Data ethics readings

Presentations: 35 minutes

• Create a 1 to 1.5 page handout
• Summarize the reading (~15 minutes) and lead discussion by posing 3–4 discussion questions (~20 minutes)
• Informal: Bullet-points are fine, feel free to include screen-grabs of (e.g.) important plots
• You can take a break from Markdown—feel free to write up a Word doc and generate a PDF!
• Share the handout on Class-Exercise-Repo/data-ethics/ as (e.g.) koenecke_handout.pdf

For the articles you’re not presenting, just skim them (like, set a timer for 10 minutes per article and don’t go over it) and pose a muddiest point or follow-up question. Add your MP or follow-up to Class-Exercise-Repo/data-ethics/ as (e.g.) bender_muddiest_Dan.md. Feel free to skip the ASR bias articles (we’ll be discussing those the same day the midterm is due).

ASR bias

• Katherine: Koenecke et al. 2020
• Dan: Markl 2022 (paper, video)

Language models

• Gianina: Bender, Gebru, McMillan-Major, & “Schmitchell” 2021
  • This paper eventually led to Google firing Timnit Gebru and Margaret Mitchell; Timnit Gebru went on to launch the Distributed Artificial Intelligence Research Institute. This is what Sen’s articles are about
• Sen: Hao 2021, Perrigo 2022
  • Let me know if you hit a paywall with the Time article

Accountability to communities

• Soobin: Medina 2022
• Mack: Sneller, Wagner, & Ye 2022