Meeting LLMs for the First Time: Understanding Trust Formation and Adjustment Among Older Adults in China

Against the backdrop of rapid population aging and the proliferation of Large Language Models (LLMs) in China, this study investigates the dynamics of trust formation and adjustment among Chinese older adults during their initial encounters with LLM agents. Through in-depth interviews conducted before, during, and immediately after 30 older adults’ first interactions with six widely used Chinese LLM applications, followed by second interviews with 26 participants, the study examines how trust was initially formed and adjusted over the short term. Findings reveal that pre-interaction responses were predominantly characterized by low interest and skepticism, rooted in sociocultural perceptions rather than just technological barriers. Analysis identified four technological factors associated with participants’ trust-related judgments during engagement, including personalization, anthropomorphism, conversationality, and response latency. Longitudinal analysis identified three behavioral patterns of trust adjustment: Limited Correction, Collaborative Reciprocity, and Cognitive Entrenchment. These findings advance theoretical understanding of trust adjustment and calibration-related processes and offer design implications for building more trustworthy LLMs.