Strategic Experimentation with Poisson Bandits
By Sven Rady, Godfrey Keller. 2009
174 citations recorded by the source corpus.
This paper studies a game of strategic experimentation with two-armed bandits where the risky arm distributes lump-sum payoffs according to a Poisson process. The intensity of this process is either high or low, and unknown to the players. We consider Markov perfect equilibria with beliefs as the state variable. There is no equilibrium where all players use cut-off strategies, and all equilibria exhibit an `encouragement effect' relative to the single-agent optimum. The fact that a success on the risky arm is not fully revealing limits the payoff asymmetry as well as the inefficiency of asymmetric equilibria. We construct the unique symmetric MPE and, for a sufficiently high difference between the two possible intensities, an asymmetric equilibrium that dominates the symmetric one in terms of aggregate payoffs.
Subfields: Management Science and Operations Research, Marketing, Safety Research
View Paper: https://www.semanticscholar.org/paper/90b4a84ac3c512b885c4ae38df0957c6bf2e1614
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Management Science and Operations Research
In Marketing
In Safety Research
Citations in the corpus
- Contests for Experimentation (2017, background): “Whether the innovation is feasible depends on the state of nature, v ∈ {G, B }, where G represents “good” and B represents “bad.””
- Negatively Correlated Bandits (2011, unknown): “It remains to be investigated whether, and, if so, to what extent, our results will carry over to settings of Poisson bandits as in Keller and Rady (2007) where a bad risky arm also has a small probability of yielding a positive payoff, thus introducing the possibility of “bad news events”.”