How Psychiatric Diagnostic Reasoning Works
A Review of Adrian Kind's "How Does the Psychiatrist Know?"
Philosophical clarity is hard-won. It may seem easy in retrospect, and the insights may seem obvious, but getting to that point is often a slog. I rely on and benefit tremendously from the work of philosophers who often don’t get enough credit for their efforts. Admiration and gratitude for hard-won conceptual clarity were on my mind as I finished reading Adrian Kind’s 2025 book, “How Does the Psychiatrist Know?: On the Epistemology of Psychiatric Diagnostic Reasoning” (pdf available open-access), which is based on his PhD thesis.
Adrian Kind is a philosopher, clinical psychologist, and psychodynamic psychotherapist in advanced training, based in Germany. He is not a psychiatrist, but it is obvious from the book that he is well-informed, has attended a lot of (excellent) psychiatric case conferences, and has a good sense of what psychiatric reasoning looks like under ideal conditions.1 Kind’s work is a refreshing departure from that of many philosophical critics of psychiatry who have never had an opportunity to see an adroit diagnostician in action.
Kind’s monograph is the first systematic, in-depth, book-length philosophical investigation into how psychiatrists arrive at their diagnostic conclusions. More precisely, it addresses the question: what is the method of proper contemporary psychiatric diagnostic reasoning? Kind’s answer is that psychiatric reasoning is fundamentally a modeling process that can be described as qualitative, constitutive diagnostic modeling. That is augmented by pattern matching and other inferential processes.
Kind proposes that psychiatric reasoning is fundamentally a modeling process that can be described as qualitative, constitutive diagnostic modeling.
To understand what this means, let’s go through Kind’s discussion of modeling in general before it is applied to psychiatry.
Modeling based on limited, initial information
Kind starts with Godfrey-Smith’s and Weisberg’s notion of indirect strategy of representation in theoretical modeling. Someone using this method follows three steps. First, they create a simplified model using limited, initial information about the phenomenon they’re studying, with parts of the model assigned to represent parts of the real system. Second, they closely examine the model they are using to understand what it says and predicts about the real-world system. Third, they compare what they’ve learned from their model to what actually happens. This helps them decide whether, and to what extent, their model is useful. (The direct strategy, in contrast, is data-driven theorizing based on generating and collecting large amounts of data.)
The construction of the simplified model itself is a messy process that is guided by the existing scientific and clinical knowledge, theoretical presuppositions, and prior experiences and expertise—something that in medicine is often described as the art and craft of medical practice.
Qualitative modeling
The theoretical model that is created in response to initial information could be quantitative or qualitative. Quantitative models capture relationships in precise, mathematical terms. In qualitative modeling, which is what Kind is interested in, aspects of the real-world system are represented in a discrete manner. The states may be described, for instance, as “present,” “absent,” or neutral,” and relationships between variables may take the form of “increases,” “decreases,” or “irrelevant,” rather than mathematical formulae. Qualitative models can be propositional, specifying elements and relationships in a model in natural language.
Qualitative models can use two forms of idealizations: Aristotelian and Galilean. In Aristotelian idealization, many aspects of the real-world system are left out of the model to make the model easier to work with. In Galilean idealization, features are included in the model, but their complexity is intentionally reduced (for example, looking at educational performance only in terms of whether someone has graduated high school or not).
Diagnostic modeling
When models are used to identify and classify irregularities in a system, it is diagnostic modeling. Diagnostic modeling uses a presupposed model of normal functioning of the system to help decide whether errors have occurred based on an examination of inputs and outputs. Differential diagnosis occurs when a system produces an error, and the person analyzing it realizes that multiple possible error models could explain this problem. In this situation, they compare each possible error model from an “error taxonomy” to the system’s behavior and choose the one that best matches the scenario. Exclusion diagnosis is used when a system produces an error that could have multiple possible causes, but the person analyzing it doesn’t have a complete set of error models to choose from. If no existing model in the error taxonomy accurately explains the error, the error is diagnosed “by exclusion.” This means they’ve determined what the error likely isn’t, even though they haven't positively identified what it is. Basically, exclusion diagnosis helps narrow down possibilities by ruling out certain causes.
The above can also be described in terms of two inferential patterns: differential diagnosis as abduction or inference to the best explanation2 and exclusion diagnosis as apophatic inference (“knowledge obtained by negation”).
What makes modeling “constitutive”?
The relevant contrast here is with causal modeling, where models represent the causes of something, while constitutive modeling is about the properties and characteristics of a phenomenon. A constitutive model tells us about characteristics and core features, but not why it happened (causes) or under what circumstances it will happen in the future (prediction). For example, a constitutive model of “delusions” doesn’t tell us what causes delusions, nor does it predict when delusions might appear, but it tells us what counts as a delusion (fixedness, false belief, resistance to contrary evidence, etc.).
How does this apply to psychiatric diagnostic reasoning?
Kind describes psychiatric assessment as consisting of initial patient screening and subsequent in-depth evaluation. The screening process uses limited, initial information about a clinical problem to determine whether an error in the system is present. This leads to theoretical modeling in natural language prepositions about various diagnostic possibilities, with details about the conditions under which those possibilities would be a good fit.
The initial information is usually gathered directly from patients during the first clinical meeting. Psychiatrists ask patients why they’ve sought psychiatric help, observe their behavior, and ask specific questions or use brief tests. Additional sources like admission records or family reports can also provide information. This first step helps psychiatrists create an initial list of possible mental health concerns. This list typically includes patients' self-reported experiences (like trouble sleeping, feeling sad, or constant worry) and observed behaviors (like restlessness, difficulty concentrating, or speaking very little). Based on this, psychiatrists develop hypotheses about what might explain the patient’s issues and decide whether they need to explore these issues more deeply to confirm if they reflect a psychiatric disorder.
Kind focuses on two detailed examples to illustrate this: poverty of speech (alogia) and constant worrying.
In the case of poverty of speech, at the screening stage, a psychiatrist may consider the following possibilities:
“A. The patient did not want to consult the psychiatrist but does so to satisfy relatives or friends who pressure him to do so.
B. The patient might have an unusually pedantic way of speaking, not associated with any morbid condition.
C. The patient may have taken drugs impairing his language-related cognition – e.g., cannabis (Dellazizzo et al., 2022).
D. The patient might have had a traumatic brain injury (TBI) that could have led to this condition.
E. The patient might suffer from specific cognitive deficits in language processing responsible for his speaking behaviour.”
And for constant worry, they may think of:
“a. The patient’s worry may be the result of an increase of arousal occurring in response to dealing with current high-stress or hostile circumstances.
b. The patient takes medication or drugs on a regular basis that, depending on the dose, can cause anxiety reactions (e.g., corticosteroids or caffeine)
c. The patient’s constant worrying turns out to consist in thoughts coming to his mind whose content is not particularly distressing but that cause higher-order distress because of their undesired persistence and their negative appraisal.
d. The patient’s worry results from the anticipation of or reaction to a specific frequently occurring stimulus (e.g., a type of situation or object) that he is afraid of to a degree that seems extraordinarily high given its nature.
e. The patient’s worry is a specific stimulus-independent reaction to expectation of unlikely menacing events and more likely but unthreatening events.”
Psychiatrists come up with such hypotheses and rely on their knowledge base to evaluate each hypothesis. They have an idea of what additional information would have to apply to the person’s presentation to support each of the options they are considering. This leads to an in-depth evaluation.
To see how this unfolds, take the possibility in the case of suspected alogia that the person may just have an unusually pedantic way of speaking.3 This possibility is further addressed through setting up the following prepositions:
“The patient can elaborate their answers if asked to.
The patient recognises that their answers are unusually short and can justify their manner of speaking by explaining their motivation (e.g., wanting to save the doctor’s time, or wanting to be as precise as possible).
The patient can report that his way of presenting information is not something that has developed recently but is rather their normal way of conveying information.
If possible, the psychiatrist also speaks to relatives, friends, or other medical professionals to verify the statement that:
People who know the patient report that the patient has always tended to speak this way.”
For the presentation of constant worry, the psychiatrist may set up the following prepositions to explore the possibility that the worry is an expected or typical response to high stress:
“The patient is currently in a highly stressful or hostile life situation (e.g., currently being threatened and followed by an ex-partner, or having lost his job and being in significant debt and about to become homeless) that makes the worry reaction seem appropriate.
The patient’s increase in worry coincides with the occurrence and duration of the stressful life circumstances.
The patient’s worries directly concern the source of worry, or the topic of worry is closely linked another worry or a hostile experience. Alternatively, the worry may concern an occurrence that while under normal circumstances would be no problem, is experienced as being an issue because it comes “on top” of the actual severe problems that cause other, primary worries.”
And to explore the possibility of worry being due to obsession and compulsions, the psychiatrist can check whether the patient has thoughts that:
“are distinct thoughts, primarily experienced as visual (i.e., visual mental imagery) entering conscious awareness
are attributed to an internal origin (i.e., the patient assumes ownership of these thoughts)
are considered unacceptable or unwanted due to their egodystonic nature (i.e., their content is inconsistent with the subject’s self-image or moral convictions)
are evoking significant feelings of shame
are interfering in ongoing cognitive and/or behavioural activity
are unintended and nonvolitional or have wilful independence
are recurrent or repetitive
are difficult or impossible to control or dispel
arise more frequently under increased stress”
You get the idea. This is all bread and butter for psychiatric clinicians, but it's wonderful to see the logic of it spelled out so clearly and methodically.
Psychiatrists observe the patient’s initial behavior and listen to their initial complaints. Based on these initial observations, psychiatrists form early ideas—diagnostic hypotheses or “models”—about what might be causing the patient’s issues. The “models” that psychiatrists use are simplified representations of psychiatric disorders and their symptoms in the form of propositions and relationships among them. These models are informed by their medical training, their scientific understanding of psychiatric phenomena, and common-sense knowledge of human behavior.
They then evaluate the patient more closely, looking specifically for signs and symptoms that fit these initial models. To do this, they compare the patient's experiences and behaviors, and sometimes results from cognitive or medical tests, against various models representing psychiatric conditions and possible alternative explanations. By carefully comparing how well each model matches the patient, the psychiatrist determines whether certain diagnostic possibilities are present or absent.
A comparison between models and clinical presentation takes place, and the occurrence of differential diagnostics and exclusion diagnostics. Psychiatrists then combine the individual models into a complete set that takes into account all of the patient’s main concerns and clinical complaints. Under ideal conditions, the psychiatrist also records the evidence that led them to select the particular models they did. This combination of selected models and supporting evidence becomes the “diagnostic case formulation,” a summary that helps clinicians think clearly about and reflect critically on their own diagnostic reasoning and discuss cases effectively with colleagues.
And finally, psychiatrists engage in a pattern-recognition process enabling them to make syndromal diagnoses in accordance with the rules of a classification system. Experienced psychiatrists often recognize patterns intuitively. They also use standardized diagnostic systems, such as the DSM or ICD, to classify symptom patterns into formal psychiatric diagnoses. Kind suggests, as have others, that this is best understood as a form of prototype-based pattern recognition. Competent psychiatrists are familiar with the relevant combinations of symptoms that are the most typical or most illustrative examples of a mental health problem.
I won’t go into a discussion of this (see chapter 5 in the book), but this modeling account fulfills many features we would desire in an account of diagnostic reasoning: it is cognitively realistic, it allows us to distinguish between misdiagnosis and diagnostic malpractice, it explains the occurrence and resolution of diagnostic uncertainty, it provides an understanding of how diagnostic disagreements occur, and it offers guidance for thinking about the relationship between diagnostic reasoning and our evolving scientific understanding of psychopathology.
In my 2024 paper, “Psychiatric Diagnosis: A Clinical Guide to Navigating Diagnostic Pluralism,” my co-authors and I wrote briefly about the diagnostic process behind the diagnostic labels:
“Diagnosis is both the process and the outcome of ascertaining the nature of a clinical problem. When we look at diagnostic manuals such as DSM, we are looking at the denotation part of the process. The way clinicians or researchers use the framework is at least as important, if not more so. For example, when it comes to the DSM, clinicians rarely, if ever, use Structured Clinical Interview for DSM Disorders (SCID), which is generally reserved for research settings. Instead, clinicians often rely on prototypical descriptions of DSM psychiatric categories. They use the presenting complaints and initial history of the patient to rapidly create a differential diagnosis, use that differential diagnosis to guide their clinical assessment, and then refine or revise the differential diagnosis as the assessment proceeds (Huda, 2019). After they have established that the presentation approximates or matches a diagnostic prototype, they might consult the precise diagnostic criteria, and determine, for instance, whether the patient meets full criteria or is subthreshold or better considered in the “other specified” category. Many clinicians complement the diagnosis with a narrative diagnostic formulation that summarizes their conceptualization of predisposing, precipitating, and perpetuating factors.” (Aftab et al, 2024)
It did not occur to me at the time to think of this as a qualitative constitutive diagnostic modeling process, but having read Kind, this conceptualization comes easily to me now. I see Kind’s book as my new go-to reference for psychiatric diagnostic reasoning, and I suspect that this will also be the case for many other academics who work in this area. Any future conceptual work on this subject would build on or would have to seriously engage with Kind in one form or another.
Kind’s analysis sets the stage for some interesting future philosophical work on how psychiatric diagnostic reasoning interfaces with metaphysical debates about the nature of psychiatric constructs, Bayesian inference, epistemic risks of diagnostic modeling under suboptimal clinical conditions, diagnostic reasoning in psychological formulations, and methodological parallels between clinical diagnosis and self-diagnosis.
I’ll give the last words to Kind himself:
“By introducing and defending the model-based account as the first systematic and fully mapped out approach to applying debates about modelling in philosophy of science to the topic of medical diagnostics, more specifically to psychiatric diagnostics, I hope to have made a stimulating contribution to debates in the epistemology of psychiatric diagnostics, a still small aspect of the overall debates in the field. I also hope to have provided an example of how debates from general philosophy of science, in this case about modelling, can be made fruitful in the application to discussions of aspects of special sciences and practices such as psychiatry. Finally, if anything I have done in these pages impacts any of the pragmatic concerns of psychiatry that I mentioned above—if it ever proves useful in medical education, inspires debates about policies on how to differentiate misdiagnosis, or inspires a developer in health tech to come up with a useful program or device, and indeed if it ever leads to something that that is of help to anyone seeking psychiatric treatment—I will consider this research to have served its purpose. Future work in these areas, through which I hope to build on this thesis, will help to make this hope a reality.” (Kind, 2025)
Psychiatry at the Margins is a reader-supported publication. Subscribe here.
See also:
The psychiatrist Henrik Walter was his PhD supervisor, along with the philosopher Sascha B. Fink.
Kind writes, “While inductions aim to make inferences about future or unobserved events, abductions aim to infer something about the unobserved causes or explanatory reasons for an observed event (Aliseda, 2006),” and quotes Schurz (2008), “the crucial function of a pattern of abduction or IBE consists in its function as a search strategy which leads us, for a given kind of scenario, in a reasonable time to a most promising explanatory conjecture test which is then subject to further test.”
Kind writes with understated humor, “This is a problem that, as the literature indicates (Andreasen, 2016), has been observed particularly in interaction with administrators, politicians, scientists, and (perhaps unsurprisingly) philosophers.” (p 101)
A nice overview of a nice book!!
Great review!
Too often I hear that psychiatric diagnosis is based purely on number of arbitrary symptoms +/- blindly following DSM criteria.
The explanations discussed here are much more nuanced / more true to my experience in clinical practice