Digital Human Diversity: Putting a Face on AI Agents

Abstract

If Siri had a human face, would you trust her more or would you feel uneasy? Digital humans are virtual agents (e.g., chatbots, personal assistants) controlled by artificial intelligence (AI) that have very human-realistic faces and voices. They are virtually indistinguishable in appearance from a real person. Digital humans can have any face and voice we want—even a celebrity. Digital humans are being deployed as customer service agents (replacing chatbots), as virtual influencers, and even as real estate agents and financial advisors.

We normally enable users to customize technology to fit their preferences, but should we let users choose the gender and race of the digital humans provided by the firms they do business with? That is, not their own digital humans, but rather the digital humans of others. Will such choice lead to a better user experience or will it reinforce harmful stereotypes—or both? We conducted an online experiment with 786 participants recruited through Cloud Research who watched a video showing a digital human acting as a personal assistant to a team meeting on a zoom call. Participants were randomly assigned to view one of four digital humans (black male, black female, white male, white female). We found no significant differences in perceptions of trust, satisfaction, or willingness to use. However, when given a choice, most chose a digital human matching their race and gender. Exposure to a digital human with a race or gender different than their own influenced half the participants to select a more diverse digital human.

This study is in progress, so I would greatly appreciate feedback and advice on it as well as what follow-up studies we should do to better understand the diversity, equity and inclusion issues surrounding the deployment of digital humans.

Presenter

Alan Dennis is a Professor of Information Systems and holds the John T. Chambers Chair of Internet Systems in the Kelley School of Business at Indiana University. The chair was established in honor of John Chambers, the former CEO of Cisco Systems Inc., the leading provider of networking technology. Alan’s research focuses on AI agents, fake news, cybersecurity, and team collaboration. A 2020 analysis listed him as one of the top 1% most influential researchers in the world, across all scientific disciplines. His teaching focuses on IT infrastructure and networks, and he has written four textbooks. He is a past President of the Association for Information Systems and served as Vice President for Conferences. He was named a Fellow of the AIS in 2012 and received the LEO Award in 2021.