1: \begin{abstract}

2: We provide a general framework to study stochastic sequences related

3: to an array of models in different literatures, including models of individual learning in economics, learning automata in computer sciences,

4: social learning in marketing, and many others. In this setup, we study

5: the asymptotic properties of a class of stochastic sequences that

6: take values in $[0,1]$ and satisfy a property that we call ``bounded

7: expected relative hazard rates.'' We provide sufficient conditions

8: for related sequences, which, compared to the original sequence, either

9: move slowly or slow down over time, that yield convergence to one with

10: high probability or almost surely, respectively. We apply these results

11: to show the optimality of the learning models in \citet{Borgersetal2004},

12: \citet{ErevRoth1998}, and \citet{Schlag1998}. 

13: \end{abstract}

14: