There is a lot of interest in estimating LLMs' uncertainty, but should LLMs express uncertainty to end users? If so, when and how?

In our #FAccT2024 paper, we explore how users perceive and act upon LLMs’ natural language uncertainty expressions.

arxiv.org/abs/2405.00623

1/6

Super honored and excited that my work on the role of tech worker positionality in shaping identity concepts in computer vision was awarded a best paper 🤯

There are many good uses for AIs in social science research, but also many clear cases when they are inappropriate as well.

This paper shows how LLMs struggle with portraying complex individual identities (like race or gender), suggesting important limitations for research use.

Anyone tempted to think AI Surrogates can replace human participants in experiments needs to read this devastating empirical critique by Angelina Wang & colleagues arxiv.org/abs/2402.01908

I'm thrilled to share that our #CSCW2023 paper has been awarded Recognition for Contribution to Diversity and Inclusion 🥳! Our work explores the experiences and challenges of Black TikTok creators and affordances that support inclusive online platforms: dl.acm.org/doi/10.1145/36…

What visual cues are correlated with gender in image datasets? Basically everything!

In our #ICCV2023 work, we explore where gender artifacts arise in visual datasets, using COCO and OpenImages as a case study.

ML-based science is facing a reproducibility crisis. We think clear reporting standards for researchers can help.

Today, we're introducing REFORMS, a consensus-based checklist authored by 19 researchers across many disciplines.
aisnakeoil.com/p/introducing-…

Check out this thread about our new project on challenging the legitimacy of a category of automated decision making system that uses machine learning to predict future human behavior!

Can machine learning improve algorithmic decision-making? Developers of ML-based algorithms have made tall claims about their accuracy, efficiency, and fairness. In a systematic analysis, we find that these claims fall apart under scrutiny. …dictive-optimization.cs.princeton.edu

Governments, banks, employers, and many other institutions make decisions about us by using machine learning to predict our future behavior. In a new paper, Angelina Wang, Sayash Kapoor, @s0l0n and I challenge the legitimacy of this type of decision making. 🧵

…dictive-optimization.cs.princeton.edu

And here’s Solon Barocas moonlighting at a philosophy workshop :) ‘against predictive optimisation’—what’s wrong with automating the process of developing decision-making rules work w Angelina Wang Arvind Narayanan and Sayash Kapoor

Anyway, check out the post for many more examples. aisnakeoil.substack.com/p/the-bait-and… We have an extensive analysis of this dubious class of applications in an upcoming paper titled 'Against Predictive Optimization' with Angelina Wang & Solon Barocas. papers.ssrn.com/sol3/papers.cf…

Cool write-up about ICML vs FAccT.
Agree w this: 'At FAccT, it felt like there was roughly gender parity...At ICML, it felt like over 80% of the attendees were masculine-presenting. Personally I felt that this contributed quite significantly to the ~vibe~ of the conferences.'

This summer I went to my first two in-person conferences in grad school, FAccT and ICML, and you’ll never believe what happened next (spoiler: my advisor suggested I write a blog post about it) medium.com/@angelinaaa/a-…