XaiJu
Yannick Trapman-O'Brien
Yannick Trapman-O'Brien

patreon


March Reading; “Skipping the Line”

A few strange occurrences have had me back on the topic of algorithms and organ donations lately - both of which I think of as gateways to larger questions about how we as individuals, groups, and a society make difficult choices. So when I saw that the NYTimes has finished up a major piece on the Organ Transplant System, it felt like yet another sign that it’s time to pick up these threads again.

Organ Transplant System ‘in Chaos’ as Waiting Lists Are Ignored

Brian M. Rosenthal, Mark Hansen and Jeremy White

On its face, this is a piece of investigative journalism about the growing tendency for officials allocating donated organs to “skip” those at the top of the waiting list. Scrolling through the testimonials and interactive graphics reveals a complicated web of competing incentives (along with the expected Scooby-Doo twist of pulling off the mask and revealing capitalism as a culprit). But this article also struck with me as an example of how thorny it can be to have algorithms and humans share decision-making power, and as a case study for the way well-intended systems in the United States can gradually buckle under pressures, distort, decay, and quickly outstrip the capacity of whatever regulatory body may exist.

Something I appreciate: in addition to a “methodology” section at the bottom of the piece, the times also published some reporting on their process which I found clarifying.

Related Readings

episode 27:  these data nerds are fixing a broken organ donation system & (literally) saving lives

System Catalysts Podcast, 2024

On reading the comments on this NYTimes piece (both on their official page and on reddit - let know one say I have not suffered for my art), one striking thing that was often referenced by people claiming to be in the Organ Donation system was the change that regulations had made in the system.. The article mentions this time specifically:

“In 2020, procurement organizations felt under attack. Congress was criticizing them for letting too many organs go to waste. Regulators moved to give each organization a grade and, starting in 2026, fire the lowest performers.”

In a later chart, it’s hard not to notice a possible correlation in the percentage of organs being allocated out of sequence (and to see seemingly little progress on the percentage of organs discarded).

I wanted to learn a bit more about this regulation, and stumbled across this podcast which tells this story from the regulator’s angle. The language of this podcast is unabashedly optimistic about systemic change (and if I was being more particular, I’d say vibe combined with the dog whistle of “government contractors” speaks to me of a Thayer-style “Nudge” incrementalism), which makes it an interesting setting to hear the guests describe the issue they hoped their legislation would resolve, even as we hold in mind the unintended consequences we as readers already know to be playing out. How do we as a society balance our concerns for fairness and efficacy?

… which reminds me of a chart I once saw…

A Human's Guide To Machine Intelligence: Chapter 9; Inside the Black Box 

by Kartik Hosanagar


… which in turn got me re-reading this oft-referenced book, and led me to a differently chapter entirely - Chapter 9. Hosanagar talks about degrees of transparency and disclosure in how, why, and the trade-offs used when algorithms make decisions, and the ways each of these affected the trust-levels and sentiments of users.

Naturally, it’s a bit mad to read an entire book in order to access what is essentially an amateur footnote on a news article—and I strongly encourage you to do so. That said, our excerpt this month will involve me reading across a few of these sources, so if you just wait enough days, I’ll do it for you.

Comments

A few years ago I fell down a rabbit hole on algorithmically-set bail. The program seemed well-intentioned — they proposed that by feeding in the actuarial data, the judiciary could set bail for the accused without the biases of a single person entering in. Unfortunately, though unsurprisingly, the algorithm literally codified the systemic bias, without the accountability a human judge would have to own for their record. Gross, but at least at the time these algorithms were still observable. As learning models evolve ever more into black boxes, the choice of trust is forced into a binary of use/abstain, and it becomes difficult or impossible to regulate nuance.

Lyra Levin


More Creators