The above exchange from the French crime series Spiral (Engrenages) speaks to the fact that from the perspective of “users” the definition of groups is far from straightforward. For Gilou and Tintin – long-time members of their small but maverick police unit – the team is very real. It binds them together and is a basis for them to not just to cooperate, but also to make significant sacrifices on each other's behalf. Indeed, the argument we see here relates to the fact that Gilou and Tintin have just lied to cover up the misdemeanours of another team-member (Laure). But this is something that JP – a newcomer to the team – was not prepared to do. And as he intimates, the fundamental reason for this is that the team is not real for him.

Trying to construct a model (computational or otherwise) of what is going on here is an important question – for all the reasons that Pietraszewski sets out in his carefully argued and well-researched paper. Indeed, for social and computational scientists alike, this is an immensely important challenge. Yet, although there is much to commend in Pietraszewski's attempt to do this, his proposed solution suffers from four core problems.

The first, and most basic problem with Pietraszewski's model is that treats the group as objectively definable based on a set of external exigencies (e.g., relating to conflict). Applied to Gilou and Tintin's team this would suggest that we could develop a computational theory which implied that so long as the team satisfied a set of predefined criteria it would be a team. Yet we see from the exchange in the Paris police station that no specification can do this without taking stock of the psychological reality of group members themselves – a reality that for Gilou and JP is structured by factors such as status (e.g., as a newcomer or old-timer), accessibility (e.g., prior experience of the group as a platform for collective action), and fit (e.g., a context of threat which makes it a meaningful entity) (Oakes, Haslam, & Turner, Reference Oakes, Haslam and Turner1994).

Following from these observations, the second problem with Pietraszewski's model is that it fails to take stock of the psychological foundations of group life. Before we can model intergroup behaviour, we need first to model the processes that make group behaviour possible (Turner, Reference Turner and Tajfel1982). After all, as the scene from Spiral shows, we can only engage in conflict with “them” if we have first developed a sense of “us” that makes cooperation among ourselves possible (Coser, Reference Coser1956). Accordingly, models that seek to capture the capabilities and affordances of groups (e.g., as a basis for trust, communication, mutual influence, and organisation; Haslam, Reference Haslam2004) need to start by understanding the foundations of the internalised sense of group membership that gives rise to a sense of social identity (Turner, Oakes, Haslam, & McGarty, Reference Turner, Oakes, Haslam and McGarty1994).

Third, a satisfactory model requires an appreciation of the internal structure of groups and of the norms that regulate their behaviour. Among other things, this is because norms determine group members' understanding of acceptable behaviour and dictate the different ways in which they should relate to each other (e.g., as outlined by Fiske, Reference Fiske1992). For example, what may look like conflict to observers may be perceived as cooperation by its members (and vice versa) and such considerations are essential for understanding whether particular behaviour presages group formation or else disintegration. As Gilou and Tintin note, although a commitment to honesty and openness will make many groups stronger, it may destroy others. Again, this is something we need to appreciate not as onlookers who stand outside the group, but as insiders for whom the group is psychologically real in ways that make it a platform for both sense-making and social action.

Developing and integrating these various observations, the fourth problem with Pietraszewski's model is that it fails to recognise that group definition and associated group processes necessarily depend on social context. Psychological group membership is fluid not fixed, labile not ossified. This means that “who we are” and what it means to be member of a given group varies across situations – for example, as a function of the groups against which we compare ourselves, the specific tasks and challenges we confront, and the way we are led. A computational model of social groups, therefore, needs to include an appreciation of social context and of its capacity to structure psychological group membership. Without this, the model's predictive power – and hence its efficacy – will be fatally compromised.

Notwithstanding these problems, we see the challenge that Pietraszewski sets out as a worthy one. And certainly the fact that, psychologically, the group is a subjective context-dependent entity rather than a static objective object, does not mean that it cannot – and should not – be modelled computationally. Indeed, the four problems we have highlighted constitute what we see as primitives for an alternative framework that might do precisely this. Critically, though, they suggest that the key challenge for those who would develop computational models of the group is to embrace an understanding of groups that is faithful to the contextuality of group members' subjective experience. In short, we need a model of “us” not “them” that speaks to the richness of social identity (i.e., a sense of “us-ness”) as a platform for the potentialities of group life.