The School of Communications faculty member explored how conversational styles can influence the way people communicate with AI and how they evaluate the responses they receive.
Qian Xu, professor of strategic communications and A.J. Fletcher Professor, collaborated with Cheng Chen, assistant professor at Oregon State University, on a study examining how AI conversational styles interact with users’ expectations of the roles AI should play.
Their article, “” was recently published in Behaviour & Information Technology, a peer-reviewed academic journal that focuses on human-centered information technology.
The experiment randomly assigned 137 college students to one of three Microsoft Copilot conversational styles of Copilot – creative, balanced or precise – and measured which role they expected the tool to fulfill: co-author, creator, curator or converser. Participants then completed at least three rounds of interaction with Copilot to develop a strategic plan addressing a public health issue in their home county.
Key findings include:
- Copilot’s conversational style affected the readability and linguistic features of its responses, including the use of words associated with analytical thinking and authenticity.
- Users adapted their writing based on Copilot’s conversational style, particularly in the readability of their prompts.
- Users and Copilot showed linguistic alignment in emotional tone and clout across all three conversational styles but diverged in analytical thinking.
- Mismatches between Copilot’s conversational style and users’ expectations of its role reduced perceived credibility and relevance of its responses.
“AI conversational style is more than a matter of preference. What matters is how well the style fits the roles users expect AI to play,” Xu said. The findings point to opportunities for more adaptive AI interfaces that adjust conversational style based on users’ expectations and tasks.
Behaviour & Information Technology is an international peer-reviewed Taylor & Francis journal covering the human aspects of information technology, including usability, user experience, human-computer interaction and human-centered design.