Partner choice increases observed reciprocity-based cooperation but decreases unobserved stake-based cooperation.

According to current theory and experiments, cooperation is more likely to evolve when organisms can choose to replace uncooperative partners with cooperative ones. However, there is a downside to this partner choice: when partners can be easily replaced, organisms have less stake in their partners' welfare and will therefore be less likely to help keep those partners alive and well enough to reciprocate. Here, I present a mathematical model showing that when a third party is present, organisms will provide more observable help to their partners (reciprocity/signalling-based helping), but less anonymous help that would keep that partner in good condition (stake-based helping). The net effect of partner choice depends on the relative strength of these two factors: partner choice has a more positive effect if interactions are short (i.e. less stake), when observers judge based on observed helping (i.e. reputation matters), and when one can have multiple cooperative partners at the same time. These results show the importance of differentiating between helping that relies on observation (e.g. reciprocity and signalling), helping that requires no observation (e.g. kinship and stake), and how the two types interact.