Saari et al.’s recent paper explores the relationship between the items within the domain of “impulse dyscontrol” of the Mild Behavioral Impairment Checklist (MBI-C) in predementia subjects with cognitive complaints using network analysis (Saari et al., Reference Saari, Smith and Ismail2022). Impulsiveness, or impulse control problems, is a common phenomenon among neurodegenerative conditions, such as Alzheimer’s disease (Rochat et al., Reference Rochat, Delbeuck, Billieux, d’Acremont, Van der Linden and Van der Linden2008), Parkinson’s disease (Imperiale et al., Reference Imperiale2018), and fronto-temporal dementia (Miller et al., Reference Miller, Darby, Benson, Cummings and Miller1997). Impulse control dysfunction can present in the preclinical stage of these neurodegenerative diseases and has been identified as a potential biomarker of the early stages of these diseases. Saari et al.’s work appears to have been motivated by the gap in our understanding of the relationship between the symptoms collectively situated under the umbrella term “impulse dyscontrol” of the MBI-C. This knowledge gap, as Saari et al. rightly point out, hinders the search for specific neural correlates of impulsiveness in persons at risk of developing dementia. Saari et al.’s use of network analysis in this context is novel; that is, to examine the clustering of items within the domain of impulse dyscontrol to identify relationships within a pool of potentially highly heterogenous symptoms. Their results indeed suggest that the impulse dyscontrol domain is heterogeneous. This is demonstrated by the fact that most of the items in the MBI-C are not closely connected, meaning that the presence of one of these items does not generally correlate with the presence of other items. The exception to this is the “stubborn or rigid,” “frustrated or impatient,” “argumentative,” and “agitation/aggressive” symptoms-cluster (hereafter “the dominant symptoms-cluster”), which consists of the most closely connected, as well as the most frequently present, items among the subjects. This suggests that the majority of individuals who score highly in this domain will present with the items from the dominant symptoms-cluster.

Two issues of concern arise from these observations. The first is in relation to the construct of the MBI-C impulse dyscontrol domain. Even though the domain is labeled as describing “the ability to delay gratification and control behavior, impulses, oral intake and/or changes in reward,” it is not immediately apparent that the characteristics of the dominant symptoms-cluster involve any delayed gratification or reward-driven choices of action. This means that the majority of individuals who score highly in this domain will not demonstrate behavior that the domain represents. Rather, rarer items such as sexual disinhibition and hoarding are more related to abnormal reward-driven behavior, even though these items are only loosely connected with each other, meaning that an individual who presents with one item is not likely to present with another, thus rendering the impact of each of these items on the total score of the domain relatively low. The second issue of concern is in relation to the overall heterogeneity of the items on the MBI-C. Given this heterogeneity, a question arises as to whether the items of this domain should all be included in one single domain, especially for the purposes of neural substrate mapping.

While Saari et al.’s analysis provides us with helpful information about the items of the MBI-C, what it cannot do is delineate how the items within the dominant symptoms-cluster relate to each other. Is there a common higher order construct that explains the clustering of these items? Or are the items interdependent in some way? Given the significance of the dominant symptoms-cluster in identifying “impulse dyscontrol,” the answer to this question is important in facilitating our understanding of the core processes underpinning “impulse dyscontrol” in predementia patients. The authors suggest that the cluster represents an “antagonistic antisocial construct,” but this proposition is disputable. While antagonism is undoubtedly a feature of these items, the concept of an “antagonistic antisocial construct” comes from the Hierarchical Taxonomy of Psychopathology consortium model (Krueger et al., Reference Krueger2018), which is associated with an “intention to hurt other people” (Krueger et al., Reference Krueger2018). There is, however, no evidence that any intention of hurting others is present in people with the dominant symptoms-cluster. The association with the “antisocial” label of this construct also carries a negative and potentially stigmatizing tone that is both not well justified and unable to provide meaningful insight into how and why people become antagonistic in the process of early neurodegeneration. Having said that, however, the presence of a social element among the behavioral items of the cluster is an important observation and should not be overlooked.

As an alternative to Saari et al.’s proposition, I propose that the emergence of the dominant symptoms-cluster can alternatively be explained by the loss of cognitive flexibility. Cognitive flexibility, the ability to adaptively shift mental processes and behavior according to the situation, is an important part of the executive function (Dajani and Uddin, Reference Dajani and Uddin2015). Gross impairment in cognitive flexibility can be seen in patients with dementia, as part of the dysexecutive syndrome (Godefroy et al., Reference Godefroy2010). It is reasonable, however, to hypothesize that more subtle forms of cognitive inflexibility might be present in a predementia state. The “rigidity/stubbornness” item, which is one of the central nodes of the dominant symptoms-cluster, could be a relatively direct behavioral manifestation of cognitive inflexibility. Cognitive inflexibility is linked to social cognition impairment (Grattan and Eslinger, Reference Grattan and Eslinger1989), which in turn can result in negative social interactions and behaviors. Cognitively inflexible individuals are also prone to cognitive overload due to reduced effectiveness in problem solving, which can in turn impair social judgment and behavior.

The distinction between an antagonistic antisocial construct and cognitive inflexibility as the higher order construct explaining the dominant symptoms-cluster is important for the purposes of understanding the dynamics of the progression of early-stage neurodegeneration, biomarker discovery, neurobiological correlations, and the design of early interventions. This alternative proposal can be tested with the help of psychometric measuring tools that measure cognitive flexibility, such as the Wisconsin Card sorting test (Miles et al., Reference Miles, Howlett, Berryman, Nedeljkovic, Moseley and Phillipou2021). Conventional psychometric tests may not, however, be able to account for the element of social interaction, which is an important factor involved in the dominant symptoms-cluster.

A computational psychiatry approach, which emphasizes the use of quantitative algorithmic models to understand and simulate psychophysical processes underlying observable phenomenology, is a novel tool that can be applicable in this context. In-depth introductions of this approach can be found in previous reviews by Adams et al. (Reference Adams, Huys and Roiser2016); Huys et al. (Reference Huys, Maia and Frank2016) and others. There are several key advantages of using a computational approach in this context. For example, in the process of explicitly modeling a phenomenon, one can take into account the interdependence of factors, which would not be limited to biological factors. In addition, explicitly defined parameters in the model would allow for quantitative model comparison, simulated data generation, hypothesis testing, and neural substrate mapping. The computational approach has achieved some success in enhancing our quantitative understanding of the core features of schizophrenia, such as the formation of psychotic symptoms (Sterzer et al., Reference Sterzer2018) and social dysfunction (Simonsen et al., Reference Simonsen2021). Recent efforts have sought to introduce the computational approach into the clinical prognosis prediction of depression (Berwian et al., Reference Berwian2020). Determining precisely how one might model the interaction of cognitive inflexibility and antagonistic behavior in a neurodegenerative context will require further exploration and is beyond the scope of this commentary.

In summary, behavioral changes are important early markers of neurodegenerative disorders, though how they should be constructed and quantified is still controversial. Grouping domains into symptoms-cluster based on clinical observation alone may not be neuroscientifically valid and may result in heterogeneity that can affect the understanding of the neurodegenerative process, as well as the search for neural correlates. Identification of the dominant symptoms-cluster is a step forward in refining the construct of “impulse dyscontrol.” The dynamic relationship between the items within this cluster, however, still needs to be clarified. This commentary proposes cognitive inflexibility as an alternative candidate to antisocial antagonism as the underlying construct of this symptoms-cluster. The use of a computational approach should be explored for the purposes of disentangling the complex dynamics of hidden cognitive constructs and observable social behavior.