AI-Generated Images: What Humans and Machines See When They Look at the Same Image

TL;DR

This paper develops and evaluates explainable AI detectors for identifying AI-generated images, emphasizing human-understandable explanations and insights into visual-language cues, trained on a large fake image dataset.

Contribution

It introduces a comprehensive detection framework with multiple architectures and XAI methods, evaluated through human surveys for clarity and effectiveness.

Findings

XAI methods can be refined to improve human understanding of AI-generated images.

The detection framework effectively distinguishes fake images from real ones across various generators.

Human preferences align with certain visual and textual explanations, enhancing interpretability.

Abstract

The misuse of generative AI in online disinformation campaigns highlights the urgent need for transparent and explainable detection systems. In this work, we investigate how detectors for AI-generated images can be more effective in providing human-understandable explanations for their predictions. To this end, we develop a suite of detectors with various architectures and fine-tuning strategies, trained on our large-scale photorealistic fake image dataset, AIText2Image, and assess their performance on state-of-the-art text-to-image AI generators. We integrate 16 different explainable AI (XAI) methods into our detection framework, and the visual explanations are comprehensively refined and evaluated through a novel approach that prioritizes human understanding of AI-generated images, using both textual and visual responses collected from a survey of 100 participants. This framework offers insights into visual-language cues in fake image detection and into the clarity of XAI methods from a human perspective, measuring the alignment of XAI outputs with human preferences.