Override versus full delegation: pricing impact

Determine whether allowing human participants to override non-binding algorithmic price recommendations (Recommendation treatment) results in higher market prices than fully delegating all pricing decisions to the same algorithm for the entire supergame (Outsourcing treatment) in the indefinitely repeated Bertrand duopoly experiment with two firms, perfectly inelastic demand from 60 consumers, integer price grid {0,1,2,3,4,5}, and continuation probability 0.95; ascertain which effect dominates between increased adoption due to override capability and potential price reductions induced by deviations from algorithmic recommendations.

Background

The study compares two modes of algorithm use: full delegation (Outsourcing) and advisory support with manual override (Recommendation). Behavioral literature suggests override capability can increase algorithm adoption, but strategic deviations from recommendations could reduce prices.

Before presenting results, the authors explicitly flag the comparative pricing effect as an open question, noting opposing forces—higher adoption under override versus potentially lower prices due to overriding collusive recommendations—and pose it as an exploratory research question.