Revisiting the competency trap

Gaël Le Mens,Jerker Denrell

doi:10.1093/icc/dtz072

Gaël Le Mens, Jerker Denrell

Open Access

https://doi.org/10.1093/icc/dtz072

Copy DOI

Abstract

Abstract We revisit the competency trap and reexamine when it occurs. We show that a bias against alternatives that improve with practice does not require that learning is myopic in the sense of lacking foresight or failing to explore. The same bias occurs even if learners engage in substantial exploration and have foresight. In fact, we demonstrate that even a rational and foresighted learner, who follow an optimal strategy for balancing exploration and exploitation, will learn to prefer alternatives with initially high payoffs that decrease with practice over alternatives, with identical expected values, that have initially low payoffs that increase with practice. Our results show that a bias against alternatives that improve with practice is due to an asymmetry in error correction rather than to myopic learning. The implication is that a wide range of selection systems, even optimally designed ones, will be biased against late-bloomers.

Highlights

Jim was always interested in illustrations of how sensible learning processes could lead to suboptimal outcomes
When the value of the known alternative is À1, the learner chooses the initially increasing type 97.9% of the time in period three and chooses the initially decreasing type 99.2% of the time. These results illustrate how even an optimal policy leads to a competency trap in the sense that more competent actors, who have a known alternative with a higher payoff, will be more biased against a new alternative with initially increasing payoff
Our results show that the bias is not is not a consequence of the “myopic” character of learning, if myopia is defined as lack of awareness of the possibility that payoffs may be increasing or if myopia is defined as a greedy policy which does not explore

Summary

Introduction

Jim was always interested in illustrations of how sensible learning processes could lead to suboptimal outcomes. Even a rational learner will behave like the individual in the parable of the competency trap and will favor the alternative with high initial performance compared to the alternative with high long-run but poor short-run performance. This modeling exercise shows that the origin of the bias is not necessarily the assumption that learning is myopic in the sense that it ignores the long-run.

Myopic learning policies are not necessary for bias emergence

Even a farsighted rational learner may be biased

Optimal policy P3

A rational learner can favor an initially increasing alternative

Binomial payoffs

Caveat

Findings

10. Conclusion