Who Gets to Make AI Safe? Toward Participatory AI Safety
As AI systems become embedded in everyday life, an increasingly important question is: who gets to define what it means for them to be safe? Technical safeguards and standardized benchmarks are essential, but users, workers, and impacted communities often encounter harms and tensions that conventional evaluations miss.

In this talk, I introduce participatory AI safety, an approach that brings these forms of situated knowledge into how AI systems are designed, evaluated, and governed. Drawing on our research spanning everyday algorithm auditing, community-centered AI design, and emerging relational systems such as AI companions, I will examine both the promise and limits of participation. Participation can reveal harms and values otherwise overlooked, but it can also shift substantial labor and responsibility onto the people experiencing AI failures. I will discuss how participatory approaches can complement technical evaluation while empowering people to shape AI rather than simply asking them to bear the burden of making it safe.

About our speaker
Dr. Hong Shen is an Assistant Professor in the Human-Computer Interaction Institute at Carnegie Mellon University’s School of Computer Science, where she directs the CARE (Collective AI Research and Evaluation) Lab. Her research focuses developing human- and community-centered approaches to understanding, evaluating, and mitigating AI harms. Her research has received awards at leading venues including ACM CHI and CSCW and has been featured in The New York Times, Forbes, Wired, BBC, and The Guardian. She is a recipient of the Google Research Scholar Award, Microsoft Research Global AI & Society Fellowship, and Google Policy Fellowship.

How to Get There
Public transportation: Take the Orange Line to Ruggles or the Green Line to Northeastern University. Upon arrival to Khoury College, West Village H, take the elevator to the third floor, then turn left and proceed to Room 366.

‘Hacking’ Assistive Technology: Creating and Customizing AI Tools for Personal Accessibility

Existing AI assistive applications for blind people tend to be designed for common use cases in order to be broadly applicable, for instance, reading printed text or identifying objects. Yet, individuals have unique skills, goals, and contexts, leading to a long-tail of tasks that one-size-fits-all assistive technology cannot address well. These gaps point to limitations in the current design process for assistive technologies: solutions are designed for broad groups of people, and participants are involved in limited stages of the design process (typically formative studies, short design check-ins, and evaluations). In this talk, I will present my research applying a Do-It-Yourself (DIY) approach to assistive technology, enabling people with disabilities to create AI-powered tools for their unique needs. In doing so, my approach shifts agency from researchers and developers to people with disabilities. I will first present my qualitative research on current technology gaps and DIY practices among blind people, showing how people with disabilities often already spend a high degree of effort customizing and creating workaround ‘hacks’ to make technology work for them. I will then present two tools aimed at harnessing and better supporting these DIY efforts, by enabling people to create and modify assistive technologies for themselves. Overall, my research aims to promote the democratization of AI technology creation, and support blind people in having greater control over AI-based technologies in their lives.

About our speaker
Jaylin Herskovitz joined the Department of Computer Science at Tufts University as an Assistant Professor in January 2026. Her research focuses on accessibility and AI-based assistive technologies, DIY technology, and AR/VR. Her work combines qualitative methods with technical system building, and she works with people with disabilities to design, develop, and evaluate AI-based assistive technology. Her work has been published in top-tier HCI conferences, including CHI, UIST, and ASSETS. Jaylin holds a Ph.D. in Computer Science from the University of Michigan, where she was advised by Anhong Guo. During her Ph.D., Jaylin was supported by an NSF Graduate Research Fellowship and completed internships with Apple AI/ML Research and Microsoft Research.

Naviagation: Enter the building through this gate and take left.

Nearest T station is Ruggles on Orange line and Northeastern University on Green line

‘Hacking’ Assistive Technology: Creating and Customizing AI Tools for Personal Accessibility

Existing AI assistive applications for blind people tend to be designed for common use cases in order to be broadly applicable, for instance, reading printed text or identifying objects. Yet, individuals have unique skills, goals, and contexts, leading to a long-tail of tasks that one-size-fits-all assistive technology cannot address well. These gaps point to limitations in the current design process for assistive technologies: solutions are designed for broad groups of people, and participants are involved in limited stages of the design process (typically formative studies, short design check-ins, and evaluations). In this talk, I will present my research applying a Do-It-Yourself (DIY) approach to assistive technology, enabling people with disabilities to create AI-powered tools for their unique needs. In doing so, my approach shifts agency from researchers and developers to people with disabilities. I will first present my qualitative research on current technology gaps and DIY practices among blind people, showing how people with disabilities often already spend a high degree of effort customizing and creating workaround ‘hacks’ to make technology work for them. I will then present two tools aimed at harnessing and better supporting these DIY efforts, by enabling people to create and modify assistive technologies for themselves. Overall, my research aims to promote the democratization of AI technology creation, and support blind people in having greater control over AI-based technologies in their lives.

About our speaker
Jaylin Herskovitz joined the Department of Computer Science at Tufts University as an Assistant Professor in January 2026. Her research focuses on accessibility and AI-based assistive technologies, DIY technology, and AR/VR. Her work combines qualitative methods with technical system building, and she works with people with disabilities to design, develop, and evaluate AI-based assistive technology. Her work has been published in top-tier HCI conferences, including CHI, UIST, and ASSETS. Jaylin holds a Ph.D. in Computer Science from the University of Michigan, where she was advised by Anhong Guo. During her Ph.D., Jaylin was supported by an NSF Graduate Research Fellowship and completed internships with Apple AI/ML Research and Microsoft Research.

Naviagation: Enter the building through this gate and take left.

Nearest T station is Ruggles on Orange line and Northeastern University on Green line