---
source_url: https://arxiv.org/abs/2604.01114
ingested: 2026-08-03
sha256: f6887df8ebe3f244c165878ea719aa52d160d6f854c620de71038b5573d4e651
---

# Trust and Reliance on AI in Education: AI Literacy and Need for Cognition as Moderators

Griffin Pitts, Neha Rani, Weedguet Mildort. arXiv:2604.01114 [cs.HC]. Full paper accepted to AIED 2026 (27th International Conference on AI in Education). v3, 8 Jun 2026.

## Design

- 432 undergraduate participants
- Students completed Python output-prediction problems while receiving recommendations and explanations from an AI chatbot, **including accurate and intentionally misleading suggestions**
- **Appropriate reliance** operationalised behaviourally: the extent to which students' responses reflected appropriate use of suggestions — accepting correct ones, rejecting incorrect ones
- Pre/post-task surveys: trust in the assistant, AI literacy, need for cognition (NFC), programming self-efficacy, programming literacy

## Findings

- **Non-linear relationship: higher trust was associated with LOWER appropriate reliance** — students with more trust discriminated less between correct and incorrect recommendations
- This relationship was **significantly moderated by AI literacy and need for cognition** (learner characteristics shape how trust translates into reliance behaviour)
- Highlights the need for instructional and system supports that encourage more reflective evaluation of AI assistance during problem-solving

## Relevance

Trust alone is not a good predictor of appropriate AI use; over-trust → overreliance. Supports designs that calibrate trust (verification prompts, confidence displays) and curricula that build AI literacy + reflective evaluation habits.