Abstract

As AI becomes part of how we learn, seek advice, and make everyday decisions, how does that affect our psychological and physical wellbeing, as well as the society we live in?

In this talk, Chayapatr Archiwaranguprok, Rachel Poonsiriwong, and Krystal Kirkland introduce ImpactBench, an open platform developed at MIT Media Lab to evaluate AI behaviors that may support or undermine human well-being. Using simulated conversations, ImpactBench examines how AI models respond in situations involving learning, mental health, relationships, and personal choices.

The speakers will also introduce AI Nutrition Labels, which translate evaluation results into accessible summaries of each model’s strengths and weaknesses. Just as food labels help people understand what they consume, these labels aim to help people make informed choices about the AI they use.

Speakers Bios

Chayapatr (Pub) Archiwaranguprok

Chayapatr (Pub) Archiwaranguprok is a Thai researcher and creative technologist at the MIT Media Lab’s Cyborg Psychology group, researching human-AI interaction, information science, and applied computing. His work has been published in venues including ACM CHI, IUI, DIS, Hypertext, AHs, SAC, ACM-IEEE JCDL, NeurIPS Creative AI, and PNAS. His work on AI-induced false memories received a Best Paper Award at CHI 2025, with coverage in TIME, Bloomberg, and The Atlantic. As a technologist, he contributed to Cyber Subin, a touring dance performance that opened the 2025 Holland Festival.

Before joining MIT, Pub co-founded Creatorsgarten, a collective of creative technologists in Thailand behind community projects such as the Stupid Hackathon Thailand, now in its ninth year. He has also been a technical advisor on the Netflix series Mad Unicorn, editor-in-chief of the Thai space outlet Spaceth.co, and a host of the ThaiPBS science program Starstuff.

Rachel Poonsiriwong

Rachel Poonsiriwong is a Singaporean researcher and interaction designer at the MIT Media Lab, where she is part of the Cyborg Psychology research group directed by Dr. Pat Pataranutaporn. Working across human-computer interaction, psychology, and artificial intelligence, she studies how AI interfaces reshape human agency, cognition, and social relationships in the short and long term. Before joining MIT, Rachel was a founding product designer at Scale AI, where she led the design of machine-learning infrastructure used to develop autonomous systems for organizations including Toyota, Nuro, Velodyne, and iRobot. She later served as design lead for Scale’s enterprise generative-AI platform.

Rachel studied at the Harvard Graduate School of Design (Master in Design Studies) before joining the MIT Media Lab, and holds a BFA in Interaction Design from the California College of the Arts.

Krystal Kirkland

Krystal Kirkland is an engineer, artist, and sociologist whose work centers on observation and storytelling across mediums. She is a senior software engineer at Arize AI, where she builds front-end systems for an AI assistant that helps developers debug, evaluate, and monitor AI models. She started at Arize six years ago in documentation and user education. Today she focuses on the interface layer of AI: how complex systems present themselves, how they earn user trust, and how empathy shapes simple, powerful interactions.

Outside Arize, she runs a creative studio for editorial photography and creative direction. In 2020, she created Machine in Relief, a textile concept that used a generative adversarial network as both a technical system and a creative medium. She also maintains an email service that shares news of deaths within Connecticut’s Jewish community, working with funeral homes to keep it running. She holds a BA in Sociology from UC Berkeley.