Users value proactive control in memory for conversational AI
Characterizing Memory Misalignment in Human-LLM Interaction From User Perspectives
Human-Computer Interaction
Summary
When AI chatbots try to remember things about users, sometimes the memories don’t match what users expect, causing confusion. The authors studied many examples of these memory mismatches from user diaries and usage logs to understand where and why the problems happen. They then worked with designers to come up with different ways users could interact with the AI’s memory to fix these issues. Testing these ideas with more users showed people prefer tools that let them actively manage memory instead of complicated or passive options. Overall, the research suggests balancing easy controls with smooth conversations is key.
What this means in practice
- •For conversational ai developers: Design memory features that allow users to proactively control and correct remembered information to better match expectations during AI chats.
- •For ux designers for digital assistants: Implement interaction strategies that balance user oversight with minimal conversation disruption to improve user trust in assistant memory.
Authors
Jingruo Chen, Shuning Zhang, Eryue Xu, Jianing Li, Xin Yi
Abstract
While memory enhances personalization in LLM-based conversational agents, it suffers from memory misalignment, where memories violate user expectations. We present a mixed-methods investigation to characterize and mitigate memory misalignment from user perspectives. First, we collected data from memory usage (N=28, 457 entries) and diary study (N=32, 304 reports), which yielded a taxonomy spanning 14 misalignment types across memory intake, storage and management, retrieval and interpretation stages. Second, four co-design workshops with 12 experienced HCI researchers derived a design space to tackle memory misalignment issues, consisting of 12 candidate interaction strategies structured across interaction form, placement and intrusiveness dimensions. Finally, a speed dating with 121 users reveals preference heterogeneity, where users prioritize proactive controls over cognitively demanding causal graph inspections or passive audit logs. Synthesizing these findings, we highlight the tension between supervisory agency and interaction overhead, and advocate for friction-aware memories that balance user oversight with conversation smoothness.