Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
This paper studies the analytical properties of the reinforcement learning model proposed in Erev and Roth (1998), also termed cumulative reinforcement learning in Laslier et al. (2001).