Content area
Full text
While artificial intelligence (AI) has demonstrably improved prediction and understanding of many environmental science phenomena (e.g., Ahijevych et al. 2016; Williams et al. 2016; McGovern et al. 2017; Gagne et al. 2017, 2019; Lagerquist et al. 2019a; Barnes et al. 2019; Reichstein et al. 2019; Boukabara et al. 2021), there is often a lack of trust by environmental science decision-makers when it comes to relying on “black box” algorithms, especially in life-or-death situations (Karstens et al. 2018; Demuth et al. 2020). Developing AI that is trustworthy and useful for environmental risk management requires fundamental natural, mathematical, and social sciences research on the AI needs and perceptions of key users. These users’ judgments and decisions may depend on their expertise and context (Larkin et al. 1980; Chi et al. 1981; Payne et al. 1992). Such research should include a users’ understanding and perceptions of the AI method, its performance, and other factors emerging in empirical and theoretical research on AI (Mueller et al. 2019; Wang et al. 2019; Glikson and Woolley 2020).
We introduce the NSF AI Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography (AI2ES), a national AI institute that conducts convergent research focused on creating trustworthy AI for the weather, climate, and ocean communities. AI2ES seeks to uniquely benefit humanity by developing novel physically based AI techniques that are demonstrated to be trustworthy, and to directly improve prediction, understanding, and communication of high-impact environmental hazards.
Developing trustworthy AI, particularly for the weather, water, and climate communities, is an urgent and timely priority (IPCC 2018; Reidmiller et al. 2018; ERISS Corporation and The Maritime Alliance 2019) at the highest levels of government and industry. NOAA has identified AI as a high priority in their strategic AI plan (NOAA 2019). Similarly, the White House continues to prioritize the development of innovative AI (National Science and Technology Council 2019b; Office of Science and Technology Policy 2019; National Science and Technology Council 2019a); the National Academies of Science, Engineering and Medicine cites improved forecasting of extreme events as a critical task (National Academies of Sciences, Engineering, and Medicine 2016); and NOAA and the National Weather Service (NWS) have increasingly focused on providing impact-based decision support services (IDSS) to reduce weather risks (Uccellini and...