Generative AI (GenAI) assistants are increasingly used in workplace and customer-facing contexts, yet their feedback mechanisms often rely on low-bandwidth signals such as thumbs up/down that offer limited actionable detail. We examine this design problem through the combined lenses of social listening and hybrid intelligence (HI). From a social listening perspective, feedback is not only an evaluative input but also a form of bi-directional communication that signals whether a system is attentive and responsive to its users. From an HI perspective, feedback can be framed as a collaborative contribution to iterative system refinement rather than as a simple end-user rating. We report a preliminary study comparing two feedback designs in a GenAI marketing-assistant prototype: (1) a conventional icon-based feedback affordance with optional comments, and (2) an HI-framed in-chat solicitation that explains how user feedback will be used to improve the system. Marketing professionals recruited via Maze.co completed 19 sessions. Likert-scale measures showed no significant differences in trust, willingness to provide feedback, or intended use. HI sessions produced significantly longer open-ended responses, and qualitative analysis suggests that the HI design reduced interpretation cost and made the feedback interaction feel more visible and collaborative. These preliminary findings suggest that designing feedback as social listening may improve the richness of user input while also surfacing trade-offs between low-friction interaction and expressive nuance.
Accepted for Hybrid Human AI (HHAI) conference 2026.