When forensic linguist Claire Hardaker’s team sent off a grant proposal on the risks of trusting artificial intelligence (AI) models in voice analysis, the Lancaster University researchers assumed it would be read by at least one person knowledgeable about the topic.

So the rejection email Hardaker received last week came as a shock. The proposal “was not selected to progress beyond the AI triage stage,” it read. A collection of AI models had evaluated 179 proposals, CRANE’s communication explained – and discarded the bottom 50% before “full human review.” More details about how the procedure actually worked would be shared with applicants at a later date, it added.

The rejection email, which Hardaker and other unsuccessful applicants shared on social media last week, has prompted outrage among many researchers, who worry about the fairness of AI-based assessment and about the potential for privacy leaks if AIs are fed unpublished ideas.


Source https://www.science.org/content/article/u-k-funder-uses-ai-triage-reject-grant-proposals-leaving-researchers-dismayed

Another high-profile case has interested scientists. Read now