Time to absorption in discounted reinforcement models. (Q2574614)

Reinforcement model is studied with the goal to establish the time to absorption when the discount \(x\) tends to zero. Consider population of at least four members where triples are formed to ``play together''. These triples change in time and it is proved by the authors [Math.\ Soc.\ Sci.\ 48, 315--327 (2004; Zbl 1091.91060)] that the process is trapped in a degenerate state such that the population is divided into subgroups of sizes 3-5 which members play within the subgroup only. The main results of the paper (Theorems 2.2 and 3.1) concern with a model where the history influence is discounted using a discount rate \(1-x\). In Theorem 2.2 it is proved that with a positive probability each player will play with each other beyond time \(\exp \{c_N x^{-1}\}\), \(N\) is the population size. Consider more specific one-dimensional case with state space \([0,1]\) and its compact subintervals \(I_x\) increasing to \((0,1)\) as \(x\) tends to zero. Then it is proved in Theorem 3.1 that the expectation of the first leave time (from \(I_x\)) of the process tends to \(\exp \{Cx^{-1}\}\) as \(x\) tends to zero.