Learning to be imperfect: The ultimatum game

Social and individual learning of helping in humans and other species

“…Individual learning is predicted to be very slow in games that are characterized by multiple Nash equilibria [15]. The ultimatum game has an infinite number of such equilibria, only one of which-contributing nothing-is stable in the ultra long run [84]. The social norm triggered initially is thus expected to dominate for a very long time before the learning dynamic finds the longrun equilibrium of the game [84].…”

“…The ultimatum game has an infinite number of such equilibria, only one of which-contributing nothing-is stable in the ultra long run [84]. The social norm triggered initially is thus expected to dominate for a very long time before the learning dynamic finds the longrun equilibrium of the game [84]. These two different proximate explanations for a fair contribution in experimental games illustrates again that the underpinning of behavior maps used by subjects in experiments remains unclear and is much debated, not to mention the effect of the internal states of individuals such as serotonin level [85].…”

Equity dynamics in bargaining without information exchange

“…If demands are jointly feasible, each player receives his demand; otherwise, 9 See also Gale et al (1995), Nowak et al (2000), Konrad & Morath (2015) for evolutionary models of "ultimatum bargaining" (Güth et al 1982), or Binmore et al (1998) for an evolutionary analysis of alternating-offer "Rubinstein bargaining" (Rubinstein 1982).…”

A survey on the bandit problem with switching costs

“…The discrete-time bandit problem is can be formulated as follows, following Frostig and Weiss (1999). 10 The state of the bandit system 7 For the literature on ultimatum games, see Gale et al (1995) and Thaler (1988). 8 Theoretically, the optimal solution for player is not the Gittins index policy since offers are not independent, that is, the probability of acceptance of the offer is weakly increasing in the size of the offer.…”