Biocognitive Classification of Antisocial Individuals Without Explanatory Reductionism

Syndrome-Based Classifications and Research Domain Criteria

Although there is no tendency in syndrome-based systems to deny that mental illnesses depend on biological processes, these processes are not relevant for classifying them. According to DSM–5, for instance, the core of the diagnosis of ASPD is defined by a list of mostly behavioral symptoms. To be classified as having ASPD, a person needs to be at least 18 years old, show conduct disorder before age 15, and satisfy at least three out of these seven criteria: (a) failure to conform to social norms, (b) deceitfulness, (c) impulsivity or failure to plan ahead, (d) irritability and aggressiveness, (e) reckless disregard for self or others, (f) consistent irresponsibility, and (g) lack of remorse (APA, 2013).

This type of syndrome-based approach was adopted beginning with DSM–III (APA, 1980) and has been influenced by empiricist and neopositivist views on classification (Fulford & Sartorius, 2009; cf. Cooper & Blashfield, 2018). The main aim was to build “atheoretical” systems of classifications that would enable reliable diagnoses that were based on directly observable behavioral symptoms and mental states or traits inferred from these behaviors. In addition, this approach was adopted for purposes of communication and standardization that would enable epidemiological studies, research, psychiatric treatment, and use of diagnoses in social practices such as insurance, forensic application, and so on (Cooper, 2005).

The DSM approach to psychiatric diagnosis has been criticized within the relevant sciences (see e.g., Buckholtz & Meyer-Lindenberg, 2012; Cuthbert & Insel, 2013; Insel et al., 2010; Lilienfeld, 2014; Lilienfeld, Smith, & Watts, 2013) but also by philosophers (Murphy, 2006, 2017; Tabb, 2015, 2019a). The core of these criticisms is that this type of approach, for the most part, produces categorical structures that have low validity and are defined by polythetic symptoms that are shared across different disorders (i.e., indicate extensive comorbidity). Thus, the ensuing classifications tend to cover widely heterogeneous groups of people that for clinical and research purposes should be kept apart (Lilienfeld, 2014; Lilienfeld et al., 2013; Murphy, 2006; Poland & Von Eckardt, 2013).

The general difficulties that stem from syndrome-based classifications have motivated the elaboration of alternative classificatory systems that put emphasis on biological factors. The RDoC project is a recent and influential example of these approaches (Buckholtz & Meyer-Lindenberg, 2012; Cuthbert & Insel, 2013; Insel et al., 2010). The goal is to develop, for the purposes of research, new classifications that will not be constrained by the categories as encapsulated in the DSM and ICD. Rather, classifications should take a dimensional approach and rebuild categories bottom up by using the available data on the genetic, neural, cognitive, and affective systems underlying the spectrum between normal and abnormal behavior. The aim of RDoC is, thus, to transform the classification systems of psychiatric conditions. The ensuing classifications will integrate medically relevant data spanning from genetics to cognitive, affective, and social neuroscience to transform diagnosis, provide valid measures of disorders, aid clinical practice, and improve health outcomes (Cuthbert, 2014; Cuthbert & Insel, 2013; Insel et al., 2010).

The RDoC approach was introduced as an initiative that might improve prospects for developing personalized medicine in psychiatry, thus getting closer to other branches of medicine in which classifications of health problems rely on objective biological markers and less on personal reports and phenomenological symptoms (Cuthbert & Insel, 2013). This approach, which emphasizes biological factors in devising classifications, is based on several assumptions (Insel et al., 2010). One is that mental disorders should be reconceptualized as brain disorders, specifically as functional disturbances of the brain circuits. The second is that dysfunctions of brain circuits could be identified by relying on the methods and data from the neurosciences. The third is that the data from neuroscience and genetics will provide biomarkers that will enhance current symptom-based ways of identifying and treating medical conditions.

The RDoC framework consists of six predefined major domains that provide a platform for carrying further research in psychopathology. These six domains include negative valence (e.g., dimensions acute threat/fear, potential threat/anxiety), positive valence (e.g., approach behavior, reward learning), cognitive systems (e.g., attention, perception, memory, language), systems for social processes (e.g., affiliation and attachment, social communication, perception of self and others), arousal/modulatory systems (arousal, sleep-wake cycles), and sensorimotor domain (e.g., motor action, agency and ownership, innate motor patterns; see National Institute of Mental Health, 2019). The dimensions comprising the six domains represent constructs that are further operationalized at different levels of analysis (encompassing data from genetics, neuroscience, and behavioral self-reports) and provide a platform for research. The constructs represent dynamic frameworks that will be validated and revised as part of an ongoing scientific investigation of different manifestations of psychopathology.

To sum up, the RDoC approach has the original aim of replacing DSM and ICD diagnostic categories with new ones based on biocognitive research. Accordingly, in early stages of research, the researchers focus on people exhibiting symptoms they would like to study. For instance, studying anxiety disorders would include going to a clinic and taking the biocognitive data from patients exhibiting anxiety symptoms without worrying how these individuals would be classified in terms of DSM or ICD categories (Cuthbert & Insel, 2013). Closer to our case, the investigation would start by visiting inmates in a forensic institution who exhibit severe forms of antisocial behavior without antecedently delineating the group according to how they would score on different syndrome-based measures (Brazil et al., 2018).

In the next section, we illustrate how the use of biological and cognitive information can be applied to overcome difficulties emerged by adopting syndrome-based categorizations of people who engage in various forms of antisocial behavior.

A Biocognitive Approach to Antisocial Behavior

The study of antisocial personalities, especially in the case of psychopathy, has recently seen considerable advances. Studies have uncovered functional, neuroanatomical, neurophysiological, and even genetic characteristics that associate with this condition. After a brief survey of these results, in the remainder of this section, we consider how these advancements but also difficulties related to the current syndrome-based categorizations of psychopathy motivate the adoption of a biocognitive approach.

A wealth of research on psychopathy has shown emotional deficits taken to be related to low self-regulation and antisocial behavior. Studies, in fact, have shown peculiar affective processing underlying associative learning, emotion recognition, and empathic responses (Blair, 2005, 2008; Gregory et al., 2015). The development of empathy seems to depend largely on the ability of experiencing own emotions, and there are indeed indications that the affective component of empathy is impaired in psychopathy (Decety, Chen, Harenski, & Kiehl, 2013). There are also studies concerning a more general aberrant physiological responsivity in autonomic systems and executive abilities underlying moral reasoning, emotional regulation, and self-regulation in personal conduct and in response to social cues (Glenn et al., 2015; Gregory et al., 2012; Kiehl, 2006; Yang et al., 2015).

Although some researchers have assumed that emotional impairments are key to psychopathy (Blair, 2013; Lykken, 1995; McCord & McCord, 1964), there is some evidence that recommends a more nuanced view. A meta-analysis by Jolliffe and Farrington (2004) found that affective empathy was only weakly related to offending. Meta-analytic studies have also challenged other dominant beliefs related to affective experience in psychopathy. For instance, it was found that psychopathic traits are also linked to reduced processing of threat-related information but not to reduced conscious experience of fear as an emotion (Hoppenbrouwers, Bulten, & Brazil, 2016). A meta-analysis by Dawel, O’Kearney, McKone, and Palermo (2012) also found no statistically significant effects supporting the notion that emotion recognition, which is believed to be key to the development of empathy, is generally impaired in adults with high psychopathic tendencies (see also Derefinko, 2015).

There is also a large body of work showing that disturbances in other cognitive domains, such as attention and reinforcement learning, play important roles in psychopathy (Blair, 2013; Koenigs & Newman, 2013; Smith & Lilienfeld, 2015). In particular, attentional deficits might be also related to deficits in self-regulation (Newman & Baskin-Sommers, 2011) given that there is correlation between aberrant focusing of attention during decision-making tasks and social cognition in psychopaths (see e.g., Drayton, Santos, & Baskin-Sommers, 2018; Koenigs & Newman, 2013; Smith & Lilienfeld, 2015).

In recent years, the biological and cognitive underpinnings of antisocial behavior and psychopathy have also been intensely investigated (for recent reviews, see e.g., Brazil & Cima, 2016; Glenn & Raine, 2014). For instance, brain-imaging studies have indicated that psychopathy is correlated with aberrant brain activity in the prefrontal, insular, and limbic regions, which might underlie maladaptive features of psychopathy that are related to emotion processing, reward learning, impulse inhibition, and social cognition (Poeppl et al., 2019). Another study indicated that different facets of psychopathy in a sample drawn from the general population correlated differentially with general efficiency of neural processing (Tillem, van Dongen, Brazil, & Baskin-Sommers, 2018). In particular, higher interpersonal-affective traits were associated with less efficiency in long-range communication in a band of neural oscillations linked to higher order processes, such as decision-making and affective theory of mind (e.g., emotion recognition), whereas lifestyle and impulsive traits were linked to hyperefficiency in a frequency band of neural oscillations that seems to be involved in lower order processes, such as sensory perception. Such differences in neural processing point toward differential patterns of correlations between the different psychopathy traits and disruptions and abnormalities in information processing and decision-making (Baskin-Sommers, Brazil, et al., 2015).

Moreover, genetic studies on antisocial conduct have indicated that genetic and environmental factors interact in determining behavioral and psychological tendencies that characterize individuals exhibiting severe forms of antisocial behavior (Burt, 2009; Rhee & Waldman, 2002; Viding, Jones, Frick, Moffitt, & Plomin, 2008; Werner, Few, & Bucholz, 2015).

These encouraging advancements, however, emerge within a scientific paradigm that is afflicted by the kind of problems associated with syndrome-based classifications. In fact, categories such as ASPD and psychopathy capture wide and heterogeneous groups of people, including individuals who often exhibit externalizing behaviors, affective deficits, and peculiar interpersonal styles, which themselves again comprise further subtypes (Brazil et al., 2018; Brzović, Jurjako, & Šustar, 2017). For instance, Brazil and colleagues (2018) wrote,

This diversity in the classification of antisocial personality and psychopathy has led to a greater understanding of the need to differentiate between the two, but also to confusions and disagreements (especially after the introduction of the DSM). In general, antisocial personality has been associated with behaviors characterized by, e.g., impulsivity, excessive reward seeking, and hostility and reactive aggression. In contrast, psychopathy has often been linked to difficulties in establishing genuine relationships, shallow affect, in addition to chronic antisocial behavior. Antisocial personality and psychopathy also show different patterns of comorbid disorders. Although the distinction between antisocial personality and psychopathy is now accepted, there is still disagreement on how to best characterize these constructs and how to specify the different subtypes. Especially the subtyping of psychopathy has received a relatively large amount of attention, but there is still no consensus on this matter. (p. 264)

Even for the psychopathy category, despite the wealth of biocognitive data, there is no agreement on how to distinguish between its subtypes. This heterogeneity is visible in different and often incompatible accounts. For instance, different authors have distinguished between successful and unsuccessful psychopaths (sometimes defined as those with and without criminal records), ¹ sociopathy and psychopathy proper (environmental vs. genetic etiologies), primary and secondary psychopathy (low-anxiety vs. high-anxiety psychopaths), and so on (Skeem, Polaschek, Patrick, & Lilienfeld, 2011). The disagreement about the best way to conceptualize psychopathy has also fueled disparity about the best way to measure the construct (Miller & Lynam, 2012; cf. Lilienfeld et al., 2012). Although diversity is not a problem per se, in the case of psychopathy, it has often led to inconsistent empirical results within and across studies (Baskin-Sommers, Brazil, et al., 2015; Brzović et al., 2017; Jurjako & Malatesti, 2018b; Schultz, Balderston, Baskin-Sommers, Larson, & Helmstetter, 2016).

For example, Baskin-Sommers, Brazil, et al. (2015) showed that within a single sample of offenders, depending on which measure of psychopathy is used, one can get opposite correlations with external criteria. In this specific study, they tested how psychopathy scores in a group of offenders correlated with a battery of executive function tests. When they used the Psychopathy Checklist–Revised (PCL-R; Hare, 2003), a measure of psychopathy commonly used in forensic settings, they either received zero or negative correlations with psychopathic traits. Different results were obtained by using a self-report measure that is based on conceptualizations according to which core psychopathic traits are characterized by fearlessness and dominance in interpersonal relations (see e.g., Fowler & Lilienfeld, 2013). These analyses showed positive correlations between psychopathy scores and performance on the same executive function tests and in the same sample, which contradict the findings obtained using the PCL-R and also common views according to which psychopaths have deficits in cognitive and affective self-regulation (see e.g., Koenigs & Newman, 2013).

Thus, important problems seem to be related with the fact that currently available syndrome-based approaches to antisocial behavior cover heterogeneous groupings outlined by a variety of measures and conceptualizations that often share symptoms with other DSM and ICD categories. This, in turn, creates difficulties in investigating the biological correlates of maladaptive behavior and using such information for devising new clinical studies and treatments. These problems are especially clear in psychopathy research, in which there is still an overarching pessimism about successful treatment and rehabilitation options (Brazil et al., 2018; Salekin et al., 2010; Skeem et al., 2011). These issues motivate biocognitive approaches to psychiatric and forensic nosology because these seem to have the potential to provide more nuanced information about group characteristics. Furthermore, and more importantly, there is a scarcity of proposals and discussions on how to develop an overarching approach aimed at using and systematizing biocognitive data into a coherent framework and translate them into therapeutic and intervention practices.

Brazil and colleagues (2018) provided such a framework in the spirit of RDoC. The principal difference between the biocognitive approaches to antisocial behavior and the syndrome-based approaches concerns the emphasis on the role of biological and cognitive data in guiding the classification of the relevant conditions. Syndrome-based diagnoses of antisocial personality disorder (DSM–5) and psychopathy, as measured with the PCL-R (Hare, 2003) or other diagnostic tools (Fowler & Lilienfeld, 2013), characterize and build up a psychosocial profile of an individual and then conduct further experimental or correlational studies. In the RDoC approach, such tools would not be completely abandoned. However, the emphasis on them would be attenuated in favor of more bottom-up approaches that would enable far greater reliance on existing biological and cognitive data to rebuild and further stratify the existing categories of antisocial individuals. In this type of classification, the crucial step is collecting all the available biological, behavioral, and cognitive data and, ideally, using them in unison to form more homogeneous subgroups. This is believed to improve predictive accuracy and facilitate research on the underlying causes of different forms of antisocial behavior, ultimately leading to more effective ways of treating and preventing antisocial behavior.

Brazil and colleagues (2018) moved beyond RDoC in several respects. Note that their framework is not bound by predefined dimensions and their operationalization, as in the RDoC approach (see Cuthbert & Insel, 2013). Nonetheless, in the spirit of RDoC, they offered an integrative approach to biological and cognitive factors that can be used to capture individual differences in antisocial populations. The approach involves three stages.

In the first stage, the available genetic, neuroimaging, cognitive, and behavioral data, as obtained by different measures and procedures, are gathered and aggregated. These multidimensional data sets will need to be large enough to unearth potentially weak effects. This might create an obstacle given that such an endeavor goes beyond the efforts of individual researchers and labs. However, the difficulty is not insurmountable, as testified by large science projects such as the human connectome project and the ENIGMA consortium (see e.g., Bearden & Thompson, 2017), which provide large databases of genetic, neuroimaging, and clinical data and have been used in research on antisocial behavior (Rogers & De Brito, 2016).

In the second stage, the collected data would be combined and analyzed by powerful data-mining algorithms. This step has two phases. The first phase corresponds to exploration that includes finding new latent variables that might characterize data sets and accordingly enable researchers to stratify, or even discover, new clusters that reflect more homogeneous groupings of individuals according to different dimensions. Ideally, these dimensions should identify markers or latent markers that will enable prediction and have diagnostic utility. For example, patterns of activation in a specific brain region could be used to reliably delineate certain groups of individuals exhibiting antisocial behavior.

A study by Fanti, Konikou, Cohn, Popma, and Brazil (2020) exemplified this approach. They investigated group differences in amygdala activation within a sample of youths at risk for antisocial behavior. By using measures for anxiety, callous-unemotional traits, and childhood abuse, they distinguished between five groups of at-risk individuals. By using functional MRI, they then tested group differences in the patterns of amygdala activity of the participants while they performed an aversive-conditioning task. Note that the study suggests that a group characterized only by high callous-unemotional traits and a group characterized by high callous-unemotional traits, anxiety, and familial abuse differed significantly in their amygdala responses during the different phases of the aversive-conditioning task. This finding indicates that the subtypes of youths at risk for antisocial behavior can also be delineated by differential amygdala activity under specific learning conditions. Likewise, data from different behavioral tasks and associated biocognitive measures could be used to stratify and provide markers of different subgroups of antisocial individuals (Brazil et al., 2012; Zeier, Baskin-Sommers, Hiatt Racer, & Newman, 2012).

After novel categories and profiles have been established, the second phase includes validation of the markers by linking them to reliable behavioral observations. For instance, if a category includes a dimension representing disturbances in cognitive inhibition, then this dimension should correlate highly with behaviors that would follow from such an impairment (e.g., sensation-seeking and reactive aggression; Blair, 2008; Brazil, Maes, et al., 2013).

Here a significant amendment to the RDoC type of approach should be noted. Often the connection between the neural data and behaviors will need to be computationally modeled to provide explanations of how neural activity translates into cognitive processes underlying different forms of antisocial behaviors. The RDoC project has been rightly criticized for not explicitly including the computational level of description within its domains and dimensions (Wiecki, Poland, & Frank, 2015; see also Lilienfeld, 2014). Computational approaches to psychiatry promise to add additional constraints on psychiatric nosology by providing a conceptual bridge between neurobiological mechanisms and behavioral manifestations of different disorders (Mathys, 2016; Stephan & Mathys, 2014). We indicate several respects in which computational models have been useful in research on psychopathy.

Research suggests, for instance, that psychopathic individuals show aberrant associative learning when it is based on rewards and punishments (Blair, Mitchell, & Blair, 2005). Psychopathic individuals seem to learn less from punishments and thus are less prone to change their behavior in response to it. However, behavioral tasks and accompanying brain-imaging studies do not allow for direct quantification how different types of information (e.g., rewards and punishments) affect associative learning on a trial-by-trial basis. Using computational modeling of task performance to estimate latent cognitive operations and mapping these to the brain processes and other relevant variables, such as personality traits, self-reports, and so on, can help solve this problem (Brazil, Hunt, et al., 2013).

In addition, computational modeling enables researchers to parse the behavioral task and cognitive processes into their more basic components. This allows researchers to discover and investigate latent variables that might be responsible for producing the observed data from brain studies and during behavioral performance. In this regard, a recent study indicated that psychopathic features are related to less accurate estimates about the relationship between the imperative stimulus and punishment during a conditioning task (Brazil, Mathys, Popma, Hoppenbrouwers, & Cohn, 2017). More specifically, the latter study pointed out that inaccuracy (i.e., “uncertainty”) in different types of representations related to the perceived changeability of the stimulus-outcome contingencies during threat conditioning were differentially correlated with different psychopathic traits. These findings support the view that understanding psychopathy and antisocial behavior requires more fine-grained insights into the cognitive mechanisms underlying aberrant behaviors, which can potentially lead to more targeted therapies to remedy the impairments. In this way, the spirit of the RDoC project can be extended to include computational considerations when devising new classifications and measures for doing research (for further extensions of RDoC, see Bolton & Gillett, 2019).

The third stage of the proposed biocognitive approach to classification involves devising effective treatments that target specific deficits, impairments, and dysfunctions that were identified in previous steps and provide grounds for forming clinically more homogeneous groupings of antisocial individuals. So far, there are not many treatment studies that are based on this type of biocognitive profiles.

A study by Baskin-Sommers, Curtin, and Newman (2015) offers a successful example of an intervention that was developed according to the differences in cognitive characteristics of the target groups. They designed a cognitive remediation treatment to alleviate psychopathy-specific impairments on the basis of evidence that individuals scoring higher on psychopathy measures show deficits in modulating attention when multiple sources of information need to be integrated to successfully solve the task (see e.g., Koenigs & Newman, 2013). After following the treatment program that was specifically designed to enhance the modulation of attention, individuals with higher psychopathic traits showed improvements while solving tasks requiring modulation of attention. Crucially, a second group consisting of offenders who did not have impairments in modulation of attention did not benefit from this intervention. This indicated that the treatment works if it is applied to a group of people who are delineated according to a meaningful biocognitive measure.

It has to be noted that adding emphasis on biocognitive data in classifying different psychiatric disorders and conditions does not entail excessive medicalization (for discussion, see Bolton, 2013; Poland & Von Eckardt, 2013). In fact, these data might help to design therapies that involve everything from pharmacological to targeted psychotherapy (Insel & Cuthbert, 2015). As illustrated by the study by Baskin-Sommers, Curtin, and Newman (2015), a focus on cognitive impairments in people exhibiting severe forms of antisocial behavior can lead to devising and implementing targeted noninvasive cognitive remedial therapies.

It is evident that the biocognitive approach to antisocial behavior delineated so far has to meet significant empirical challenges concerning data collection and their analysis (Brazil et al., 2018). However, there might be conceptual difficulties that might discourage even embarking on this project. In the next section, we consider a conceptual or theoretical objection directed against the biocognitive approaches in mental health sciences.

An Objection to Classification Based on Biocognition

Although the vision encapsulated in proposals such as RDoC has boosted research into the biological basis of psychopathology, it has also spawned critique. Borsboom et al. (2019) recently offered a detailed criticism of the RDoC-type approaches to psychiatric research. Specifically, they argued that approaches that adopt the “mental disorder is a brain disorder” ² view are empirically and theoretically untenable (Borsboom et al., 2019, p. 3). Their criticism elaborates the general objection that projects such as RDoC are committed to a form of reductionism that could hinder or be detrimental to psychiatric practice (Parnas, 2014). Given that the biocognitive approach to categorizing and investigating antisocial behavior shares conceptual assumptions with RDoC, it might be susceptible to the same objections.

Borsboom and colleagues maintained that biologically oriented approaches to psychiatry involve explanatory reductionism. This is the view that mental disorders “are caused by ‘genes for mental disorders’, neurobiological mechanisms, deficient brain circuits, and other biological factors” and “the belief that mental disorders can ultimately be explained in terms of specific dysfunctional neurobiological conditions” (Borsboom et al., 2019, p. 1). Moreover, they maintained that explanatory reductionism “implies the possibility of constructing a biological definition of . . . mental disorders” (Borsboom et al., 2019, p. 3). The idea is that explanatory reductionism aims at identifying higher level phenomena related to mental disorders and their symptoms with their presumed underlying biological causes. For the explanatory reduction to be successful, they maintained that lower level phenomena must be independently identifiable from the higher level phenomena that it causes. The characterization of syphilis as a disease caused by the bacterium Treponema pallidum exemplifies a successful categorization on the basis of explanatory reductionism (see Borsboom et al., 2019, p. 1). According to this characterization, symptoms such as fever, weight loss, hair loss, headaches, and sores in genitals are higher order phenomena that characterize a development of syphilis, and T. pallidum is the lower level phenomena constituting the entity that causes the higher level symptoms. Note that these two levels can be independently identified, thereby constituting a successful reductive explanation.

Borsboom and colleagues (2019) argued that explanatory reductionism in psychiatry and in the classificatory approaches that presuppose it is empirically untenable. Concerning the empirical inadequacy of explanatory reductionism, they claimed that there does not seem to be many discoveries in psychiatry that can be reduced to unique biological causal factors, despite the wealth of research into the biological underpinnings of mental disorders (see also Adam, 2013). Studies have shown that mental disorders, as currently categorized, are multifactorial and involve causes that span from genetic and neurobiological to psychological and social levels (Kendler, 2012). Moreover, Borsboom and colleagues wrote that several theoretical/conceptual problems afflict the reductionist program.

To appreciate one of Borsboom et al.’s (2019) conceptual criticisms, we need to clarify first their claim that certain mental states that constitute symptoms of mental disorders (as defined in the DSM or ICD) are intentional. In accordance with a consolidated philosophical tradition, several types of mental states, such as beliefs and desires, have intentionality (from Latin tendere, meaning “directedness toward an object”; Brentano, 1874). In this context, intentionality refers to the property of being about something or of having a semantic content. The belief that it is raining outside, for example, is intentional in this technical sense of the word because it is about the fact that it is raining outside. This fact is said to be the semantic content, or content for short, of that belief. Specifically, the content of mental states is taken to discern the states from each other. For instance, the belief that today is Tuesday and the belief that today is Wednesday differ because their contents are different; they are about different states of affairs. Likewise, desires and some feelings are individuated by their content.

In case of the DSM–5, Borsboom et al. (2019) mentioned that the definition of alcohol use disorder refers to “craving,” which is defined as a desire or an urge about drinking alcohol. Likewise, symptoms of depression include feelings of worthlessness and excessive guilt, which are mental states with content. They maintained, furthermore, that symptoms with intentional content might give rise to other symptoms and therefore explain them because of rational relations (see Borsboom et al., 2019, pp. 5–6). For instance, psychodynamic explanations of psychopathy claim that lack of remorse and grandiose sense of self-worth might be a coping strategy that makes sense (i.e., is rational) from the perspective of trying to protect one’s self-image from feelings of shame and worthlessness caused by abusive parents and other adverse events that happened during childhood (see Prosser, Friston, Bakker, & Parr, 2018).

According to Borsboom et al. (2019, p. 6), the intentionality of symptoms and their rational relations create two sorts of difficulties for explanatory reductionism. One is that the rational relations between symptoms are not recognizable at the neurobiological level. Therefore, by endorsing explanatory reductionism and, thus, attempting to identify mental disorders only through neurobiological properties, we might miss important patterns of causal relations between intentional mental states.

Borsboom et al. (2019, p. 6) thought that a second challenge for explanatory reductionism derives from the multiple realizability of intentional mental states. Many philosophers of mind endorse the view that the same type of mental state can be realized by multiple physical states (for a review, see Bickle, 2019). The underlying assumption for this view is that the different types of mental states can be completely characterized independently from the biological mechanisms that underpin their occurrence. A classic example is the conceptualization of pain through its typical causal role. According to this characterization, pain is the internal state caused by certain nociceptive stimuli and can cause avoidance behaviors and other mental states, such as the desire to get rid of the pain. The causal role of pain is most certainly implemented in certain neuronal circuits, but the characteristics of these circuits may differ across individuals and species. For instance, if octopuses can have pain, then it is probably implemented in a different physical structure than in humans given that we have different brain structures.

Closer to the case at issue, Borsboom et al. (2019) appeared to argue that given that intentional mental states are identified by their content, they can be implemented or realized by different physical systems. For instance, a person can have a perception of a mouse, and a cat can perceive the same mouse. Thus, they can have a perceptual state with the same content, although their physical correlates might differ given the differences between human and cat brains. Moreover, given the plasticity of the brain, the same psychological functions and mental states might in principle be implemented by different brain structures in different people. On theoretical grounds, Borsboom and colleagues offered, as a likely possibility, that the contents of mental states can be multiply realizable, in the sense that “two people may both believe they are being spied on by the CIA, but this belief may be coded differently in their brains” (Borsboom et al., 2019, p. 7). The analogy they offered is that symptoms of mental disorders that are characterized by intentional content, such as having delusional beliefs or different types of phobias (e.g., fear of heights or closed spaces), “can be physically realized in different ways in different people, just as a photograph can be stored on a microfilm, as some digits in your computer, or on a piece of printed paper” (Borsboom et al., 2019, p. 7). Accordingly, they maintained that if symptoms of a mental disorder involve contentful mental states, it is possible that the disorder will be multiply realizable, that is, implemented in different physical states. The multiple realizability of symptoms and mental states is taken to raise a problem for explanatory reductionism. If mental states can be implemented in radically different biological structures, then it is unlikely that for each mental disorder and its associated symptoms there will be a biological structure that uniquely realizes or causes it. ³

The third argument is that against a further consequence of explanatory reductionism, the network of symptoms that characterizes different psychiatric conditions is relative to specific cultural and historical contexts. This makes it unlikely that we will always be able to capture them in scientifically neutral biological terms. Borsboom and colleagues (2019) clarified this point with the example of substance abuse disorder:

The probability that alcohol withdrawal symptoms in substance abuse will lead to legal problems (both diagnostic criteria in DSM–5) is clearly different for an American living now, as compared with one living at the time of the Prohibition; likewise, the consequences of public drunkenness are completely different in, say, Riyadh versus New Orleans. (pp. 7–8)

Borsboom et al. maintained that if explanatory reductionism is true, then all the different manifestations of substance abuse should have a homogeneous “biological essence” that cuts across different contexts and is “responsible for these different expressions” (Borsboom et al., 2019, p. 8). However, it is unlikely that mental disorders that are (partly) defined in terms of violations of social expectations (i.e., social norms), as is the case in ASPD and substance abuse, share a biological essence.

A final objection to RDoC-inspired approaches could be that given that social criteria individuate symptoms of a disorder, it is unlikely that we will be able to find separate (independent) criteria for individuating the biological correlates without a reference to the higher order social criteria. As we saw in the case of syphilis and its cause, T. pallidum, Borsboom and colleagues (2019) seemed to think that if this requirement is not satisfied, then we do not have a proper reductive explanation of a disorder.

Borsboom et al. (2019) used these objections to argue that RDoC-inspired approaches are reductionistic and are likely to turn out to be less powerful for advancing research in psychopathology. In the next section, we argue that the objections related to explanatory reductionism do not undermine empirical or theoretical credentials of the RDoC-type approaches to investigating antisocial behavior.

Biocognitive Classification Without Reductionism

Before offering our principal responses to Borsboom et al.’s (2019) theoretical objections, it should be remarked that their claim that current empirical research can be adduced as a ground for pessimism about prospects of biology and cognition-inspired classificatory approaches is misplaced. As we have seen, such projects are principally motivated by the dissatisfaction with syndrome-based categorizations of mental disorders based on scoring of behavioral symptoms, as implemented in different versions of DSM and ICD (Cuthbert & Insel, 2013; Poland & Von Eckardt, 2013; Tabb, 2015). The large body of research showing that subtypes of antisocial individuals engage in similar behaviors but present with different patterns of cognitive, genetic, and neurobiological features already indicates that it is unlikely that the impairments seen in these subgroups can be reduced to clearly delineated genetic and neurobiological mechanisms that are shared among the groups (Adam, 2013; Cuthbert & Insel, 2013; Insel & Cuthbert, 2015). Accordingly, the aim of biocognitive approaches, such as RDoC, is to revise our current classifications of mental disorders by including data types ranging from genetic and neurobiological to cognitive and behavioral factors rather than reducing them to unique neurobiological mechanisms (Cuthbert, 2014; Cuthbert & Insel, 2013; Wiecki et al., 2015). In this regard, biocognitive frameworks should be more properly regarded as revisionary approaches to psychiatric classification instead of reductive approaches.

The considerations above apply to research on antisocial behavior. The goal is not to reduce ASPD and psychopathy to their biological underpinnings; it is, rather, to devise and revise classifications that will reflect better classifications of individuals who exhibit severe, pervasive, and early onset of antisocial behavior and will be conducive to better therapies, intervention, and prevention programs (Fanti et al., 2020). Thus, the empirical failures under current syndrome-based classifications considered by Borsboom et al. (2019) do not offer sufficient grounds for abandoning approaches that seek to anchor psychiatric classifications in biological and cognitive systems. Instead, the failure of empirically aligning syndrome-based classification with biocognitive discoveries and theories is a strong reason to revise or abandon the current categories and to focus on the further development of bottom-up classifications (Tabb, 2019b).

Nonetheless, endeavors that incorporate biological and cognitive factors, and specifically the biocognitive approach to stratification of individuals exhibiting severe forms of antisocial behavior, put emphasis on biological and cognitive factors on which new classifications should be rebuilt. Thus, it has to be seen whether this type of approach implies a commitment to explanatory reductionism. Recall that according to Borsboom et al. (2019), explanatory reductionism involves the following assumptions: (a) Mental disorders can be explained by (unique) biological causes, and (b) mental disorders can be identified with brain disorders by a process of reduction. Recall that here reduction involves providing separate criteria for identifying the higher level phenomena (i.e., mental disorders and their symptoms) and their reductive base (i.e., their biological implementation).

Regarding the first assumption, it should not be overlooked that many biocognitive approaches to classification do not imply that a mental disorder is caused by unique biological causes. Instead, as we have seen, the emphasis is on recategorizing mental disorders using, among other factors, biological causes and correlates also but not at the expense of environmental and social factors (Brazil et al., 2018; Insel et al., 2010). The real issue with the explanatory reductionism that Borsboom et al. (2019) seemed to be pointing out in this context must be that it fails because we cannot identify the lower level biological underpinnings without relying on antecedent identification of higher level symptoms. This is what they wrote in this context:

If these realizing brain states cannot be characterized as neurobiological phenomena (i.e., one cannot say that John believes the CIA spies on him solely on the basis of knowledge about his biological constitution), they cannot be identified independently of the phenomenology. In this case, explanatory reductionism fails to provide theoretical concepts at the biological level that can be systematically identified with theoretical concepts at a higher-level, and the idea that mental disorders are brain disorders becomes void. (p. 7)

Thus, it seems that the intentional content or context-specific manifestation of different symptoms makes the idea of explanatory reductionism conceptually implausible. To be more specific, it is not that states or mechanisms in the brain cannot represent things in the world (i.e., have intentionality or content) or that they manifest differently across cultures and contexts. Rather, the fact that we cannot identify those brain states and mechanisms without knowing what they represent or without referring to the social criteria for their individuation is taken to undermine explanatory reductionism. However, if this is what explanatory reductionism entails, then the proper response to this worry should be that the RDoC-type approaches do not presuppose explanatory reductionism.

In general, biocognitive approaches are committed neither to identifying mental disorders just through biological criteria nor to the idea that higher level cognitive constructs will be discarded in explanations or procedures for diagnosing illnesses when their lower level correlates are identified. For instance, Cuthbert and Insel (2013) indicated that the first step in implementing RDoC to psychiatric research presupposes that neural functions and dysfunctions should be determined by examining what behavioral tasks the brain is supposed to perform given its evolutionary history. We can then start investigating their specific neural correlates after this has been achieved. Thus, the specification of neural systems conceptually depends on identifying the cognitive and behavioral tasks that the brain is supposed to perform. Of course, given the revisionary nature of RDoC-inspired projects, it is an open empirical possibility that future developments of categories and explanations in psychiatric and forensic sciences might reduce or even eliminate the higher level constructs in favor of neurobiological constructs (for discussion, see Murphy, 2017; Pernu, 2019). However, we do not take a stance on this empirical issue. We argue that there is no necessary conceptual connection between biocognitive approaches and the idea that neurobiological mechanisms should be identified independently from higher order cognitive processes or that higher order constructs will be eliminated in the process. More importantly, we argue that this is the case regarding the biocognitive approach to classifying antisocial behavior.

According to Brazil et al.’s (2018) proposal for stratifying and rebuilding categories of antisocial behavior, the first step is to delineate individuals who exhibit extreme forms of antisocial behavior defined by deviations from socially adopted standards. Because it is accepted that these standards vary cross-culturally, this implies that the categorization, biological substrates, and symptom severity might vary cross-culturally as well (Jurjako et al., 2019; Sadler, 2008). The next step is to collect and integrate within this behaviorally defined group of people genetic, physiological, cognitive, and brain data to form subtypes that will reflect more homogeneous clusters of people. But, here again, there is no presupposition that antisocial behavior will fully be explained in neurobiological terms because (a) we already started with a group that is behaviorally/normatively individuated and (b) we know that etiology of antisocial behavior is correlated with growing up in abusive and depriving social environments, which often systematically influences the developmental disturbances in the activity of different brain areas that correlate with pervasive forms of antisocial behavior (Burt, 2009; Connolly & Beaver, 2014; Gao, Raine, Chan, Venables, & Mednick, 2010; Glenn & Raine, 2014; Raine, 2018). Thus, these factors will also be weighted and integrated with the data delineating more homogeneous clusters (Fanti et al., 2020).

Moreover, Brazil et al. (2018) explicitly discussed computational/representational parameters and variables to measure latent cognitive/affective processes, such as attention, confidence, uncertainty in beliefs, and expectation-based reward learning. These computational models pertain to quantifying and disentangling different internal cognitive processes and play important intermediary roles between clinical and phenomenological manifestations of psychiatric illnesses and their neurobiological underpinnings (Mathys, 2016; Stephan & Mathys, 2014; Wiecki et al., 2015). Accordingly, these approaches are expected to further the integration of, rather than eliminating, behavioral processes and life experiences with variables pertaining to biocognitive information to devise better data-driven categories (Brazil et al., 2018; Insel & Cuthbert, 2015).

The considerations presented so far are enough to show that Borsboom et al.’s (2019) objections do not undermine the biocognitive approach to research of antisocial behavior and psychiatric classifications more generally. However, before closing our discussion, we would like to address another possible difficulty. In the criticism by Borsboom et al. might linger a further objection. It appears legitimate to conclude that one of their messages is that if RDoC-type approaches are not construed as committed to explanatory reductionism, then their insistence on biological or, as in the case of the approach at issue, biocognitive variables appears to be unmotivated.

Our response is that they and other opponents of biocognitive approaches to psychiatry fail to notice that the core motivation behind a framework such as RDoC is better understood as a methodological injunction to go beyond self-reports and behavioral data to integrate already available and new biocognitive data into the nosology. This interpretation also applies to the primary motivation for the original RDoC proposal. Insel and Cuthbert (2015) suggested that one major challenge for developing precision medicine in psychiatry is to find ways of dealing with the many imprecise concepts that have been dominating the research and practice. Among the problematic and imprecise concepts, they included terms such as behavioral health conditions, mental disorders, or mental health conditions because they give the impression that there is an autonomous mental substance to which mental problems pertain disconnected from its physical underpinnings. All parties to this debate seem to agree that mind and body are not really different substances and that biological factors have a role in psychiatric research (Bolton & Gillett, 2019; Borsboom et al., 2019). Nonetheless, there is a tendency to overemphasize observable behavior, subjective reports, and relations between symptoms of mental disorders that creates an artificial wedge between psychiatry and the rest of medicine (Insel & Cuthbert, 2015; White, Rickards, & Zeman, 2012). Such tendencies could be seen as impeding advances in clinical and scientific research in psychiatry and forensic sciences (Cuthbert & Insel, 2013; Wiecki et al., 2015).

Given this backdrop, RDoC-type approaches are motivated by an injunction that should contribute to avoiding arbitrary divisions between bodily medicine and mental health conditions. Such a reading recommends, in fact, reconceiving and not reducing psychopathology as dysfunctions of brain and bodily systems. Progress in diagnosing and treating illnesses is more likely to be achieved by recognizing that beside subjective reports, phenomenological experience, and observable behavior, we have to include other biocognitive data. Integrating such data will bring psychiatry closer to the rest of medicine, in which biotechnological advances already greatly enhanced diagnosing, predicting, and treating of illnesses (Cuthbert & Insel, 2013; White et al., 2012). Thus, instead of criticizing biocognitive approaches on the basis of their alleged reductionist pretensions, the critical evaluation should focus on their potential to deliver the promised practical benefits of better diagnosing, preventing, and treating health issues in psychiatric and forensic sciences.

Conclusion

The study and treatment of antisocial personality disorder and psychopathy is fraught with difficulties that to a significant extent stem from classifications relying on heterogeneous behavioral symptoms. We argued that the prospect of progressing toward more effective treatments depends on recategorizing people falling under these constructs through initiatives that focus on the integration of biological and cognitive information. Such approaches should enable the refinement of old and relatively stable diagnostic categories and the discovery of new categories relevant for psychiatric research and practice. These refined or new categories are intended to reflect more homogeneous groupings of individuals that are suitable for establishing tailored therapeutic interventions. Moreover, we indicated ways in which a proper biocognitive classification of antisocial behavior can go beyond the currently available constructs of RDoC by including the computational level of description to provide better constructs and classifications.

Given that biocognitive approaches put emphasis on cognitive and biological factors in psychiatric nosology, we addressed the question whether they are committed to an illegitimate form of explanatory reductionism. We argued that in its general conception and when applied to research on antisocial behavior, the biocognitive approach is not committed to explanatory reductionism. Moreover, we indicated how this type of approach already shows real promise in devising targeted therapies for ameliorating deficits correlated with different types of antisocial behavior.