Reality Monitoring in Large Language Models: Self-Knowledge That Transforms with Conversation Memory
2026-07-27 • Artificial Intelligence
Artificial IntelligenceComputation and LanguageComputers and Society
AI summaryⓘ
The authors studied whether large language models (LLMs) can tell the difference between what they said and what a user said, a skill called reality monitoring in humans. They found that LLMs struggle with this, especially when memory demands increase, sometimes mixing up their own statements with those of users. Additionally, some models become more confident even when they are wrong, revealing issues not detected by current tests. The authors suggest that simply knowing facts isn't enough for AI; it also needs to track where those facts come from.
conversational AIreality monitoringlarge language modelssource attributionepisodic memoryconfidence calibrationhallucinationsmulti-turn dialogueparameter countbenchmarking
Authors
Saurabh Ranjan, Konstantina Sokratous, Brian Odegaard
Abstract
A conversational AI that cannot tell its own output from what a user said will treat its own mistakes as user-provided facts. In humans, this capacity is called reality monitoring, and its failures are linked to hallucinations, delusions, and confabulation, yet whether LLMs possess it remains untested. Here we show, across two experiments and six LLMs, that source attribution depends on how conversational memory is structured: ceiling accuracy for self-generated content under minimal memory demands reverses to a fragile external-item advantage once episodic delay removes that shortcut. Feedback exposes two failures: in some models, internal and external judgments swap; in others, accuracy improves while confidence decouples from correctness, dissociations invisible to existing benchmarks. Across models, this pattern implicates active, not aggregate, parameter count. This suggests that as AI systems take on autonomous, multi-turn roles, evaluating what they know is not enough: tracking where that knowledge came from may matter equally.