Humans can learn to detect AI-generated texts, or at least learn when they can't

2025-05-03

AI Summary

This study investigates whether immediate feedback can improve individuals’ ability to distinguish AI-generated text (GPT-4o) from human-written text (Koditex corpus) and enhance confidence calibration. A behavioral experiment with 255 native Czech speakers employed a two-alternative forced-choice task, coupled with confidence and readability rating scales. Results demonstrate that immediate feedback significantly increases classification accuracy and reduces confidence calibration error by 62%. It effectively corrects systematic misperceptions—such as the belief that AI text is inherently more rigid and more readable—and provides the first empirical evidence that metacognitive awareness (“knowing what one does not know”) can be strengthened through brief feedback-based training. The key contribution lies in establishing a feedback-driven rapid recalibration mechanism, offering critical empirical support for interventions aimed at improving AI literacy and discernment of synthetic content.

Technology Category

• Humans and AI: AI for Accessibility
• Machine Learning: Calibration & Uncertainty Quantification
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success

Abstract

This study investigates whether individuals can learn to accurately discriminate between human-written and AI-produced texts when provided with immediate feedback, and if they can use this feedback to recalibrate their self-perceived competence. We also explore the specific criteria individuals rely upon when making these decisions, focusing on textual style and perceived readability. We used GPT-4o to generate several hundred texts across various genres and text types comparable to Koditex, a multi-register corpus of human-written texts. We then presented randomized text pairs to 255 Czech native speakers who identified which text was human-written and which was AI-generated. Participants were randomly assigned to two conditions: one receiving immediate feedback after each trial, the other receiving no feedback until experiment completion. We recorded accuracy in identification, confidence levels, response times, and judgments about text readability along with demographic data and participants' engagement with AI technologies prior to the experiment. Participants receiving immediate feedback showed significant improvement in accuracy and confidence calibration. Participants initially held incorrect assumptions about AI-generated text features, including expectations about stylistic rigidity and readability. Notably, without feedback, participants made the most errors precisely when feeling most confident -- an issue largely resolved among the feedback group. The ability to differentiate between human and AI-generated texts can be effectively learned through targeted training with explicit feedback, which helps correct misconceptions about AI stylistic features and readability, as well as potential other variables that were not explored, while facilitating more accurate self-assessment. This finding might be particularly important in educational contexts.

Problem

• Can humans learn to distinguish AI-generated texts with feedback
• What criteria do people use to identify AI-generated texts
• Does feedback improve accuracy in detecting AI-generated content

Innovation

• Used GPT-4o to generate diverse AI texts
• Provided immediate feedback to improve accuracy
• Analyzed stylistic and readability criteria