Date
Publisher
arXiv
We present an empirical study of how both experienced tutors and non-tutors judge the correctness of tutor praise responses under different Artificial Intelligence (AI)-assisted interfaces, types of explanation (textual explanations vs. inline highlighting). We first fine-tuned several Large Language Models (LLMs) to produce binary correctness labels and explanations, achieving up to 88% accuracy and 0.92 F1 score with GPT-4. We then let the GPT-4 models assist 95 participants in tutoring decision-making tasks by offering different types of explanations. Our findings show that although human-AI collaboration outperforms humans alone in evaluating tutor responses, it remains less accurate than AI alone. Moreover, we find that non-tutors tend to follow the AI's advice more consistently, which boosts their overall accuracy on the task: especially when the AI is correct. In contrast, experienced tutors often override the AI's correct suggestions and thus miss out on potential gains from the AI's generally high baseline accuracy. Further analysis reveals that explanations in text reasoning will increase over-reliance and reduce underreliance, while inline highlighting does not. Moreover, neither explanation style actually has a significant effect on performance and costs participants more time to complete the task, instead of saving time. Our findings reveal a tension between expertise, explanation design, and efficiency in AI-assisted decision-making, highlighting the need for balanced approaches that foster more effective human-AI collaboration.
Study at a glance
- Question: How do AI assistance and explanation styles affect decision accuracy, time, and reliance among expert tutors and novices?
- Setting and data: Online experiment with 47 trained tutors and 48 Prolific participants evaluating 216 simulated math tutoring praise responses.
- Method: Online experiment using a counterbalanced repeated-measures design across five interface conditions.
Why it matters
Educators face heavy cognitive demands when evaluating student work and feedback. While AI could assist these educational decisions, little research examines how large language models and different explanation styles affect human reliance and judgment in tutoring.
Key findings
- AI assistance improved human decision accuracy compared to working alone, but human-AI collaboration consistently underperformed an AI-only accuracy baseline of 0.8.
- Novices benefited more from AI suggestions than experienced tutors, who frequently overrode correct AI advice, revealing a paradox of expertise.
- Textual reasoning explanations reduced under-reliance when AI was correct but increased over-reliance when incorrect, while both explanation modalities increased completion time without improving accuracy.
Summary drafted with AI from the full paper and may contain errors. See the full paper for details.
What is the application?
Who age?
Why use AI?
Study design
