Venturing into computational psychiatry

03/04/2025

Next year, provided exams go well, I will be starting my intercalated BSc in Neuroscience. As part of this, I will be undertaking a lab project at the Applied Computational Psychiatry Lab, with whom I worked last summer as well.

Given that I only have nine months to work on this, it seems likely that I will be hyper-focused on a highly scoped project, so I worry that I will be subsumed by minutiae and lose my grip on the bigger picture. To this end, I have been reading Seriès's Computational Psychiatry: A Primer. I have been enjoying the historical view of how this niche has emerged. Its development was less stochastic than I imagined it to be; a handful of researchers, often motivated by powerful personal experience with psychopathology, sought to apply the well-established methods of computational neuroscience to the far murkier waters of phenomenal experience. Emotively, this certainly appeals to my medic-brain: mapping out higher-level emotional circuitry feels far more tangible and urgent.

Equally, I have some scruples. I think I could very easily fall into some conceptual traps while working in this area. The main one is reifying computational models of the mind and treating them as the mind qua itself. Many have dismissed computational approaches altogether on these grounds, but I believe it is possible to apply them to the psyche whilst being aware that you are not capturing anything of the underlying phenomenal state – just systematising some of the inputs and outputs of the state.

With this in mind, the central question I will be taking into this project is whether it is actually possible to make computational models that do not compromise the primacy of patient phenomenology in psychopathology.

Some more basic questions I have: