AI's Clever Hans Effect in Chemistry: Style Signals Mislead Activity Predictions

This research highlights a critical vulnerability in AI models applied to chemistry, demonstrating that they can be misled by stylistic features in datasets rather than truly understanding chemical properties. This has significant implications for the reliability of AI-driven drug discovery and materials science.

• •AI models can be fooled by superficial stylistic cues in chemical data.
• •The research emphasizes the importance of robust data and model evaluation.
• •Findings suggest a need for improved AI training and validation methodologies in chemistry.