
despite all my rage, I am still just a statistically probable sequence of observable online behaviors

mostly not here anymore because Elon ruined it. unclear why folks are here contributing to his weird distortions, but alas many are.

despite all my rage, I am still just a statistically probable sequence of observable online behaviors

BEST panel I ever attended at a conference! Great format and fantastic discussion!
Excited for our #CSCW2024 panel at 2:30 pm today in the Heredia room - "Is Human-AI Interaction CSCW?" This will be presented as a formal debate with me and @msbernst competing against @andresmh and @jeffbigham

somehow I hadn’t noticed that TSA wait times at @PITairport are powered by Zensors! This was originally based on a couple of papers we wrote — great to see how it’s still going! cs.cmu.edu/~jbigham/pubs/pdf… cs.cmu.edu/~jbigham/pubs/pdf…

Queenie officially retired from the @cmuhcii today — as PHD Program Director since the beginning, she has been a part of literally every one of our HCI PhDs journeys, and from the outpouring of thanks it’s clear she is highly appreciated! HCI would be a different field without Queenie’s care, compassion and leadership!

this is one of the deepest insights i've gotten from a CHI paper -- not surprised it came from @fabulousQian, @nikola_banovic and @cmuhcii's own JZ if you're creating with GenAI, sit with it for a moment dl.acm.org/doi/pdf/10.1145/3…

pretty good performance for 15 years before LLMs! benchmarks are very different (we didn't try to generate the text to enter into a form field, for instance), but really not that far off ;P

here's a paper by ME from 2009, it used a large corpus of web automation scripts to learn to predict the next action given a description (look familiar??)… we had the user choose which to go forward on to protect agency webinsight.cs.washington.edu…

there was a lot of stuff back then, one of my favorite is PLOW -- super compelling, it turns out that it's not as hard as you think to do "ok" on agent tasks b/c there's only so many things you can do on a web page This was Best Paper at AAAI in 2007 cdn.aaai.org/AAAI/2007/AAAI0…

remember, web/UI agents is a field of AI research that was invented in 2024, and not at all in the early 2000s by AI+HCI people ;P
Remember: Agents is a field of AI research that was invented in 2024, and not at all in the 1970s

It was fun chairing #UIST2024 with @lining_yao and @hciprof! And it was great to have @wobbrockjo give the closing talk on his super impactful work. He started his talk with this quote from Scott Hudson.


largest #uist2024 ever fills the beautiful ballroom at the Omni William Penn

I brought my two older kids to the #uist2024 demo session last night -- @AmandaL79938988 & @jasonwuishere did a great job presenting our work on user interface design evaluation to them, … but I think they were more excited about the VR demos :)

understanding whether a UI design is good is important for understanding user interfaces & vital for automatic UI generation at #uist2024, we presented a significant step forward in automatic UI evaluation - check out thread below! (paper: arxiv.org/abs/2404.12500)
When you ask an LLM to generate a UI, how do you know if it's good? It is often easy to recognize a good design but difficult to explain why and reason about/evaluate tradeoffs. UIClip is a data-driven model that assigns numerical scores to represent quality of UI screenshots


another heads up for #UIST2024 folks -- the buses from downtown (westin) to CMU are pretty fast and frequent -- best is the 67 (but it's hourly), the 61s come every 10 minutes or so -- buy a ticket in the Transit app ($2.75) or on board