What are the worst ethical disasters in NLP history?
(I'm teaching "ethics of NLP" tomorrow and history is good for teaching this topic.)
Most are data breaches/releases (AOL search logs, OKCupid profiles, Finnish therapy records...) but what others?
I'll put some other examples in thread --> 1/n
Very related to what a lot of other folks on the ATProto are building and imagining, e.g. bsky.app/profile/semb...
And use the demo here!
bsky.app/profile/kenn...
See Kenny's post for his vision for a more exploratory social media! He's articulating (and building!) an inspiring possibility of how modern language technologies can support as opposed to degrade human understanding
The Tech Team at ACLU is hiring! We are looking for a Data Scientist with expertise in NLP and AI ethics to work on using language tech to support ACLU's mission. Come help us tackle questions about how AI systems can be carefully applied to support the public interest. www.aclu.org/careers/appl...
Brief fun survey from Jessica, Andrew & myself:
If you are a faculty member, research scientist, postdoc, or senior Ph.D. student in any area of science, please take five minutes and fill it out. We’ll share the results widely along with some reflections.
docs.google.com/forms/d/e/1F...
@sjgreenwood.bsky.social would have more recent suggestions but when we were starting, we forked this repo github.com/MarshalX/blu.... We walk through architecture + code here! arxiv.org/abs/2601.04253
you can build this on bluesky!
RCV has been used for over a century, reaching over ~14M US voters.
Yet theoretically, it is vulnerable to strategic manipulation. The catch? Showing the strategy space empirically is computationally difficult.
We designed algorithms to do so and applied to over 100 elections. New paper in JCSS!
I just did this as a pre-submission checklist item for my own paper, and it was very useful! AI is quite good at this.
My takeaway from Atmosphere: it needs to be WAY easier for builders to get paid
So I built at.fund
at.fund detects the tools you use, and tells you how to fund the builders who make those tools. And if you're a builder, it takes seconds to add records to that tell your users how to fund you 👍
I regret missing #atmosphereconf but would love attendees to check out and give feedback on our labeler at @stechlab-labels.bsky.social -- a first-pass implementation looking at how labelers can improve the surfacing of transparent, objective, useful ATProto information about account behavior.
@sjgreenwood.bsky.social is!
I'll add that this demo illustrates the possibilities of the AT Proto -- open APIs, and encouraging people to build new interfaces #atproto #AtmosphereConf
www.skytrails.org
Allowing people to "browse" -- jumping from topic to related topic -- is an old, founding idea in information storage/retrieval and the internet. The new insight is that language models are good enough to *build* these trails at scale, to enable human understanding.
Led by @kennypeng.bsky.social, we've been working on a new implementation of an old vision, to go beyond "feeds" to "interconnected trails". Demo on Bluesky/ATProto posts over a week!
www.skytrails.org
Pleased to share our new paper forthcoming in @icwsm.bsky.social! We introduce a novel framework to measure value expressions in social media posts at scale, leveraging personalization to handle the inherent subjectivity of human values.
arxiv.org/abs/2511.08453
Interesting paper and thread on interdisciplinarity
Ah interesting! Thanks for sharing
Fun! And I know my students have appreciated having you around :)
At some point, would love to catch you to understand the difference between doctrine and policy...
Really glad to see @gsagostini.bsky.social recognized, he's a great urban data scientist!
agreed! variance is just quite high with a phd across the board
yeah, it's high variance (many of my phd friends clearly did come out ahead!) and if it's also the path to immigrate to the US, then it also might be positive expected value!
This is great!
@debunkbot.bsky.social is this true? are you live on bluesky?
Yeah, maybe. Looking at is thinking traces it seemed like it was trying multiple things like direct fetch and headless browser. I was using it for something similar, trying to grab a set of cited papers in a paper so I could read
Huh. Codex has been pretty good for me on this. Working with the Zotero MCP to put it in there, on the other hand, not so good.
Worthwhile new essay "Mathematicians in the Age of AI" by Jeremy Avigad, CMU professor and director of the NSF Institute for Computer-Aided Reasoning in Mathematics (ICARM) at CMU: www.andrew.cmu.edu/user/avigad/...
A picture of Joe Halpern smiling in green shirt in front of a blue background.
Today arXiv remembers our colleague Joe Halpern, who was instrumental in founding arXiv's CS section.
Joe's passions ranged far & wide and we're lucky that arXiv was one of them. Joe, thank you for giving so much to arXiv - you are missed.
blog.arxiv.org/2026/02/27/remembering-joe-halpern
