100: Can We Trust AI? Personas, Stress and Human Unicorns

100: Can We Trust AI? Personas, Stress and Human Unicorns - with Tamilla Triantoro

Episode Details

Summary

For the 100th episode of AI Snacks, Anastassia talks with Tamilla Triantoro, who researches how people work, learn and trust alongside AI. They start with a simple question that turns out to be hard: what is trust in AI, and is it really trust, or only confidence in a system? Tamilla explains why fluent language makes us believe a machine understands us, and why we forgive a colleague's mistake but not a chatbot's.

She shares a study in which people worked with a supportive AI persona or a hostile one. Their bodies showed stress even when they said everything was fine, and the hostile group did worse work. Tamilla also describes the four AI roles she tested while writing Converging Minds, her "human unicorn" advice for small business owners, and why schools should teach students to question AI rather than ban it.

This matters for families, educators and leaders: the AI we choose shapes how people feel and think, so we need to choose it with care.

Guest Bio

Tamilla Triantoro is an Associate Professor of Business Analytics and Information Systems at Quinnipiac University's School of Business, where she co-directs the M&T Bank Center for Women and Business. Her research looks at the human costs and benefits of AI at work, frameworks for human and AI collaboration, and avatars in education and business. With Aleksandra Przegalinska she wrote Converging Minds: The Creative Potential of Collaborative AI (CRC Press, 2024), and she is working on a book about AI and human identity. She sits on the scientific board of the Human+AI Institute at CampusAI and is a fellow of the Global Justice Program at Yale. She holds a PhD from the City University of New York.

Key Takeaways

Trust or confidence? Tamilla separates trust in a person from confidence in a system, like the confidence we place in roads and traffic rules when we drive.
Fluency feels like truth. Because AI speaks smoothly from day one, we judge it as we judge a good speaker and quickly assume it understands us.
An unfair double standard. We forgive a colleague's mistake, but when AI gets something wrong, our trust collapses much faster.
Your body notices first. In Tamilla's experiment, people working with a hostile AI persona showed more physical stress than they reported, and they performed worse.
Ask how the AI will talk to your people. Organisations choose AI on speed and cost, but its personality affects staff wellbeing and output.
AI as critic works best. Of four roles Tamilla and Aleksandra tested in 2023, only the AI critic truly helped; back and forth beats taking the first answer.
Think first, then ask. Write down your own bullet points before you prompt, then compare and debate with the AI.
Teach, do not ban. The new divide is between students taught to question AI and students left alone with it; Socratic tools can help.

Guest Hyperlinks

Episode cover art for 100: Can We Trust AI? Personas, Stress and Human Unicorns - with Tamilla Triantoro