vol 1 (6/28-7/12)

let’s get started!

Some of the best and most relevant items from the last two weeks:

flimsy but existent Biden EO on reproductive privacy + rights (story by Tonya Riley), follow up blogpost from FTC on protecting some digital privacy and user data, and a blog post on Privacy post Dobbs by my colleagues @EPIC Sara Geoghegan and Dana Khabbaz (July 8-11ish)

Kian Vesteinsson @kianvest

Great to see the President directing FTC and HHS to protect the data privacy of those seeking reproductive healthcare. But this shouldn't require rapid response. Dobbs is a historic attack on rights, with worse to come. We need a federal data privacy law.

whitehouse.govFACT SHEET: President Biden to Sign Executive Order Protecting Access to Reproductive Health Care Services | The White HouseTwo weeks ago, the Supreme Court issued a decision that overturned Roe v. Wade and eliminated a woman’s Constitutional right to choose. This

---

Chris Gilliard (@hypervisible on TW) published School Surveillance Will Never Protect Kids From Shootings on WIRED (June 30)

---

there’s another shitty predictive policing program, which claims to be able to predict crime ahead of time. Bloomberg had a story that communicated. this was my take, and they updated it (jul. 3)

Ben Winters @Wenbinters

this is a reckless and nonsense article not sure if the headline, the lede, or the absolutely inaccurate framing of “in contrast, more data is used to support the prediction which makes it…better?” Is the worst part also, US govt, stop funding this garbage

Surveillance Killjoy @hypervisible

Here we fucking go again. https://t.co/yZv4EruI34

I don’t necessarily recommend reading that, but i can recommend some great reading on predictive policing in general: From Brennan; “Predictive policing algorithms are racist. They need to be dismantled” by Will Douglas Heaven; and Dirty Data, Bad Predictions by Rashida Richardson, Jason Schultz, and Kate Crawford

---

great story by Khari Johnson on WIRED exploring some interesting parts of the EU AI Act, which has some infrastructure for a risk-based regulation, but the devil is in the details on what gets banned, what triggers higher obligations etc…Huge to watch for AI reg frameworks that are “risk-based” — who gets to decide what is what level of risk? (6/30)

---

incredibly timely empirical study and recs from algo harm allstar team at the Algorithmic Justice League of Sasha Costanza-Chock, Deb Raji, and Joy Buolamwini on the state of the algorithmic auditing ecosystem (published June 2022)
Who Audits the Auditors? Recommendation from a field study of the algorithmic auditing ecosystem — which debuted at the ACM conference on Fairness Accountability and Trust. lots of great reading always come from that conference — and they always come in 12-15 page doses

another incredible FaCCT (future issues will highlight other papers from the conference)

-the fallacy of AI functionality by Deb Raji, Elizabeth Kumar, Aaron Horowitz, and Andrew Selbst, including this really helpful “failure taxonomy”

---

on period trackers in the wake of Dobbs: Samantha Cole of Motherboard surveyed popular apps on their data policies. Really helpful but few specific new promises, and all from the companies themselves (june 28)

---

the AP and Motherboard reported about Meta (IG and FB) removing posts and flagging accounts of members that offered to order legal abortion pills for individuals in states where abortion was banned (june 28)

the speed with which this was taken down brings to mind Facebook’s overbroad claims of the use of AI to detect hate speech, TOS violations, and more — IMO, this is a harmful and BS claim, with these anecdotes showing automated takedowns are not necessarily warranted or correct takedowns

Ben Winters @Wenbinters

Meta spokesperson at AI in content mod panel at #CPDP2022 bumps claims that their hate speech AI improved substantially because AI now flags ~95% of their hate speech auto takedowns up from 80% — a claim they’ve been making for a while. this⬆️ % doesn’t mean improvement (1/2)