AI shows clear racial bias when used for job recruiting, new tests reveal
AI has yet to solve hiring discrimination, and it might be making it worse.
Via Not A Bot
“Could AI BE any more biased? |
| Now read that title again but in your best Chandler Bing voice 🙂 |
Diversity issues in AI have long been a concern. A recent Bloomberg experiment seems to confirm it ️ |
| GenAI technology, specifically GPT-3.5, exhibits racial biases in job recruitment. |
| The study method involved systematically assigning names associated with specific racial and ethnic backgrounds to resumes of equal qualification. |
| AI’s task? To rank these resumes for various job openings. |
| The results? Disconcerting. GPT-3.5 showed a clear preference for names from specific demographics, failing standard benchmarks for job discrimination. |
| Of particular concern were: |
|
| In response to these findings, OpenAI claimed that these results do not reflect their clients’ typical use of their software, many of whom refine responses to reduce bias. |
| But regardless, the investigation, backed by insights from over 30 experts across various fields, emphasizes the crucial ethical issues of heavily relying on AI without addressing its inherent biases.” |
️
0 Responses
Stay in touch with the conversation, subscribe to the RSS feed for comments on this post.