Photo by Shantanu Kumar on Unsplash
Are people more likely to rely on and interact with a chatbot when it seems more human? That was the question NIU College of Business faculty members Andy Jeon, Devaki Rau, Matt Pickard and Biagio Palese set out to answer when they joined forces for an AI-age research project.
In February, after four years of research, the authors published their joint research article, “Reluctant to ask a human-ish chatbot? When anthropomorphic cues help and when they backfire,” in the Marketing Education Review. Through their research, the four faculty members conducted two interesting studies that examined the results of college students’ interactions with chatbots.
“Interacting effectively with AI is rapidly becoming a required skill in the workforce, much like knowing how to use a spreadsheet or write an email. It’s no longer optional,” said Palese. “That’s why it’s critical that students get hands-on experience with AI tools while they are still learning — in an environment where mistakes are safe and guidance is available — so they graduate truly ready for what employers expect of them.”
From the beginning of this process, the authors challenged what seemed to be a commonly held assumption: That it is more ideal to “humanize” a chatbot to make it feel friendly and approachable. From there, the team piloted a more humanized chatbot design and found that some students actually disliked this humanlike chatbot.
“That result made us curious, so we explored further to understand why,” Jeon said. “We theorized that the same humanlike cues are perceived differently depending on what we call a ‘schema,’ or people’s mental framework and prior expectations that shape how they interpret what they see.”
For the next several years, the team looked at how students, especially marketing students, respond when technological tools such as chatbot assistants attempt to mimic a human instructor. Across two experimental studies, their work examined how the effects of humanlikeness in chatbot assistants (anthropomorphization) depend on students’ cultural values and expectations about authority. Challenging the assumption that adding humanlike features to AI is inherently desirable, results from Study 1 find that the presence of humanlike cues in chatbot assistants does not universally enhance perceptions of humanlikeness. Instead, students who are more accepting of hierarchical relationships perceive anthropomorphic chatbots as less humanlike than nonanthropomorphic chatbots. This effect was more pronounced among marketing than nonmarketing students.
“For me, introducing a chatbot is exactly like introducing a new friend, colleague, peer or ‘tutor,’” said Jeon. “As social beings interacting with another social entity, we first need to understand how humans respond to such interactions. Rather than assuming that making a chatbot more humanlike is always better, our findings show that different schemas lead students to react differently. Therefore, we need to understand our students carefully and design chatbot features that are thoughtfully tailored to their expectations and cultural values.”
Study 2 built upon the first study’s findings by analyzing real-time student interactions with a custom-built chatbot. While perceived humanlikeness increases engagement and subsequently improves attitudes toward use, students with high conformity to authority ask significantly fewer questions when the chatbot appears highly humanlike. These findings reveal a novel cultural boundary condition for effective student interactions with humanlike technology: When students value deference to authority, humanlike bots may inhibit rather than enhance student engagement with chatbots.
Beyond the importance of this research in a world that is increasingly integrated with AI, this research is unique because it brought together researchers with very different ways of viewing the world. Rau, Jeon, Pickard and Palese work in the college’s departments of management, marketing, accountancy, and operations management and information systems, respectively.
“It was great to have faculty from different areas working on this research together because we each brought different perspectives to the problem,” Rau said. “As a result, we got a more holistic picture of the problem than if we had stayed within our silos. Working with a diverse team of individuals requires us to step out of our comfort zones. However, the richer outcomes — more interesting results with direct implications for the classroom — are well worth the extra effort.”
Known for its cutting-edge teaching and research, the NIU College of Business facilitated this publication. Specifically, the college enacted an interdisciplinary research grant in summer 2021 to connect these researchers and the research committee provided feedback on the proposal the researchers submitted.
“Interdisciplinary research is challenging and interesting. Yet it reveals some of the most useful, nuanced and interesting insights,” said Pickard. “The College of Business’ support of interdisciplinary research was a nice gesture validating the value of interdisciplinary research.”
Learn more about this article.
