Social relationships shape how we behave toward others – whether to lead or follow, when to offer assistance, and from whom we can expect support and to what extent. Thomas does an excellent job tying together a wide range of empirical findings into a unified framework. And we agree with the central premise: the human mind, even from infancy, must contain relationship representations and procedures operating upon these representations in order to navigate the social world effectively. The strength of the model is that it is explicit, and like any good model, is bound to be wrong in some respects; this is not a shortcoming, but entirely the point. Here, we take aim at one possible shortcoming, not to criticize, but to use the model as a scaffolding for pushing further inquiry, and preview some work we’re currently doing in our lab that may speak to some larger issues.
Our constructive gripe runs as follows: Broadly, we wonder if the three categories of relationships presented are in fact distinct. In the process, we worry that fuzziness of the boundaries between categories may undermine some of the functional clarity that the model seeks to provide. Specifically, we worry that the lines denoting interactions between agents hide considerable ambiguity that warrants sussing out, to the point that at least one of the three categories may be an instance of a class of outcomes involving the other two.
In particular, equality matching seems conceptually different from the other two. Rather than describing a type of relationship, equality matching underpins all successful social exchanges. The ability to detect violations of fairness and punish free riders is a foundational adaptation that undergirds the evolution of “altruism” (direct and indirect reciprocity). In this sense, equality matching may not be a distinct relationship model but rather, may have the same status as what Thomas calls “strength”: adjectival modifications of the “types,” not the types themselves. That is, a hot pie and a cold ice cream describe only two foods, not four: hot and cold are adjectival modifiers, and so, we suggest, is equality matching. More literally, the human mind contains adaptations that are concerned with tracking the degree to which all relationships involve equality versus inequality, although not necessarily in kind. Equality matching is not a relationship type, but a particular value along this adjectival continuum.
Thomas’ evidence supports this: in the absence of any relationship cues, infants default to expecting resources to be distributed equally. Their sensitivity to inequality (e.g., noticing when a character “steals” from another) reflects an expectation for fairness, not necessarily an assignment of a relationship “type” (see Tooby & Cosmides, Reference Tooby and Cosmides1996). Within non-kin relationships, roughly-speaking, humans’ cognitive architecture should be sensitive to cues denoting some sort of reciprocation. Indeed, humans experience anger when they feel they are not being properly valued in a relationship and modulate the costs they incur for someone based on the costs the other incurs for them (Sell et al., Reference Sell, Sznycer, Al-Shawaf, Lim, Krauss, Feldman and Tooby2017). Even within authority rank relationships, the expectation of reciprocity persists. Infants in Thomas’ cited studies selectively deferred to “legitimate” leaders rather than bullies, suggesting that following a leader is contingent on some perception of shared or aligned interests. And Price & Van Vugt, Reference Price and Van Vugt2015) notion of “service for prestige” is predicated on a reciprocal relationship between leaders and followers: meaning that a type authority ranking is caused by a kind of equality matching (just one not in kind).
Even leaving equality matching aside, the remaining categories are still not quite distinguishable. Communal sharing is defined by acting as one, having shared fate, and expecting care without direct reciprocity, but all of these features can also apply to hierarchical relationships that are cooperative rather than coercive. Leader–follower dynamics, when not based on dominance, still involve mutual obligation and shared goals. In coordinated group efforts, individuals take on asymmetric roles that collectively function as a single entity. From a macro-level view, they “act as one,” but the micro-level ties that support this unity are still built on reciprocal benefit – even if that reciprocity is uneven and role-dependent. Soldiers follow a commander not just out of deference, but because the leader represents the group’s interests, and the costs of obedience are incurred on behalf of a shared cause.
In sum, and with due deference to Fiske’s model upon which Thomas builds, there is considerable trouble here, which suggests that while the model may hold adequate water for the descriptive purpose of organizing research findings (of which there is great value), the model does not quite decompose relationship types into their cognitive primitives; we should not confuse it for a computational theory of relationship types, but rather a framework to push on (as we are doing here) to get there.
On a final note, one key insight that we are now just beginning to appreciate in recent work in our lab is a point alluded to by Thomas – namely, that infants use third-party interactions to infer relationships. While this a well-worn observation outside of infancy, the function of using third-party relationships, we contend, has been underestimated and undervalued – even by ourselves.
While further refinement and empirical validation are ongoing, our results from a number of agent-based modeling simulations suggest that reasoning about third-party relationships may be necessary for the emergence of coordinated, stable social structures. In these simulations, agents form ties by scoring others based in part on how they relate in directed pursuits to other agents (four strategies based on the primitives introduced in Pietraszewski, (Reference Pietraszewski2022); see Figure 1). When agents use any of these triadic evaluation strategies (including any and all permutations of them), groups naturally emerge: densely connected cliques where every member is tied to every other – while direct reciprocity alone fails to produce group structures. It is no accident that young infants not only track their own relationship status with others but also track the relationship between others (third-parties), because without the ability to track others’ relationships, our ability to pursue our own relationships falters, particularly if we want those relationships to be stable.
Figure 1.
Adapted from conflict interactions, above are the four ways in which groups might be transitively inferred through dyadic pursuits, with group-constitutive roles depicted in red. From left to right, A’s pursuit of B is followed by: A also pursuing C; C pursuing B; B pursuing C; C pursuing A.
You have
Access
Social relationships shape how we behave toward others – whether to lead or follow, when to offer assistance, and from whom we can expect support and to what extent. Thomas does an excellent job tying together a wide range of empirical findings into a unified framework. And we agree with the central premise: the human mind, even from infancy, must contain relationship representations and procedures operating upon these representations in order to navigate the social world effectively. The strength of the model is that it is explicit, and like any good model, is bound to be wrong in some respects; this is not a shortcoming, but entirely the point. Here, we take aim at one possible shortcoming, not to criticize, but to use the model as a scaffolding for pushing further inquiry, and preview some work we’re currently doing in our lab that may speak to some larger issues.
Our constructive gripe runs as follows: Broadly, we wonder if the three categories of relationships presented are in fact distinct. In the process, we worry that fuzziness of the boundaries between categories may undermine some of the functional clarity that the model seeks to provide. Specifically, we worry that the lines denoting interactions between agents hide considerable ambiguity that warrants sussing out, to the point that at least one of the three categories may be an instance of a class of outcomes involving the other two.
In particular, equality matching seems conceptually different from the other two. Rather than describing a type of relationship, equality matching underpins all successful social exchanges. The ability to detect violations of fairness and punish free riders is a foundational adaptation that undergirds the evolution of “altruism” (direct and indirect reciprocity). In this sense, equality matching may not be a distinct relationship model but rather, may have the same status as what Thomas calls “strength”: adjectival modifications of the “types,” not the types themselves. That is, a hot pie and a cold ice cream describe only two foods, not four: hot and cold are adjectival modifiers, and so, we suggest, is equality matching. More literally, the human mind contains adaptations that are concerned with tracking the degree to which all relationships involve equality versus inequality, although not necessarily in kind. Equality matching is not a relationship type, but a particular value along this adjectival continuum.
Thomas’ evidence supports this: in the absence of any relationship cues, infants default to expecting resources to be distributed equally. Their sensitivity to inequality (e.g., noticing when a character “steals” from another) reflects an expectation for fairness, not necessarily an assignment of a relationship “type” (see Tooby & Cosmides, Reference Tooby and Cosmides1996). Within non-kin relationships, roughly-speaking, humans’ cognitive architecture should be sensitive to cues denoting some sort of reciprocation. Indeed, humans experience anger when they feel they are not being properly valued in a relationship and modulate the costs they incur for someone based on the costs the other incurs for them (Sell et al., Reference Sell, Sznycer, Al-Shawaf, Lim, Krauss, Feldman and Tooby2017). Even within authority rank relationships, the expectation of reciprocity persists. Infants in Thomas’ cited studies selectively deferred to “legitimate” leaders rather than bullies, suggesting that following a leader is contingent on some perception of shared or aligned interests. And Price & Van Vugt, Reference Price and Van Vugt2015) notion of “service for prestige” is predicated on a reciprocal relationship between leaders and followers: meaning that a type authority ranking is caused by a kind of equality matching (just one not in kind).
Even leaving equality matching aside, the remaining categories are still not quite distinguishable. Communal sharing is defined by acting as one, having shared fate, and expecting care without direct reciprocity, but all of these features can also apply to hierarchical relationships that are cooperative rather than coercive. Leader–follower dynamics, when not based on dominance, still involve mutual obligation and shared goals. In coordinated group efforts, individuals take on asymmetric roles that collectively function as a single entity. From a macro-level view, they “act as one,” but the micro-level ties that support this unity are still built on reciprocal benefit – even if that reciprocity is uneven and role-dependent. Soldiers follow a commander not just out of deference, but because the leader represents the group’s interests, and the costs of obedience are incurred on behalf of a shared cause.
In sum, and with due deference to Fiske’s model upon which Thomas builds, there is considerable trouble here, which suggests that while the model may hold adequate water for the descriptive purpose of organizing research findings (of which there is great value), the model does not quite decompose relationship types into their cognitive primitives; we should not confuse it for a computational theory of relationship types, but rather a framework to push on (as we are doing here) to get there.
On a final note, one key insight that we are now just beginning to appreciate in recent work in our lab is a point alluded to by Thomas – namely, that infants use third-party interactions to infer relationships. While this a well-worn observation outside of infancy, the function of using third-party relationships, we contend, has been underestimated and undervalued – even by ourselves.
While further refinement and empirical validation are ongoing, our results from a number of agent-based modeling simulations suggest that reasoning about third-party relationships may be necessary for the emergence of coordinated, stable social structures. In these simulations, agents form ties by scoring others based in part on how they relate in directed pursuits to other agents (four strategies based on the primitives introduced in Pietraszewski, (Reference Pietraszewski2022); see Figure 1). When agents use any of these triadic evaluation strategies (including any and all permutations of them), groups naturally emerge: densely connected cliques where every member is tied to every other – while direct reciprocity alone fails to produce group structures. It is no accident that young infants not only track their own relationship status with others but also track the relationship between others (third-parties), because without the ability to track others’ relationships, our ability to pursue our own relationships falters, particularly if we want those relationships to be stable.
Adapted from conflict interactions, above are the four ways in which groups might be transitively inferred through dyadic pursuits, with group-constitutive roles depicted in red. From left to right, A’s pursuit of B is followed by: A also pursuing C; C pursuing B; B pursuing C; C pursuing A.
Acknowledgements
N/A.
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interests
The authors declare none.