From a developmental standpoint, the main virtue of David Pietraszewski's theory is its ability to deal effectively with the reduction problem. At its core is the claim that the gamut of multi-agent conflict can be decomposed into four types of triadic interactions, each specifying a distinct way by which third parties can be drawn into conflict. By constraining the hypothesis space to a finite repertoire of coalitional schemata, the theory offers a developmentally plausible way in which young learners may infer role assignments across multi-agent configurations. Once the appropriate stance is adopted, attributing group membership becomes an eminently tractable task (cf. Thomsen & Carey, Reference Thomsen, Carey, Banaji and Gelman2013).

Nothing, however, guarantees that third-party interactions end up being interpreted through coalitional lenses. Any given social behavior can be, in fact, apprehended under multiple and equally compatible frames. By way of example, let us consider the “generalization” event, in which A imposes costs on B and C. By Pietraszewski's account, this event should license the inference that B and C belong to the same group. This is the case, however, only if the observer is trying to map the observed interactions in terms of coalitional units. Other interpretations are certainly possible. For instance, the observer may take the across-patient consistency of the agent's action as evidence of an underlying trait (i.e., A is a bully; Boseovski & Lee, Reference Boseovski and Lee2006). By focusing on the attribution of individual dispositions, this stance inhibits utilizing B and C's common interaction role (as victims) as group-diagnostic information. Reasoning about traits is, of course, not the only contender. Triadic interactions may be revelatory of underlying social structures. Let us consider the “displacement” case: A imposes costs on B, who does the same on C. Under Pietraszewski's account, A and C should be represented as belonging to the same group. This interaction can, however, also be taken to indicate a linear dominance structure in which a high-ranking agent (A) attacks a subordinate (B), who redirects its aggression against a lower-ranking agent (C) to prevent further aggression from third parties (for evidence of this behavior in nonhuman animals, see Kazem & Aureli, Reference Kazem and Aureli2005). This interpretation, although similarly requiring the embedding of pairwise relations within a larger structural template, does not license the assumption that A and C are in the same group.

As these examples illustrates, there are multiple frames to choose among. This plurality is to be expected, as it reflects the variety of fitness-relevant affordances in the social world. If the interpretation of social interactions can be considered akin to the phenomenon of multistable perception, systems for arbitrating among these competing frames then must be expected, especially when the concurrent activation of multiple frames may lead to contradictory inferences (as in the “displacement” scenario, in which two agents would be represented at the same time as being part of the same group as well as placed at the opposite ends of a hierarchy). If Pietraszewski's computational theory manages to effectively imprint deterministic force to group-based reasoning in a specific coalitional frame, it offers little guidance with respect to these broader arbitration issues. This architectural blind spot is, however, orthogonal to the internal validity of Pietraszewski's theory: Once a suitable event description (in terms of groups) has been selected, membership attribution does indeed become a cognitively straightforward operation. Furthermore, this limitation is by no means unique to Pietraszewski's account. Several authors made the empirical case for each of these interpretive strategies in infancy and toddlerhood (trait-based, Hamlin, Reference Hamlin2013; relational, Thomsen & Carey, Reference Thomsen, Carey, Banaji and Gelman2013; group-based, Powell & Spelke, Reference Powell and Spelke2013; Rhodes, Hetherington, Brink, & Wellman, Reference Rhodes, Hetherington, Brink and Wellman2015), yet explicit discussions about how young learners may adjudicate among these have been lacking.

In fact, Pietraszewski appears aware of this issue. In discussing the “defense” scenario (in which agent C, having witnessed A attacking B, then retaliates against A), Pietraszewski concedes that observers may interpret A's behavior not as indicative of their common group membership, but rather of the agent's sensitivity to moral demands (e.g., punishing unprovoked aggressors, Geraci, Reference Geraci2021; Kanakogi et al., Reference Kanakogi, Inoue, Matsuda, Butler, Hiraki and Myowa-Yamakoshi2017). To determine which of these interpretations is being upheld, Pietraszewski proposes a way to assess the cross-situational consistency of role assignment: If A and C belong to the same group, they should also fill the appropriate group-constitutive roles in the remaining three scenarios; no such consistency should be expected if A's intervention was instead prompted by moral demands. Although falling short of solving the arbitration problem, this verification strategy suggests that different frames may yield representations of different granularity: A trait-based frame, for instance, may not encourage the encoding of the patients' identity (because not essential to predicting behavioral consistency) as much as a relational frame, or, conversely, it may be compatible with a wider range of pairwise combinations than a relational structure based on assumptions of linear ordering (Mascaro & Csibra, Reference Mascaro and Csibra2014).

In sum, although Pietraszewski's sterling contribution to social cognition may not shed new light on the still-neglected issue of how observers can decide among competing frames, it advances a computationally tractable and straightforwardly testable theory of how humans reason about social groups, which developmental scientists, in particular, will undoubtedly benefit from.