Emma Harvey
@emmharv
PhD student @CornellInfoSci | Sociotechnical fairness & algorithm auditing | Previously @MSFTResearch FATE, @Penn | Also http://emmharv.bsky.social
🌟Applying for a PhD @CornellInfoSci? Our student-run application feedback program is open NOW through Oct 31🌟 We're open to all and especially hope to support scholars who have navigated barriers in higher ed - please share to help us broaden access! infosci.cornell.edu/phd/admissions
Researchers @CornellInfoSci find that ignoring race in college admissions leads to an admitted class that’s much less diverse, but with similar academic credentials to those where affirmative action is factored in. @whynotyet @cornell_tech @TheOfficialACM news.cornell.edu/stories/2024/1…
Gaps Between Research and Practice When Measuring Representational Harms Caused by LLM-Based Systems: arxiv.org/abs/2411.15662 TL;DR: We identify challenges to using publicly available measurement instruments for measuring representational harms. Led by our intern @emmharv.
In the final presentation session, a team from @Cornell examines the impacts of ending #affirmativeaction on college admissions. Their findings reveal that removing race from applicant ranking algorithms harms diversity without increasing academic merit. #HigherEd 👨🎓
Many thanks to @garanceburke & @HilkeSchellmann for covering our work and raising awareness about the potential harms of speech-to-text hallucinations in downstream applications! apnews.com/article/ai-art… For a refresher on our paper's findings, see🧵below:
🎷Excited to present our paper, “Careless Whisper: Speech-to-text Hallucination Harms” at @FAccTConference! 🎷We assess Whisper (OpenAI’s speech recognition tool) for transcribed hallucinations that don’t appear in audio input. Paper link: arxiv.org/abs/2402.08021, thread 👇
Researchers say an AI-powered transcription tool used in hospitals invents things no one ever said apnews.com/article/ai-art…
📢📢📢 Do you work on LLM model/system evaluation, including measuring harms, adverse impacts, or other possible undesirable LLM behaviors? Please consider participating in this paid interview study led by @emmharv !👇🙏 #NLProc #ACL2024 #EMNLP2024 x.com/emmharv/status…
🔔 Are you an AI practitioner whose work involves evaluating LLM-based systems for harms, adverse impacts, or other undesirable behaviors? Consider signing up for a (paid!) @MSFTResearch interview: forms.office.com/r/JBjhDRnaLY @ewsheng @hannawallach @sulin_blodgett @o_saja @achould
🔔 Are you an AI practitioner whose work involves evaluating LLM-based systems for harms, adverse impacts, or other undesirable behaviors? Consider signing up for a (paid!) @MSFTResearch interview: forms.office.com/r/JBjhDRnaLY @ewsheng @hannawallach @sulin_blodgett @o_saja @achould
I am in Rio for #FAccT2024!! If you are too, let’s hang out!
I’m officially in Seattle for the summer to intern at @MSFTResearch #FATE, where I’ll be working on bridging research and practice to measure harms from LLMs with the inimitable @ewsheng! If you’re also here, LMK - I’m excited to explore and I’d love to grab coffee, hike, etc!!
🎉 “The Cadaver in the Machine” has received a Best Paper Honorable Mention at #CHI2024!! 🎉 @haukesa @az_jacobs @MannyMoss @mona_sloane
So excited to share that "The Cadaver in the Machine: The Social Practices of Measurement and Validation in Motion Capture Technology" with @haukesa @az_jacobs @MannyMoss @mona_sloane is conditionally accepted to #CHI2024 (and you can read it now on arXiv)!!
I’ll be working from Pittsburgh (my former home!) next week - if you are there and want to get a coffee and/or chat about algo fairness etc., please DM me!! Firetree pic attached because it is my favorite ~Pittsburgh thing~
