# Evaluating Epistemic Guardrails in AI Reading Assistants: A Behavioral Audit of a Minimal Prototype

arXiv: https://arxiv.org/abs/2604.27275 · HTML: https://mlsanity.com/paper/2604.27275/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 34 references that are in the map.
- First posted on arXiv 30 Apr 2026.

## Builds on (most cited first, 23 of 23)

- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [Instruction-Following Evaluation for Large Language Models](https://mlsanity.com/paper/2311.07911/) (2023)
- [Who Validates the Validators? Aligning LLM-Assisted Evaluation of LLM Outputs with Human Preferences](https://mlsanity.com/paper/2404.12272/) (2024)
- [Trust or Escalate: LLM Judges with Provable Guarantees for Human Agreement](https://mlsanity.com/paper/2407.18370/) (2024)
- [Towards interactive evaluations for interaction harms in human-AI systems](https://mlsanity.com/paper/2405.10632/) (2024)
- [EducationQ: Evaluating LLMs' Teaching Capabilities Through Multi-Agent Dialogue Framework](https://mlsanity.com/paper/2504.14928/) (2025)
- [RubricEval: A Rubric-Level Meta-Evaluation Benchmark for LLM Judges in Instruction Following](https://mlsanity.com/paper/2603.25133/) (2026)
- [Impact of Guidance and Interaction Strategies for LLM Use on Learner Performance and Perception](https://mlsanity.com/paper/2310.13712/) (2023)
- [Reading Subtext: Evaluating Large Language Models on Short Story Summarization with Writers](https://mlsanity.com/paper/2403.01061/) (2024)
- [Learning to Judge: LLMs Designing and Applying Evaluation Rubrics](https://mlsanity.com/paper/2602.08672/) (2026)
- [A review on the use of large language models as virtual tutors](https://mlsanity.com/paper/2405.11983/) (2024)
- [Firm or Fickle? Evaluating Large Language Models Consistency in Sequential Interactions](https://mlsanity.com/paper/2503.22353/) (2025)
- [Should agentic conversational AI change how we think about ethics? Characterising an interactional ethics centred on respect](https://mlsanity.com/paper/2401.09082/) (2024)
- [Beyond the Single Turn: Reframing Refusals as Dynamic Experiences Embedded in the Context of Mental Health Support Interactions with LLMs](https://mlsanity.com/paper/2602.01694/) (2026)
- [Classifying Epistemic Relationships in Human-AI Interaction: An Exploratory Approach](https://mlsanity.com/paper/2508.03673/) (2025)
- [Relational Dissonance in Human-AI Interactions: The Case of Knowledge Work](https://mlsanity.com/paper/2509.15836/) (2025)
- [A Meta-Analysis of LLM Effects on Students across Qualification, Socialisation, and Subjectification](https://mlsanity.com/paper/2509.22725/) (2025)
- [Alignment Verifiability in Large Language Models: Normative Indistinguishability under Behavioral Evaluation](https://mlsanity.com/paper/2602.05656/) (2026)
- [The Slow Drift of Support: Boundary Failures in Multi-Turn Mental Health LLM Dialogues](https://mlsanity.com/paper/2601.14269/) (2026)
- [Watching AI Think: User Perceptions of Visible Thinking in Chatbots](https://mlsanity.com/paper/2601.16720/) (2026)
- [When does AI support thinking, and when does it replace it? Learners' conceptualisations of AI as a dynamic cognitive partner: A typology](https://mlsanity.com/paper/2602.15638/) (2026)
- [Quantifying Conversational Reliability of Large Language Models under Multi-Turn Interaction](https://mlsanity.com/paper/2603.01423/) (2026)
- [Generative AI User Experience: Developing Human--AI Epistemic Partnership](https://mlsanity.com/paper/2603.23863/) (2026)

## Abstract (arXiv)

Large language model (LLM) reading assistants are increasingly used in settings that require interpretation rather than simple retrieval. In these contexts, the central risk is not only error or unsafe output, but interpretive displacement: the transfer of meaning-making work from reader to system. This paper examines that problem through the concept of epistemic guardrails, defined here as constraints on how an artificial intelligence (AI) system participates in reading and interpretation. Using TextWalk, a minimal reading-support prototype designed as a co-reader rather than an answer-provider, the study applies a fixed ten-prompt protocol to twelve analytical texts spanning four categories of argumentative prose. The protocol escalates from baseline reading support to interpretive inquiry, boundary stress, and explicit shortcut pressure, enabling guardrails to be examined as behavioral properties observable in interaction rather than as static instruction features. Results show strong baseline stability, measurable strain during interpretive inquiry, partial recovery under direct boundary stress, and late-stage stabilization under escalation pressure. The most consequential weaknesses did not appear as overt collapse, but in a middle zone between support and substitution, where the system remained grounded and pedagogical while redistributing too much interpretive labor away from the reader. The paper contributes a protocol for evaluating epistemic guardrails as interactional phenomena in conversational AI reading assistants, an empirical account of their behavioral dynamics under pressure, and an emerging model of interpretive boundary function in reading-support AI.