Research Talk by Arnaud De Bruyn, ESSEC Business School (FR)

04/12/2023

Bots Bargaining with Humans: Building AI Super-Bar- gainers with Algorithmic Anthropomorphization

In our recent Research Seminar Series, we had the privilege of hosting Arnaud De Bruyn from ESSEC Business School, who presented his latest research focusing on the rare domain of algorithmic anthropomorphization, showing signs of being an important consideration in AI development.

Arnaud explores the growing trend of AI-driven automation in bargaining processes by companies and its potential psychological impact on consumers. His study introduces the concept of "algorithmic anthropomorphization" as a novel approach to address algorithm aversion in coopetition contexts like bargaining games. In this approach, a bargaining AI is trained within a Generative Adversarial Network (GAN) framework to achieve superior economic outcomes while maintaining a "human" appearance in its interactions.

This algorithmically anthropomorphized AI is compared with two alternative bot specifications: a primitive bot mimicking human behavior and a purely economically efficient bot. The results reveal that despite making a bot appear more human-like, superficial anthropomorphization does not significantly improve subjective evaluations. On the other hand, algorithmic anthropomorphization shows promise as a solution, although imperfect. However, even when bots exhibit behavior indistinguishable from humans, they encounter the "uncanny valley" phenomenon, leading to lower subjective evaluations, regardless of their economic performance.

Considering that subjective evaluations of negotiations can predict the outcomes of future negotiations, the study discusses the potential negative impact of AI bargaining algorithms on long-term customer relationships. The findings underscore the need for a careful balance between economic efficiency and human-like interaction to ensure positive consumer perceptions in AI-driven bargaining scenarios.

We thank Arnaud for sharing such insightful results and sparking fruitful discussions on the future of AI.

Back to overview