Clearing up biases in artificial intelligence
Group's goal is to help environmental scientists learn the basics of AI


Date:
April 20, 2022
Source:
University of Oklahoma
Summary:
Scientists have noticed grave disparities in artificial
intelligence, noting that the methods are not objective, especially
when it comes to geodiversity. AI tools, whether forecasting hail,
wind or tornadoes, are assumed to be inherently objective, says
one of the researchers. They aren't, she says.



FULL STORY ========================================================================== There's no doubt that artificial intelligence is embedded in our everyday lives. From smartphones to ridesharing apps to mobile check deposits,
AI is so pervasive that we rarely think about how it works.


==========================================================================
For one University of Oklahoma scientist, however, artificial intelligence
and machine learning are at the forefront of her work -- expressly as
it relates to weather. Amy McGovern, Ph.D., leads the National Science Foundation AI Institute for Research on Trustworthy AI in Weather,
Climate, and Coastal Oceanography at the University of Oklahoma.

An American Meteorological Fellow, McGovern has been studying severe
weather phenomena since the late 1990s. During her career, she has
witnessed a rapid emergence in the AI field, all while developing what she hopes are trustworthy AI methods to avert weather and climate disasters.

Lately, however, McGovern and researchers from Colorado and Washington
have noticed grave disparities in AI, noting that the methods are not objective, especially when it comes to geodiversity.

"Artificial intelligence algorithms are based on mathematical formulas
that are seen as objective; however, there is a bias toward areas with
higher populations, as well as areas that are more affluent," said
McGovern, a professor at OU's School of Computer Science and School
of Meteorology.

"For example, if more people live in an area, there is a higher chance
that someone observes and reports a hail or tornado event. This can
bias the AI model to over-predict hail and tornadoes in urban areas and under-predict severe weather in rural towns," she said.

AI tools, whether forecasting hail, wind or tornadoes, are assumed to
be inherently objective. They aren't, McGovern says.

Raising Awareness The team recently published a paper titled "Why We Need
to Focus on Developing Ethical, Responsible, and Trustworthy AI Approaches
for Environmental Sciences." Published by Cambridge University Press, the
paper will appear in the inaugural issue of Environmental Data Science.

The researchers are exploring ethical AI methods, specifically in the
field of environmental sciences. "Whether involved in teaching, industry
or government, environmental scientists are absolutely essential for
developing meaningful AI tools, and more educational resources are
needed to help environmental scientists learn the basics of artificial intelligence so they can play a leading role in future developments,"
McGovern said.

The group sees ethics in AI in the environmental sciences as an emerging
trend in education. "With the rapid emergence of data science techniques
in the sciences and the societal importance of many of these applications, there is an urgent need to prepare future scientists to be knowledgeable," McGovern said.

AI systems can be as flawed as the people who create them and can unintentionally do more harm than good if not developed and applied responsibly, McGovern says. "We hope our work is a major step toward
making AI systems more ethically informed in environmental science."

========================================================================== Story Source: Materials provided by University_of_Oklahoma. Note:
Content may be edited for style and length.


========================================================================== Journal Reference:
1. Amy McGovern, Imme Ebert-Uphoff, David John Gagne, Ann Bostrom. Why
we
need to focus on developing ethical, responsible, and trustworthy
artificial intelligence approaches for environmental science.

Environmental Data Science, 2022; 1 DOI: 10.1017/eds.2022.5 ==========================================================================

Link to news story: https://www.sciencedaily.com/releases/2022/04/220420133607.htm

--- up 7 weeks, 2 days, 10 hours, 51 minutes
* Origin: -=> Castle Rock BBS <=- Now Husky HPT Powered! (1:317/3)