AI Communication and Technology

Perceived Humanness of Online Chat Agents

Imagine a common, if not daily, experience in today’s online society: you’re scrolling through a website and ping a chat box pops up, stating “How can I help you today?”

Organizations are adopting these live-chat systems that employ either human agents or AI-driven conversational agents. How consumers perceive these live-chat conversations is important in understanding how organizations build strong relationships with them. However, perceptions of human-to-human and human-to-machine interactions can be a challenging concept to understand and study as technology like ChatGPT continues to blur the boundaries between what we assume to be humans and machines.

A team led by University of Florida Graduate School Associate Dean for Graduate Affairs Tom Kelleher, an Advertising professor at the UF College of Journalism and Communications, sought to ease this challenge by developing a way to measure consumer perceptions of these interactions. The team focused on humanness, or perceptions of how human-like these live chat interactions are.

Their measure to understand consumers’ perceptions of the human-like qualities of both human- and machine-driven chats involved three concepts: conversational human voice (how natural and engaging interactions between consumers and organizations are), anthropomorphism (the tendency to apply human-like qualities to nonhuman agents), and social presence (how real an actual person is perceived to be online). Combining these three concepts, the team created a scale that measures consumer perceptions of humanness.

To test and apply their new scale, the team conducted two studies that used both human and machine-generated live chats. In the first study, most, but not all, participants correctly identified whether they were chatting with a human or machine from companies such as Express, Amazon or Best Buy. However, more than a third incorrectly reported that the artificial conversational agent with whom they communicated was human or that a human agent was a machine. Results of the second study showed that perceived humanness of conversational agents is associated with higher levels of trust in an organization through perceived investment. In other words, the more human-like a conversational agent is, the more consumers believe the organization invests in the relationship, and thus, the more trust they have in the organization.

The research shows the importance of human-like qualities in online agents for organizations to build trust in consumer relationships. It also lays a foundation for continued research, offering a valid way to measure the perceived humanness of online content. To conclude, the authors emphasize the applicability of the study both theoretically, to drive future research in this area, and practically, to establish the clear benefits of humanness in organizational online agents.

The original article, “Measuring Consumer-Perceived Humanness of Online Organizational Agents,” was published in Computers in Human Behavior, Volume 128, March 2022.

Authors: Lincoln Lu, Casey McDonald, Tom Kelleher, Susanna Lee, Yoo Jin Chung, Sophia Mueller, Marc Vielledent, Cen April Yu