Drew Dimmery

Another blog post: separatedhyperplane.substack.com/p/we-cant-le... We can't let frontier labs choose their own auditors - or their own questions. Triangulating between the near-past lens of algorithmic accountability, the financial crisis, and the AI evaluation landscape as it looks now:

AI evaluation is all the rage these days. I'm excited to be teaching a class on it this Fall; I wrote up how I'm thinking about it: ddimmery.com/posts/ai-eva...

🚨New Science Advances🚨 Untrustworthy sources on Facebook and Instagram in 2020: Concentrated exposure but no attitudinal effects www.science.org/doi/10.1126/... -Exposure very low but highly concentrated -Reducing exposure by ~70% for months had no measurable effects *even among frequent consumers*

Last monday we published our paper on AI Labels and a potential spillover effect on unlabeled posts on platforms such as Instagram in the ICWSM proceedings. If you are interested in our work you can now find it here: ojs.aaai.org/index.php/IC... Huge thanks to my co-authors @ddimmery.com and Pu Yan!

The Hertie School Data Science Lab is hiring a postdoc. This is an appealing position with fairly low teaching/admin responsibilities and a lot of freedom to develop your research agenda. Feel free to reach out if you want to know more! hertie-school.dvinci-easy.com/en/p/en/jobs...

Who produces hate speech? And how does that matter for content moderation? We show that across different countries and platforms, a relatively small share of users are responsible for a very large share of hate - overall, 5% write 83-100% of hateful content. www.cambridge.org/core/journal...

When you see "AI-generated" on Instagram, does it make you more skeptical – or more trusting of everything without a label? New research from German Instagram users has answers for policymakers pushing for AI labeling laws. Good news and a warning👇

Is online discourse getting worse? To bring actual data to bear on this question, we launched the Public Discourse Indicator, a dashboard tracking online comments submitted to several major Swiss newspapers. Our aim here is twofold: www.public-discourse.org/en/public-di...

🚨My first preprint is out on @socarxiv.bsky.social! How do AI-generated content labels shape what people see as authentic on social media — and do labels have unintended side effects? osf.io/preprints/so... A thread 🧵

Join us next week for the Brown Bag Seminar! We’re excited to host Professor @ddimmery.com for a discussion on how to design studies that preemptively address challenges, improving inference and generalization before data collection even begins. Register 👉 www.hertie-school.org/en/datascien...

The Experiments Section Newsletter is BACK! Latest issue just dropped: what do we actually do while conducting experiments, and how do we handle unexpected events?

The end of 538 is a huge shame - both for the incredible people who worked there, and for political and data journalism as a whole. I was lucky enough to work beside them for a few years and want to say a bit about what I think was so valuable that I hope doesn't vanish from the media landscape: 🧵

RCTs are good, but not (just) because they make the statistics easy.

Interested in how people talk to each other online? Care about causal inference and/or NLP? Want to design and implement field experiments? Come do a PhD with Dominik Hangartner, me, and a bunch of awesome people at IPL in Zurich: jobs.ethz.ch/job/view/JOP...

The US2020 project was not independent (and that's good) open.substack.com/pub/drewdimm...

I riffed on Ivan Illich, Smalltalk and This Place. open.substack.com/pub/drewdimm...

A thread about being wrong: 5 years ago, we wrote a paper about how how newly enfranchised 16-year-olds vote in Austria. But we were wrong. This year, @elisabethgraf.bsky.social, @schnizzl.bsky.social, Sylvia Kritzinger and I are setting the record straight: authors.elsevier.com/c/1juT5xRaZk...

Welcome to Hertie, @ddimmery.com ! 🤝  The Professor of Data Science for the Common Good begins teaching today. "Algorithms, code and machine learning are an increasingly large part of how our world works. I'm excited to empower students to understand these topics!" @hertiedatascience.bsky.social

Using the reframing of "the Algorithm" to "the Apparatus" should shape how we evaluate what it is/does open.substack.com/pub/drewdimm...

Continuing with the controversial takes, in this post I argue that folks should plot their data (based on a new paper at ISR with Yan Leng) open.substack.com/pub/drewdimm...

Got a hot take: open.substack.com/pub/drewdimm...

Does anyone know of a good intellectual history of ML/AI? Something that doesn't just survey the development of methods, but includes things like "Papert and Minsky think solving computer vision will be an undergrad summer project in the 60s"

Your weird linear algebra fact of the day: b_ols = (X'X)^{-1} X'y = X^+ y where X^+ is the Moore-Penrose inverse and X'X is invertible.

Really nice post by Kevin about the (insane) techno-optimist manifesto. There's also an implicit meta-level critique about modes of thinking that's worth considering, too: the Manifesto was basically a tweet-thread; Kevin's post is not. open.substack.com/pub/kevinmun...

Wrote some thoughts on microblogging (derogatory)

Meanwhile, in Austria, US soft power is undefeated www.politico.eu/article/let-...

My coauthor Drew Dimmery (formerly of Meta) on how the 2020 US Facebook and Instagram Election Project offers a template for reporting by social media platforms under future regulation - recommended

I finally got around to posting a package to PyPi (https://pypi.org/project/bwd/). Geez, the comparison to the CRAN process is like night and day. Zero worrying about a petty tyrant yelling at me for not following some abstruse and poorly documented procedures.