I disagree that a thing has to be decidable for there to be a relevant algorithm; an algorithm is a defined set of steps to take for a given input. And there is always the option of modelling social interaction as a decidable system then come up with an algorithm for the model (which is how most algorithms are used in practice, any decidable algorithm performs strangely in a physical world when it has to cope with cosmic rays hitting the CPU).
For people to be able to interact socially they have to make predictions about the consequences of their actions. That can be formulated as some sort of stochastic model. This is enough of a hook that algorithms can be used.
Anyway, that is fairly academic because in practice humans are obviously very predictable and tend to have quite stable personalities over time. There is a wealth of psychological literature classifying major personality traits [eg, 0] and most people respond in a normal way to incentives and social hierarchy with some adjustments for their cultural background.
For people to be able to interact socially they have to make predictions about the consequences of their actions. That can be formulated as some sort of stochastic model. This is enough of a hook that algorithms can be used.
Anyway, that is fairly academic because in practice humans are obviously very predictable and tend to have quite stable personalities over time. There is a wealth of psychological literature classifying major personality traits [eg, 0] and most people respond in a normal way to incentives and social hierarchy with some adjustments for their cultural background.
[0] https://en.wikipedia.org/wiki/Category:Personality_traits