Mengqi Shi
← Back

Responsible AI in Supportive Systems

I am interested in responsible AI questions that arise when conversational systems are used in emotionally sensitive settings. My recent work focuses on older adults' emotional support-seeking with conversational AI and the ways safety interventions can shape those experiences.

Rather than treating safety only as a system constraint, I am interested in how users experience redirection, refusal, and interruption in practice, and how those mechanisms affect emotional pacing, dignity, and agency.

"Responsible AI should be attentive not only to system-level risk, but also to how safety is experienced by people in vulnerable moments."

Research Questions

  • When do older adults turn to conversational AI for emotional support, and what situational factors shape those decisions?
  • How do safety interventions such as redirection or interruption affect emotionally supportive interactions in practice?
  • How can conversational AI preserve user agency while still responding to real safety concerns?

Current Project Areas

  • Interview-based studies of older adults' emotional support-seeking with conversational AI.
  • Research on how system-level constraints, safety responses, and alignment practices are experienced during emotionally vulnerable moments.
  • User-centered design implications for emotionally supportive AI that respects context, pacing, and dignity.