AI Bias

Notice anything interesting about this AI-generated image (besides the mélange of towers grafted onto a UW Tacoma-like building)?

How about the apparently minimal diversity of race and gender? 

The prompt used to create this image was: "create a picture of university professors discussing artificial intelligence with their students at the University of Washington Tacoma (in landscape orientation)." 

ChatGPT and other "AI" systems are only as good—and bad—as the input they are given. This picture reflects the statistical preponderance of images and text that ChatGPT has ingested, for better or worse.

Considering bias in AI is an important part of crafting classroom policy, integrating AI, and engineering prompts that return information that is as unbiased as possible. AI-generated images often illustrate the biases more starkly, but they are just as prevalent—and problematic—in text.

Some readings to learn more:

• Humans Absorb Bias from AI—And Keep It after They Stop Using the Algorithm

• Shedding Light on Ai Bias With Real World Examples

• There’s More to AI Bias Than Biased Data, NIST Report Highlights