Podcast Summary
Podcast: AI and I
Host: Dan Shipper
Episode: The Future of AI in Medicine: From Rules to Intuition | Awais Aftab, Psychiatrist and Writer
Date: June 4, 2025
Episode Overview
This episode dives deep into the intersection of psychiatry, philosophy of science, and artificial intelligence, exploring how AI might be poised to revolutionize medicine—and more specifically, mental health diagnosis and care. Host Dan Shipper talks with Dr. Awais Aftab, psychiatrist, writer, and philosophy-savvy thinker, about the complexity of mental health conditions, the limitations of current diagnostic systems, the value of explanatory pluralism, and how AI might help us move from rigid rules to fluid intuition in both classification and care.
Key Discussion Points & Insights
1. The Complexity of Psychiatric Diagnosis
- Dan shares his personal OCD journey and the difficulty of obtaining an accurate diagnosis, highlighting how fuzzy and context-dependent these labels can be.
- Dr. Aftab explains clinical misdiagnosis: “OCD, as far as we know, it's not one thing. In fact, even from a symptom standpoint, it has fuzzy boundaries. So we have to respect that heterogeneity.... It's an entity that exists from a pragmatic point of view, but not from an essentialistic point of view.” (00:08–00:43)
- The discussion frames psychiatric categories as pragmatic tools rather than strict natural kinds.
2. Explanatory Pluralism in Psychiatry
- Shipper introduces the concept: In complex domains like psychiatry, single-theory explanations rarely suffice; instead, multiple perspectives yield a fuller picture.
- Aftab elaborates: “Scientific pluralism… things can be explained… through a variety of different theoretical perspectives… there isn't necessarily one single correct true perspective. But things are multifaceted, and that multifaceted nature is not captured by one theoretical… perspective.” (03:29–05:23)
- Draws a contrast between “hidden essence” scientific objects (like chemical elements) and messier, multidimensional psychiatric phenomena.
3. Limits of Scientific Essentialism & the Role of Context
- There's a recurring skepticism toward overly reductionist, rule-based approaches—both in science generally and in mental health specifically.
- Memorable moment: “The history of science is like, if you go back to Newton, is finding something that we think is super general and then realizing that it's actually contextual, we just didn't have a way to access the context in which it wouldn't hold.” (15:09)
- Discussion of realism vs. anti-realism: Do clinical categories represent ‘real things’ or just useful models? Dr. Aftab favors a pragmatist philosophy: “It shifts the focus away from certain metaphysical questions towards our ability to make use of concepts in the world.” (16:37)
4. The Realities and Pitfalls of Labels
- Shipper and Aftab discuss the power and limitations of diagnosis: For some, a label like OCD is liberating; for others, it may be harmful or misleading.
- Aftab: “...For some people, the label is bad. For me, it was quite freeing because I was like, well, you know, for OCD, typical talk therapy tends not to work that well until… treatment is much better managed. And even once I had gotten into the correct treatment… it was a nightmare to get on medication. But once I did, it's been amazing.” (18:53–19:01)
5. Checklists vs. Clinical Intuition—and Parallels in AI
- Checklist approaches (and their limitations) in both clinical practice and early AI development:
“There’s a lot of pressure to be able to reduce these things down into a checklist... but really a good clinician has a little bit of a smell for it.” (24:03) - Parallel to machine learning: Rule-based AI approaches failed at pattern recognition; success came with deep learning’s flexible, example-driven “intuition.”
“That's exactly how machine learning researchers started in AI and exactly what didn't work, what ended up working is deep learning.” (00:44–01:21, recapped at 31:04)
6. AI and the Reimagining of Psychiatric Classification
- Current approaches: The DSM and ICD began with clinician observation, refined through operational definitions. Newer approaches like HiTOP use statistical factor analysis, but the “space of psychopathology is very flexible” and resists one-size-fits-all mapping. (28:36–31:04)
- Aftab: “AI and... deep learning can help us identify new patterns and new ways of talking about the space of psychopathology that we might not have thought of ourselves... but I think in theory, that's possible.” (28:36–31:04)
7. Predictive Power vs. Explanatory Power
- Shipper argues that science undervalues pure prediction:
“In science we prioritize explanatory power over predictive power a lot. Because… in order to get your paper accepted… you have to like, have a theory. And deep learning doesn't have theories, it just says we know it works. And so my... stump speech is basically like, we should just like throw out explanations... and build predictive models… even if we don't have a theory.” (33:06–34:38) - Aftab notes that big prediction efforts (e.g., suicide risk) often don’t generalize—but agrees more data and advanced AI may change this, echoing progress in language models.
8. Practical AI Tools in Clinical Practice
- Medical note-writing: AI tools that transcribe and structure clinical encounters are already providing value, and could potentially “exceed the average clinician” in thoroughness for diagnostic interviews. (38:38–41:12)
- Democratizing support: “While ChatGPT is not being billed as… a therapist, I would bet you if you look at the way people are using it, a lot of them are using it for things that are sort of therapy like and that it's, that's actually probably a good thing... democratizing access to like the most basic level of mental and emotional support...” (41:22–42:35)
9. AI as Support and Therapist—Strengths and Limitations
- Aftab on AI tools: Generally optimistic, especially for mild/moderate cases or as a supplement—though heavier/entrenched issues may still require human support.
- “A lot of times clinicians… have all of the quirks of humanity. They can be arrogant, they can be dismissive, they can be rushed. Right. Versus chatgpt is patient. It's always there. It's non-judgmental.” (42:35–43:45)
10. AI-Driven Classification and Cultural Variation
- Novel categorization schemes: LLMs could generate new groupings, beyond current categories, by being tasked with novel prediction/classification objectives, unconstrained by current labels. (46:19–47:56)
- Cross-cultural perspectives: Psychiatric classification is “emergent and high-dimensional,” akin to musical genres—different cultures may slice mental phenomena in ways that don’t map to current nosology. (47:56–51:10)
Notable Quotes & Memorable Moments
-
“One of the diagnosis that seems to have been missed by other clinicians a lot is obsessive compulsive Disorder.”
— Dr. Awais Aftab (00:08) -
“Checklist approaches… in both clinical practice and early AI development: That’s exactly how machine learning researchers started in AI and exactly what didn’t work. What ended up working is deep learning.”
— Dan Shipper (00:44; 31:04) -
“I take a more pragmatic kind of view towards these things… it shifts the focus away from certain metaphysical questions towards our ability to make use of concepts in the world.”
— Dr. Awais Aftab (16:15–16:37) -
“There’s a lot of pressure to reduce these things down into a checklist… but really a good clinician has a little bit of a smell for it…”
— Dan Shipper (24:03) -
“In science we prioritize explanatory power over predictive power… deep learning doesn’t have theories, it just says, we know it works.”
— Dan Shipper (33:06) -
“To the extent that there are problems that can be addressed in a somewhat satisfactory manner by AI, then why not, why not utilize that?”
— Dr. Awais Aftab (43:45)
Timestamps for Key Segments
| Time | Segment | |----------|-----------------------------------------------------------| | 00:00 | OCD misdiagnosis, fuzzy boundaries, pragmatic categories | | 03:30 | Explanatory pluralism in psychiatry explained | | 11:08 | Philosophy of science: realism, pragmatism, context | | 18:53 | Dan’s OCD journey, good/bad of psychiatric labels | | 24:03 | Clinical intuition vs. checklists, parallels in AI | | 28:36 | AI/deep learning for psychiatric classification | | 33:06 | Predictive power vs. explanation in science | | 38:38 | AI tools in medicine: notes and interviews | | 41:22 | ChatGPT as basic mental health support | | 43:45 | Human vs. AI therapy: strengths/limitations | | 46:19 | LLMs and emergent, novel categorization schemes | | 47:56 | Cultural variation, high-dimensionality in classification |
Final Thoughts
This episode provides a rich, philosophical, and technical journey exploring how AI could transform medicine’s most complex and “messy” arenas—like psychiatry—by helping us move from rigid, rule-bound systems to more fluid, example-driven, context-sensitive approaches. Dr. Aftab’s pluralism and pragmatism provide an illuminating framework for thinking about both clinical care and the future of AI-driven research.
For more from Dr. Aftab:
- Substack: Psychiatry at the Margins
- Conversations in Critical Psychiatry (Oxford University Press)
- Social: X (Twitter), BlueSky
Summary prepared for listeners who want deep context on AI, psychiatry, and philosophy—with memorable insights and logical connections to practical tech.
