Servant Leaders and Dark Triads: What Two AI Personas Reveal About User Experience 

Abstract 

Much of the conversation about AI persona design treats it as a styling layer, applied on top of an otherwise neutral system. Our research points to a more complex picture, in which persona choices shape user experience in ways that are not always accessible to the users themselves.

In this talk, we share findings from a psychophysiological study in which participants completed structured workplace interactions with one of two AI personas: a Servant Leader persona built around humility and developmental feedback, and a Dark Triad persona expressing narcissism and Machiavellian qualities. We measured electrodermal activity and facial EMG continuously, alongside conventional self-report measures.

The pattern we observed was a partial decoupling between physiology and self-report. Participants in the Dark Triad condition showed elevated physiological reactivity, while their self-reported experience of the interaction did not fully track that signal. The body registered the interaction in ways that conscious reports did not always capture. Standard interaction-quality measures, such as survey instruments and behavioral proxies, appeared to under-detect aspects of what the dark persona was doing.

We suggest that persona functions less as a stylistic choice and more as a design layer with cognitive and physiological consequences, and that research on AI for human flourishing may benefit from measurement tools that complement self-report. We close by discussing implications for personality calibration in deployed systems and for the methodological habits of human-AI interaction research.

Speaker Bios

Aleksandra Przegalinska

Aleksandra Przegalinska is the Vice-President and Professor at Kozminski University, and a Fellow at the Center for Labor and a Just Economy at the Harvard Law School. Her work focuses on human–AI collaboration and the future of knowledge work. A keynote speaker and a leader of the EUonAIR European University Alliance, her work focuses on human–AI collaboration and the future of knowledge work, with particular attention to how people experience, trust, and work alongside intelligent systems. She is the co-author of "Collaborative Society" (MIT Press) and "Strategizing AI in Business and Education" (Cambridge University Press), both written with Dariusz Jemielniak. She has held a research position at Center for Collective Intelligence at MIT, and speaks widely on artificial intelligence, the changing nature of work, and the social dimensions of emerging technology. Her current research combines behavioral and psychophysiological methods to study what AI systems actually do to the people who use them.

Tamilla Triantoro

Tamilla Triantoro is Associate Professor at Quinnipiac University's School of Business. Her research on human-AI interaction has evolved from early studies of how people work with intelligent systems to current investigations of how AI personality and design affect human judgment, trust, and wellbeing. Her work has appeared in leading information systems journals and has been presented at Harvard, MIT, and Yale. She is co-author of Converging Minds: The Creative Potential of Collaborative AI(Routledge, 2024) with Aleksandra Przegalińska. She serves on the advisory board of EUonAir, an alliance of European universities advancing AI in education.