Procedural memory mismatch does not always disrupt web task behavior
Procedural Memory Under Change: Reuse and Interference in Controlled Web Tasks
Artificial Intelligence
Summary
Procedural memory helps language agents reuse routines they learned before, but this only works if the routines still apply. The authors tested what happens when the stored routines no longer exactly match the current task in web-based shopping decisions. They found that even when there are mismatches, these do not necessarily cause errors or disruptions in behavior. This shows that procedural memory can sometimes handle changes without breaking functionality, though the authors did not determine what conditions cause errors.
What this means in practice
- •For web interface developers: Design adaptation strategies for web agents that tolerate interface changes without retraining stored procedures.
- •For ai system engineers: Build AI systems with procedural memory that can handle some mismatches in task settings without performance loss.
Authors
Yanze Cao
Abstract
Procedural memory lets language agents reuse successful routines, but reuse presumes that a stored routine remains applicable. We study what happens when that presumption is deliberately violated. The study combines a retrospective, human-assisted interface-adaptation case from BrowserGym TimeWarp with controlled frozen-memory comparisons on synthetic shopping decisions. During the documented WebShop V1-V6 development path, interface-specific code was adapted while the separately stored high-level procedure was not reported to change; this phase does not constitute an autonomous memory-agent evaluation. In the controlled phase, an early pilot produced one task on which two memory conditions selected a more expensive item while the no-memory condition selected the reference minimum. Follow-up probes did not establish a recurring row-order or identity-binding pattern. We then tested four forms of mismatch: changed quantities, a different evidence representation, a conflict between local and global optimization, and distributed promotion evidence, across 32 formal cells. Each cell used one temperature-0 generation with the same local qwen3:8b configuration and no adaptive retry. Across these pairs, none of the predefined diagnostic interference signatures appeared on the tasks for which they were defined when current-task evidence was explicit and sufficient. The result identifies a tested region of non-interference: a procedural memory can be mismatched without becoming behaviorally disruptive. It does not establish general safety or a mechanism. The remaining question is which additional conditions turn applicability mismatch into observable, memory-caused error.