AI reading visualization accuracy measured with Stevens power law
A Stevens's Power Law Check-up of GPT-5.5's Image-Based Visualization Reading
Computer Vision and Pattern Recognition
Summary
It can be hard to tell how well AI understands pictures like charts and graphs. This paper tests how AI models estimate amounts shown in different images without any explanations or legends. The researchers used a method called Stevens's power law to see how AI perceives quantities compared to humans. They looked at twelve different ways visuals can represent data to make AI reading easier to understand and compare with how people see things.
What this means in practice
- •For data visualization designers: Assess how well AI can understand different visualization methods to improve automated visual data interpretation tools.
- •For user interface developers: Design AI-assisted interfaces that better interpret images without relying on legends for faster user support.
Authors
Kaichun Yang, Jian Chen
Abstract
We adapt Stevens's power law to measure the innate ability of AI models to read visualizations, which can reveal the built-in perceptual mechanisms of algorithmic models. In our pilot study, models see no legend. A model first views a reference visual representation and estimates its magnitude, then estimates the magnitude of each subsequent image of the same representation relative to that reference. Our evaluation of twelve visual variables makes how algorithmic models read visual encodings measurable, comparable with human perception, and more interpretable to humans.