Five Main Research Areas
In general, my research can be divided into five closely related areas:
Conceptual questions in philosophy of medicine and psychiatry. Here I ask question like: What is disease? Is there such a thing? Why might that matter? What is the relation between diagnosis and explanation? Should nosology reflect natural kinds?
Modelling disordered conditions: With the help of cognitive sciences, I sketch models of conditions such as delusions, addiction, depression, and so on, in order to make them legible to ordinary psychological explanation.
Ethics and political philosophy of medicalized conditions. Here I as questions such as: Should MAiD be permitted for patients whose only diagnosis is psychiatric? Is autism a medical (as opposed to a political) category?
Moral psychology and the moral Emotions: What are the conditions on justified resentment? What are the condition on forgiveness? How do they apply in cases where people suffer mental illness?
Questions in philosophy of AI: Do mass-market LLMs understand? Do they meet the normative conditions on language use? What are the harms of use of AI for companionship or therapy? What are the social consequences of widespread AI use?
All of this is working out a tension that arises at the interface between what Sellars called the manifest and scientific images of the world. On the one hand we have the descriptions of us from the point of view of the best sciences; on the other hand, we have our normatively-laden folk psychological picture of world where persons and their reasons are central. Our moral practices — holding responsible, forgiving etc. — all presuppose the manifest image, so we have to either modify the image, modify the practices, or work out a synthesis.
It is generally thought that medicalized states are states that it is not good to be in. But many are psychological states with their own independent appropriateness conditions (e.g., justified depression); others (e.g. autism) are only controversially considered medical. That is the manifest image asserting itself over and against the scientific image. I want to know: in what cases should things go in that direction and in what cases should it go in the other direction?
Examples From Each Area
Conceptual Questions in Philosophy of Medicine: It is often assumed that it matters whether some controversial condition (e.g., addiction, obesity, depression) counts as a disease. But to see whether this is sensible, we need to know what diseases are. And the truth is: no one knows. Philosophers are divided over where the accept naturalism, the view that disease is a purely biological (value free) category, or normativism, the view that the concept of disease ineliminably involves appeal to values.
Most philosophers acknowledge that naturalism is false: there is no value-free way of saying what disease is. (Talk of disorder, dysfunction, pathology, etc. is covertly value-laden, perhaps.) But normativism faces an insuperable difficulty: distinguishing the disease bad (cancer) from the non-disease bad (poverty, conventional ugliness) in a non-circular way.
So, I argue that we should be eliminativists about the concept of disease. That is: There is nothing in the world that answers to our irremediably confused notion of disease. This might seem bad. Don’t we need this concept? I argue that instead of thinking in terms of diseases, we should simply talk about different bodily states and the various value judgments we might make about them. So, we describe obesity, addiction, depression, or any controversial condition as the kind of state that it is, and then say whether it is good/bad in a given context, for a given person, etc. That’s all, and that’s all that is needed.
Modelling Disordered Conditions: Cognitive scientists have worked up detailed models of many disordered conditions, but one thing we want to know is how they matter for ordinary psychological interpretation. For instance: Is the delusional person rational? Does the addicted person have control? Is the depressed person competent? Are any of these people morally responsible? Consider addiction: It is typical to wonder whether addiction is a disease (again, assuming that matters.) But part of what we care about most in addiction is how it should modulate our judgments and feelings towards those who suffer from it. That doesn’t seem to have anything to do with whether it is a disease. In a recent paper, I appeal to a leading neuroscientific theory of addiction (the incentive sensitization theory of Kent Berridge et al.) to argue that addiction involve sui generis psychological states which I call hybrid intentions. These states help us to model addiction at the level of ordinary psychological explanation, allowing ordinary judgment to get a purchase.
Or, consider delusions: Clinical delusions are both continuous with non-clinical irrationality and profoundly puzzling. How can someone seriously believe that they are dead, or that their spouse has been replaced by an impostor? In a recent paper, I argue (with Adam Bradley) that many theorists over-rationalize delusions such as these, mistaking the goal of theorizing (making sense of something) for the goal of rationalizing (making sense of someone)
Ethics and Political Philosophy of Medicalized Conditions: In Canada (following Belgium, the Netherlands), a recent legal change has opened the door for non-terminal patents to request MAiD. This could include patients whose only diagnosis is psychiatric. In a current paper, I argue that expanding MAiD to depression would be a mistake. Some severely depressed patients cannot seriously contemplate treatment alternatives and clinical tools for assessing decision-making capacity can’t distinguish disabling from non-disabling depression.
I have also written (with Sarah Arnaud) about the move to consider autism a political rather than a medical category. This position, advocated by those in the neurodiversity movement, has been subject to a lot of mischaracterization which is the result, I argue, of misunderstanding the sort of identity in term of which activist couch their position. Properly understood, the neurodiversity position is neither anti-science nor opposed to access to medical treatment for autistic subjects.
Moral Psychology and the Moral Emotions: Going back to my doctoral dissertation, I have had an interest in moral responsibility and the moral emotions. (I believe these are deeply connected, as I am a Strawsonian about responsibility.) In a recent iteration of this sub-program, I argue that forgiveness has externalist identity conditions: whether a psychological process is one of forgiveness depends constitutively on whether the reasons for the psychological change away from resentment are of the right type.
So resentment-reduction is a disjunctive type: there can be phenomenologically indistinguishable processes of coming to be less angry which are nevertheless metaphysically distinct because one and not the other was done in response to the right reasons This is an example of a more general process-product connection in psychology: sometimes the identity conditions on a psychological state or process are normative (such as in self-deception, or in veridical perception).
If the normative conditions on forgiveness — when should forgiveness be given? — are the same as the identity conditions on forgiveness — what makes something forgiveness — then the traditional ‘paradox’ of forgiveness, according to which it is not possible to satisfy the normative and identity conditions on forgiveness simultaneously, is dissolved.
Philosophy of AI: In a paper in-progress I argue that the use of AI as companions and therapists. I argue, following John Campbell, that there is a form of understanding — knowledge of the causal structure of another persons’ mind — that is not available to an LLM trained by gradient descent. The causal structures which are grasped by our valuable companions and those who have therapeutic insight into our minds are irreducibly particular and are only know by their counterfactual profile. But the capacity for this kind of counterfactual grasping presupposes propositionally structured thought and the ability to entertain a content under different modes.
Thought in these different modes is governed by norms which, I argue, should be given a Wittgensteinian analysis and to which no LLM is sensitive. Insofar as this form of interpersonal understanding is essential for companionship and therapy, AI companions and therapists are defective and pose risks not only to human well-bring, but to our typical conception of what it is to share mental life with another.
Future Work
I plan to continue working in all of the areas outlined above. In the near-term I plan to revisit some of my earlier work on self-deception. Earlier I argued for a deflationist view of self-deception, one according to which the deception of self is not done intentionally (de dicto). I now think that there are two different kinds of self-deception that correspond to two different kinds of belief.
Further Work in Philosophy of AI: In addition to the above-mentioned article in-progress, I have two others I am planning to present and write up in the near future. In the first of these I argue that the there is a constitutive connection between experience (which LLMs presumably lack), meaning, and normativity. I argue that according to a Wittgenstein-inspired social-pragmatist view of linguistic normativity (such as Brandom’s), a Kant-inspired internalist view of normativity (such as Ginsborg’s), and a existentialist phenomenologist view of linguistic normativity (such as Heidegger’s), experience, is a necessary condition for being a genuine meaning user. Each view illuminates a distinct way in which interaction with LLMs degrades our capacities for meaning-making.
In the second, I focus and the social and political risks of continued unfettered AI development, particularly the push towards AI with more sophisticated agentive capacities — e.g., embodiment, the ability to track patterns in human behavior, revise its strategies, and maintain long-term interactions. I argue that every problem that we should currently take seriously when it comes to dismembodied LLMs, related as they are, — user addiction and the erosion of social togetherness, hallucination and misinformation, the concentration of cloud capital in the hands of private powerful firms, and the existential risk to humanity — is exacerbated by the development of AI agents
Book Project: My most significant near-to-medium term project is a monograph about addiction, tentatively called Addiction From the Human Point of View. The book expands on my earlier work on addiction, briefly described above. The first section of the book considers in detail the so-called incentive sensitization theory of addition from Kent Berridge and colleagues. According to this theory, long-term exposure to addictive drugs causes neurobiological changes which sensitize attentional circuits to the relationship between drugs and their cues. I argue on grounds of empirical adequacy and explanatory power that this theory is preferable to its rivals and that, in some cases, is complementary to them rather than being explanatory competition with them.
I argue that the incentive sensitization theory entails the existence of a distinctive motivational state in addiction which I call a hybrid intention. Hybrid intentions are in some respect desire-like and in other respects intention-like. They are desire-like in that they come unbidden — or, more precisely, they come in response to encountering drug cues — and they are not under the direct control of the will; but they are intention-like in that they typically lead to action unless inhibited.
This framework allows for detailed subsequent discussions of:
Addiction and pleasure: Addictive behavior being caused by hybrid intentions explains why addiction persists even when — as consistently reported by long-term drug users — use no longer produces pleasure. The positive value of drug use, when it does produce pleasure, is orthogonal to the moral disvalue of addiction asa condition of agency.
Addiction and judgment shift: Just as some purely affective theories of addiction fail, purely cognitive theories, which often explain addiction in terms of judgment shift, also fail. Judgment shift is secondary to the operation of hybrid intentions — the triggered hybrid intention causing the judgment shift—and/or is epiphenomenal — the ‘judgment’ in question lacking the customary connection with volition typically present in practical judgment.
Addiction and the good life: Being vulnerable to having one’s action partially determined by a drug cue is a liability with respect to acting well. Insofar as living a good life requires being able to orient oneself successfully towards the good in action, having the neuroadaptations underlying a susceptibility to hybrid intentions in an obstacle to living well.
Addiction, abstinence, and 12-step programs: Traditional 12-step programs recommend total abstinence. Theorizing addiction in terms of hybrid intentions explains why this strategy is successful despite not being strictly necessary: Avoiding environments with cues is key to avoiding the triggering of hybrid intentions. Moreover, because drug consumption itself is a drug cue, moderate consumption is difficult to maintain.
Addiction and responsibility: Hybrid intentions do not determine that the agent will act on them, as evidenced by the fact that successful abstention is an everyday occurrence. All the same, because they typically lead to action unless inhibited, they change the pattern of default action. The pertinent question about addiction and responsibility then becomes: is a subject blameworthy for a failure to overcome a pattern of default action? This question is answered, in part, with reference to the degree of social and decision-architectural support available.
Addiction and disease: The traditional theoretical question about addiction is whether it is a disease. However, the whole story offered here is independent of that issue. What matters, from this point of view, is how the cognitive dynamics — of which we get a detailed picture from this theory — make a difference to what matters from the human point of view.