tried to show is that we don’t have scientific explanations for mental illness because our research methods aren’t trying to find them. What the methods are trying to do, without us being explicit about this, is assist us in philosophy. We treat patients using philosophy. And we often succeed. But the rate of knowledge growth in ours and many other biological fields pales in comparison with physics because physics historically made greater use of experimental falsification, where the aim was to find an occurrence that was in direct, logical, non-probabilistic contradiction of the theory being tested.

There are reasons why physics found it easier than biology to do this, one of which being that there are limits to the experiments we can ethically perform on humans. By some accounts physics is also moving away from the use of falsification, which sounds to me like a bad thing. Furthermore, humans are more complex than planets and atoms. This makes falsification of theories pertaining to humans very difficult to do, and especially when the theory is meant to apply to humans in general. The clear solution to this is to see that each patient requires their own modified theory of mental illness, and the requisite approach to psychiatry is one of being epistemically energised. This might sound like advocating eccentricity, but actually most psychiatrists already understand what I’m arguing. It would not be rare for an experienced psychiatrist to look at a powerpoint slide and express skepticism about the usefulness of whatever the research is purporting to prove.