Language agents improve pricing and persuasion in economic games
How Far Can an Agent Push? Trust-Safety Margins in Live Language-Based Economic Games
Multiagent SystemsComputer Science and Game Theory
Summary
The researchers studied how computer agents negotiating or persuading people in live language-based games behave when they push boundaries. They found that adjusting prices using learning strategies made offers better accepted in negotiation. In persuasion, they showed that recommending every unit sold can increase sales, but only when the customer is willing to accept the recommendation. Their agent Milton ranked 11th out of 460 in bargaining by following strict rules. They also found some errors by comparing what the agents did to what the design intended.
What this means in practice
- •For online marketplace developers: Improve pricing strategies in automated negotiations by integrating learned pricing adjustments shown to increase offer acceptance rates.
- •For call center technology teams: Design persuasion scripts that recommend every unit only when customers are predisposed to follow advice, enhancing sales effectiveness in live chats.
Authors
Dima Galat, Marian-Andrei Rizoiu
Abstract
The previous version contained Unicode em dashes. Paste this ASCII-only version: Game theory identifies when a strategic action should be accepted, but not how long a live, language-based counterpart will continue to accept it. We entered the GLEE competition in its final week with one protected agent, Milton, and four instrument agents that ran deterministic, per-game randomized trials of policy changes; Milton received changes only through pre-specified gates or registered, rollback-guarded production rules. In negotiation, learned pricing improves adjusted percentile by 0.071 (n=21,168); in bargaining, a continuation model conditioned on the rejected offer did not improve on a recency-weighted incumbent and was retired. In persuasion, a preregistered five-dose experiment (7,391 assigned games, blinded until a fixed stop) shows that recommending every unit raises sales by 0.194 of the maximum (95% CI [0.147, 0.241]) where the receiver's slack makes obedience rational (52% of the Bayesian benchmark), lowers them where it does not (-0.067), and produces a larger early-to-late purchase decline than honest recommendations (-0.123). A rule fixed before unblinding now routes Milton's dose per cell. Milton stood 11th of 460 agents in bargaining at the fixed stop; we also report the close, where the final day's pool shift moved all five agents' bargaining ratings together while their play was unchanged, and three dispatch defects were found by checking delivered behavior against the design.