Ethical and Responsible AI for the Environmental Sciences

As a key part of our mission in developing trustworthy AI for a variety of weather and climate applications, AI2ES is leading research on ethical and responsible AI across the environmental sciences.  This page describes our current work in this area.

  • Leaders: Amy McGovern (OU), Imme Ebert-Uphoff (CSU), David John Gagne (NCAR), Ann Bostrom (UW)
  • Members: Marie McGraw, Kate Musgrave (CSU), Randy Chase (OU)

Goals

The introduction of AI in many fields has had many unintended side effects, such as coding bias into algorithms (see Coded Bias for example). We want to avoid repeating these ethical mistakes for our weather, climate, and coastal oceanography applications while advancing the use of AI in environmental science applications.

The goals of this effort are:

  • Develop clear guidelines on how to ethically and responsibly develop AI for environmental science applications
  • Develop clear guidelines on how to ethically and responsibly deploy AI for environmental science applications
  • Highlight the ways in which AI can go wrong for environmental sciences and bring this to the attention of AI developers and users 
  • Develop approaches to address many of the potential issues for AI developers, such as approaches to handling biased data or to automatically identifying bias in training data.
List of issues slide
A non-exhaustive list of issues that can arise though the use of AI for environmental science applications.  This list comes from our paper.

Progress

  • Published paper:
    • McGovern, A., Ebert-Uphoff, I., Gagne, D., & Bostrom, A. (2022). Why we need to focus on developing ethical, responsible, and trustworthy artificial intelligence approaches for environmental science. Environmental Data Science, 1, E6. doi:10.1017/eds.2022.5
    • This article has already attracted enormous attention not only in the environmental science community, but also in other disciplines. It has been mentioned in articles by several news outlets, including The Conversation and Science Daily.
  • In current work, we are focusing on the specific topic of how bias in data and AI architectures can affect AI models.  We will demonstrate how different approaches to bias affect the models specifically for weather and climate data.

Resources for more learning

  • Trustworthy Artificial Intelligence for Environmental Science Summer School 2021
  • Computer Science 5970: AI, Ethics, and Geoethics
  • AI2ES Site-wide talks:
    • October 27 – Ethical AI presentation by Dean Kaye Husbands Fealing, Georgia Tech. Recording
    • April 28 – Responsible Use of AI – Imme Ebert-Uphoff, CSU. Presentation Slides; Recording
  • Panel Discussion: Responsible and Ethical Use and JEDI Issues for ML/AI in Weather, Climate and Earth System Science. National Academies of Sciences, Engineering and Medicine, Workshop on Machine Learning and Artificial Intelligence to Advance Earth System Science: Opportunities and Challenges. Panelists: D. Danks, I. Ebert Uphoff, P. Donti, and A. Gupta; Moderators: A. McGovern and A. Bostrom. 2022. link
  • Presentation: McGovern, Amy; Ebert-Uphoff, Imme; Bostrom, Ann; Gagne, David J. (2022). Ethical and Responsible AI and Trust for Weather and Climate. 21st Conference on Artificial Intelligence for Environmental Science, Houston, TX, American Meteorological Society. link
For more information, please contact Amy McGovern or Imme Ebert-Uphoff. 

AI2ES logo equation