Identifying and Categorizing Bias in AI/ML for Earth Sciences

Abstract Artificial intelligence (AI) can be used to improve performance across a wide range of Earth system prediction tasks. As with any application of AI, it is important for AI to be developed in an ethical and responsible manner to minimize bias and other effects. In this work, we extend our previous work demonstrating how AI can go wrong with weather and climate applications by presenting a categorization of bias for AI in the Earth sciences. This categorization can assist AI developers to identify potential biases that can affect their model throughout the AI development life cycle. We highlight examples from a variety of Earth system prediction tasks of each category of bias. Significance Statement As artificial intelligence (AI) grows in popularity, its methods are being applied to a wide range of Earth system prediction tasks. Although AI can facilitate more accurate prediction at many tasks, it is not without potential pitfalls, especially if the developers are not as familiar with its potential drawbacks. In this paper, we provide a classification system for the types of bias that one is likely to see in applying AI to Earth sciences. Our classification system will assist current and future AI developers to recognize where their AI system or data are biased so they can take steps to alleviate this bias.