Lightnews — Scholar-powered news

Fernando Diaz

@841io.bsky.social

1.7K followers 610 following 84 posts

Associate Professor, CMU. Researcher, Google. Evaluation and design of information retrieval and recommendation systems, including their societal impacts.

Posts Media Videos Starter Packs

Reposted by Fernando Diaz

arXiv cs.HC Human-Computer Interaction @cshc-bot.bsky.social · 14h

Renee Shelby, Fernando Diaz, Vinodkumar Prabhakaran: Taxonomy of User Needs and Actions https://arxiv.org/abs/2510.06124 https://arxiv.org/pdf/2510.06124 https://arxiv.org/html/2510.06124

Reposted by Fernando Diaz

Dylan Foster 🐢 @djfoster.bsky.social · 6d

MSR NYC is hiring spring and summer interns in AI/ML/RL!

Apply here: jobs.careers.microsoft.com/global/en/jo...

Microsoft Research Lab - New York City - Microsoft Research

Apply for a research position at Microsoft Research New York & collaborate with academia to advance economics research, prediction markets & ML.

www.microsoft.com

Fernando Diaz @841io.bsky.social · 6d

Lots of adjacent conversations in NLP [1] and vision [2] and probably more I'm overlooking. Excited to see this community continue to grow and (hopefully) congeal. 3/3

[1] c3nlp.github.io
[2] sites.google.com/view/ec3v-cv...

Fernando Diaz @841io.bsky.social · 6d

This is the most recent iteration of a series of workshops that we have been co-organizing, starting at NeurIPS 2022 (ai-cultures.github.io), where I presented this slide, anticipating the next few years (with apologies to Moritz, who I think originated the fairness version it's based on). 2/3

Fernando Diaz @841io.bsky.social · 6d

In January, Asia Biega (MPI), Georgina Born (UCL), Mary Gray (MSR), Rida Qadri (G), and I ran a Dagstuhl Seminar bringing together folks from CS and the broader social sciences to discuss questions around AI and culture. Dagstuhl has just posted our report, 1/3

drops.dagstuhl.de/storage/04da...

Fernando Diaz @841io.bsky.social · 8d

Chapter 1 of Porter's "Trust in Numbers" touches on reproducibility, though outside of the context of the crisis.

Reposted by Fernando Diaz

Alexandra Olteanu @aolteanu.bsky.social · 8d

This was accepted to #NeurIPS 🎉🎊

TL;DR Impoverished notions of rigor can have a formative impact on AI work. We argue for a broader conception of what rigorous work should entail & go beyond methodological issues to include epistemic, normative, conceptual, reporting & interpretative considerations

Alexandra Olteanu @aolteanu.bsky.social · Jun 18

We have to talk about rigor in AI work and what it should entail. The reality is that impoverished notions of rigor do not only lead to some one-off undesirable outcomes but can have a deeply formative impact on the scientific integrity and quality of both AI research and practice 1/

Print screen of the first page of a paper pre-print titled "Rigor in AI: Doing Rigorous AI Work Requires a Broader, Responsible AI-Informed Conception of Rigor" by Olteanu et al. Paper abstract: "In AI research and practice, rigor remains largely understood in terms of methodological rigor -- such as whether mathematical, statistical, or computational methods are correctly applied. We argue that this narrow conception of rigor has contributed to the concerns raised by the responsible AI community, including overblown claims about AI capabilities. Our position is that a broader conception of what rigorous AI research and practice should entail is needed. We believe such a conception -- in addition to a more expansive understanding of (1) methodological rigor -- should include aspects related to (2) what background knowledge informs what to work on (epistemic rigor); (3) how disciplinary, community, or personal norms, standards, or beliefs influence the work (normative rigor); (4) how clearly articulated the theoretical constructs under use are (conceptual rigor); (5) what is reported and how (reporting rigor); and (6) how well-supported the inferences from existing evidence are (interpretative rigor). In doing so, we also aim to provide useful language and a framework for much-needed dialogue about the AI community's work by researchers, policymakers, journalists, and other stakeholders."

Fernando Diaz @841io.bsky.social · 13d

very nice! we found that a large fraction of users literally read w their cursor and/or highlight paragraphs as they read.

Fernando Diaz @841io.bsky.social · 14d

[5] F Diaz, Q Guo, R White. Search result prefetching using cursor movement. SIGIR 2016.
[6] R White, F Diaz, Q Guo. Search result prefetching on desktop and mobile. TOIS 2017.

Fernando Diaz @841io.bsky.social · 14d

[3] P Metrikov, F Diaz, S Lahaie, J Rao. Whole page optimization: how page elements interact with the position auction. EC 2014.
[4] D Goldstein, S Suri, R P McAfee, M Ekstrand-Abueg, F Diaz. The economic and cognitive costs of annoying display advertisements. Journal of Marketing Research, 2014.

Fernando Diaz @841io.bsky.social · 14d

[1] M C Chen, J Anderson, M H Sohn. What can a mouse cursor tell us more?: correlation of eye/mouse movements on web browsing. CHI 2001.
[2] F Diaz, R White, G Buscher, D Liebling. Robust models of mouse movement on dynamic web search results pages. CIKM 2013.

Fernando Diaz @841io.bsky.social · 14d

11/10 Given the impact on search evaluation/optimization, I'm excited to see how this develops w new conversational interfaces. 11/10

Fernando Diaz @841io.bsky.social · 14d

A nice thing about cursor/viewport tracking (and other implicit signals) is that it's less prone to self-selection bias present in explicit feedback (e.g., thumbs up/down) and available from almost all users, if done right. "Done right", ofc, means that you're rigorous in who and how you log. 10/10

Fernando Diaz @841io.bsky.social · 14d

Around the same time, folks like Ryen White, Jeff Huang, Eugene Agichtein, Qi Guo, and Vidhya Navalpakkam were exploring cursor tracking for a broad range of use cases. 9/10

Fernando Diaz @841io.bsky.social · 14d

When mouse is unavailable (e.g., on mobile), you can use viewport [6]. 8/10

Fernando Diaz @841io.bsky.social · 14d

and reduce latency [5]. 7/10

Fernando Diaz @841io.bsky.social · 14d

study effects of visual presentation in advertising [3,4], 6/10

Fernando Diaz @841io.bsky.social · 14d

You can build models of visual attention [2], 5/10

Fernando Diaz @841io.bsky.social · 14d

Depending on the interface, there can be a strong correlation between eye fixation and cursor position (on search, it's about 100 pixels) [1]. This means you can use a baby amount of js to get a cheap, noisy eyetracker. With enough users, that's not too bad. 4/10

Fernando Diaz @841io.bsky.social · 14d

Let's discuss a stream of work I initiated at Yahoo Research and continued at MSR, focusing on behavioral signals derived from cursor and viewport information. 3/10

Fernando Diaz @841io.bsky.social · 14d

On the one hand, pausing rendering can lead to big cost savings if you're saving on inference or bandwidth. At the same time, it would be A Good Idea to use scrolling and other behavioral feedback for evaluation and optimization. i.e., viewport is a pretty good engagement signal. 2/10

Fernando Diaz @841io.bsky.social · 14d

I'm noticing that some conversational AI interfaces are pausing text generation (or at least rendering) until the user scrolls. A thread on attention modeling. 👀 🧵 1/10

Reposted by Fernando Diaz

Miro Dudik @mdudik.bsky.social · 20d

🚨Microsoft Research NYC is hiring🚨

We're hiring postdocs and senior researchers in AI/ML broadly, and in specific areas like test-time scaling and science of DL. Postdoc applications due Oct 22, 2025. Senior researcher applications considered on a rolling basis.

Links to apply: aka.ms/msrnyc-jobs

Microsoft Research Lab - New York City - Microsoft Research

Apply for a research position at Microsoft Research New York & collaborate with academia to advance economics research, prediction markets & ML.

Fernando Diaz @841io.bsky.social · 27d

yeah i agree. for me, one question is, given that "agent" as a term that is hard to pin down, are there aspects of models identified in the paper (see "scaffolding") that we can point to for more precision about what is happening?

Fernando Diaz @841io.bsky.social · 27d

it may be worth talking about these dimensions more or being specific about what we mean when we say agentic wrt them.