ChatGPT Replicates Gender Bias in Recommendation Letters

Trouble viewing? View in your browser. View all Scientific American publications.

    

November 28, 2023

In order to spit out human-sounding sentences, large language models must be trained on oodles of human writing. As a result, they're also trained on the inadvertent biases that we have encoded in our words throughout history. That's a problem when AI chatbots like ChatGPT are used to, say, write recommendation letters for employees. A new study found that bots were more likely to describe men with terms like "expert" and "integrity," and women with words like "warm" and "emotional." Using such AI tools in the office could entrench existing gender bias—and thus aggravate the discrimination women already face in the workplace.

Sophie Bushwick, Associate Editor, Technology

Inequality ChatGPT Replicates Gender Bias in Recommendation Letters A new study has found that the use of AI tools such as ChatGPT in the workplace entrenches biased language based on gender By Chris Stokel-Walker

Policy Firearm Forensics Has Proven Reliable in the Courtroom. And in the Lab Despite criticism, a slate of new scientific studies show that forensic firearms analysis is a reliable scientific discipline that the criminal justice system should trust By Raymond Valerio,Nelson Bunn

Artificial Intelligence When It Comes to AI Models, Bigger Isn't Always Better Artificial intelligence models are getting bigger, along with the data sets used to train them. But scaling down could solve some big AI problems By Lauren Leffer

Climate Change Air-Conditioning Discovery Eliminates Harmful Gases Heat pumps are ubiquitous in the form of air conditioners. Scientists just invented one that avoids harmful refrigerant gases By Davide Castelvecchi,Nature magazine

Behavior It's Not All in Your Head--You Do Focus Differently on Zoom Virtual meetings and video calls don't quite stack up to in-person interaction—and a new study proves it By Lauren Leffer

Defense What Would It Mean to 'Absorb' a Nuclear Attack? The missiles on the Fort Berthold Reservation in North Dakota make it a potential target for a nuclear attack. And that doesn't come close to describing what the reality would be for those on the ground. By Ella Weber | 19:05

Culture The Science to Be Grateful for This Year A year of exciting ideas and research has given us much to be grateful for By Lori Youmshajekian

Robotics Robotics 'Revives' a Long-Extinct Starfish Ancestor Engineers and paleontologists teamed up to reconstruct an ancestor of starfish from the Paleozoic era and figure out how it moved By Lauren Leffer

Policy AI Needs Rules, but Who Will Get to Make Them? Skirmishes at the U.K.'s AI Safety Summit expose tensions over how to regulate AI technology By Chris Stokel-Walker

Defense The Members of This Reservation Learned They Live with Nuclear Weapons. Can Their Reality Ever Be the Same? The Mandan, Hidatsa and Arikara peoples are learning more about the missiles siloed on their lands, and that knowledge has put the preservation of their culture and heritage in even starker relief. By Ella Weber | 14:55

Evolution Our Evolutionary Past Can Teach Us about AI's Future Evolutionary biology offers warnings, and tips, for surviving the advent of artificial intelligence By Eliot Bush

QUOTE OF THE DAY "The AI authors' writing often sounds like it was written by an alien; one Ortiz article, for instance, warns that volleyball 'can be a little tricky to get into, especially without an actual ball to practice with.'" Maggie Harrison, Futurism

FROM THE ARCHIVE Humans Absorb Bias from AI--And Keep It after They Stop Using the Algorithm People may learn from and replicate the skewed perspective of an artificial intelligence algorithm, and they carry this bias beyond their interactions with the AI By Lauren Leffer | October 2023

LATEST ISSUES

Questions?   Comments?

Download the Scientific American App

To view this email as a web page, go here. You received this email because you opted-in to receive email from Scientific American. To ensure delivery please add newsletters@scientificamerican.com to your address book. Unsubscribe     Manage Email Preferences     Privacy Policy     Contact Us

Scientific American

1 New York Plaza, FDR Dr, Floor 46, New York, NY 10004

Unsubscribe - Unsubscribe Preferences