Psychiatric semiology, that is, the science of clinical manifestations, considers that both symptoms and signs are “units of analysis.” These units are actionable psychopathological features that are essential in practice for making the diagnosis and prognosis that underpin clinical decision-making (Micoulaud-Franchi et al. Reference Micoulaud-Franchi, Quiles, Batail, Lancon, Masson, Dumas and Cermolacce2018). Psychiatric semiology is, therefore, very important. We comment on how the concept of thinking through other minds (TTOM) and the associated computational psychiatry model proposed by the authors not only questions the mechanisms underlying psychiatric manifestations (see Veissière et al. target article, sect. 5.2, para. 3), but also the ways in which these manifestations are expressed by patients and captured by psychiatrists. Indeed, we need a computational model that questions how clinical manifestations are expressed and captured. In the field of transcultural (Kirmayer & Crafa Reference Kirmayer and Crafa2014) and phenomenological psychiatry (Nordgaard et al. Reference Nordgaard, Sass and Parnas2013), it is widely considered that semiology is partially based on social and cultural construction (i.e., history of medicine, consensus of experts, folk psychology, etc.; Kirmayer & Ramstead Reference Kirmayer, Ramstead, Durt, Fuchs and Tewes2017). Moreover, with a hand outstretched to medicine, psychiatry considers that symptoms and signs are in some way linked to physiologic disturbances in the brain, as investigated by neuroscience (Micoulaud-Franchi et al. Reference Micoulaud-Franchi, Dumas, Quiles and Vion-Dury2016). Since Jaspers’ work, clinical manifestations have been taken to reflect both physiologic disturbances and patients’ attitudes toward them (Stanghellini et al. Reference Stanghellini, Bolton and Fulford2013). In this view, clinical manifestations are not just related to an underlying physiologic cause but also to cognitive-interpretive and interpersonal processes that are at play during the constitution of symptoms and signs (Kirmayer & Ramstead Reference Kirmayer, Ramstead, Durt, Fuchs and Tewes2017; Kirmayer & Sartorius Reference Kirmayer and Sartorius2007). Although theories have been proposed to account for these two dimensions central to psychiatric semiology (Berrios Reference Berrios1996; Borsboom et al. Reference Borsboom, Cramer and Kalis2018; Kirmayer & Ramstead Reference Kirmayer, Ramstead, Durt, Fuchs and Tewes2017), none has been formalized with a computational model. Although TTOM is a welcome addition that could help in formalizing clinical manifestations expressed and captured at the cultural/social level, we think that it should also take recent advances in the neuroscience of social interaction into account.

In the last decade, social neuroscience has indeed become interactive in acknowledging the impact of interpersonal social dynamics on intra-personal neurobehavioral dynamics (Hari & Kujala Reference Hari and Kujala2009; Redcay & Schilbach Reference Redcay and Schilbach2019). The second-person perspective (Schilbach et al. Reference Schilbach, Timmermans, Reddy, Costall, Bente, Schlicht and Vogeley2013) has already led to the development of a “second-person neuropsychiatry” that considers psychiatric disorders as “disturbances of social cognition” (Schilbach Reference Schilbach2016). Complementary to this perspective, two-body neuroscience and two-person physiology (Bolis & Schilbach Reference Bolis and Schilbach2018b; Dumas Reference Dumas2011) follow the call for radical embodiment in cognitive science (Thompson & Varela Reference Thompson and Varela2001) and emphasize the constitutive role of interpersonal dynamics in individual cognition. Hyperscanning, that is, the simultaneous brain recording of several people (Montague et al. Reference Montague, Berns, Cohen, McClure, Pagnoni, Dhamala, Wiest, Karpov, King, Apple and Fisher2002), has demonstrated how non-verbal interaction through sensorimotor loops modulates individuals respective internal neurophysiological dynamics and how interpersonal dynamics are measurable at the electrophysiological level through inter-brain synchronizations (Dumas et al. Reference Dumas, Nadel, Soussignan, Martinerie and Garnero2010). Although it does not negate the existence of higher-order representations (e.g., language, cultural habitus; Shea et al. Reference Shea, Boldt, Bang, Yeung, Heyes and Frith2014), it supports the development of a “social physiology” that continuously integrates sensorimotor and representational levels of analysis. Interestingly, TTOM is already trying to resolve the difference between the individual and the inter-individual, and the authors present embodied interaction as always being culturally coded by implicit cultural learning. However, in our opinion, there is also a more basic non-culturally coded and non-representational layer of interaction that is directly rooted in early developmental processes.

This layer of interaction is particularly relevant for studying neurodevelopmental disorders such as autism and schizophrenia. In autism, multi-scale approaches have already attempted to solve this social paradox (Bolis et al. Reference Bolis, Balsters, Wenderoth, Becchio and Schilbach2017; Dumas et al. Reference Dumas, Kelso and Nadel2014). In schizophrenia, the phenomenological approach has been used to account for impairment in the ability to learn the implicit social senses and use them in non-verbal communication (Fuchs Reference Fuchs2015). This involves a nonverbal and pre-individual layer of relationship (Lavelle et al. Reference Lavelle, Healey and McCabe2014) that is closely linked to minimal-self disorder as an alteration of the first-person perspective (Parnas & Zandersen Reference Parnas and Zandersen2018). This alteration raises the question how symptoms are expressed by patients and captured by psychiatrists.

This issue has been analyzed within the framework of the classical concept of “praecox feeling.” This “feeling” of bizarreness in interaction can be considered as a crucial determinant of medical decision-making in psychiatry (Cermolacce et al. Reference Cermolacce, Sass and Parnas2010; Gozé et al. Reference Gozé, Moskalewicz, Schwartz, Naudin, Micoulaud-Franchi and Cermolacce2018), because it is directly rooted in minimal-self disorders (Parnas Reference Parnas2011; Sass et al. Reference Sass, Borda, Madeira, Pienkos and Nelson2018). These well-documented first-person accounts suggest that patients are affected at a more basic level than TTOM. Hence, this suggests the existence of a sub-layer of TTOM itself which could be impaired, so the mechanisms underlying the involvement of TTOM require their own model. To meet this objective, social physiology calls for a computational model under (non-representational) and beyond (implicit or explicit social representations) the individual. Although Bayesian statistics can virtually integrate these dimensions (Friston & Frith Reference Friston and Frith2015a), they need to be captured by generative models that are based on other types of computational formalism (Friston et al. Reference Friston, Redish and Gordon2017; Montague et al. Reference Montague, Dolan, Friston and Dayan2011), especially biophysically-based neural circuit models (Wang & Krystal Reference Wang and Krystal2014). In our view, although TTOM provides a good matrix to model psychiatric semiology and its relationship both to physiology and to social interaction, it also requires the development of complementary computational models to account for physiological brain mechanisms and non-representational interpersonal dynamics (Dumas et al. Reference Dumas, Chavez, Nadel and Martinerie2012). The goal is to encompass the neurophysiological level and not separate the “implementation” from the “computational” in Marr's sense (Marr Reference Marr1982). Such TTOM combined with social physiology, including all three of Marr's levels (computational, algorithms, and implementation), could offer great perspectives for our understanding of how psychiatric manifestations are expressed and captured. This could help computational psychiatry to structure the classification of mental disorders, even including the more tacit mechanisms of intuition in clinical decision-making.