RT @YueJiang_nj: 🌟 I am on the job market now!!! 🌟. My research lies at the intersection of AI and HCI, aiming to develop human-centered te….

RT @angie_boggust: In an age of LARGE models, how do we support people in making them SMALLER? Compress and Compare is an interactive visua….

D3 animations are amazing for SVG-based vis, but hard to scale to Canvas/WebGL. I made Counterpoint to help me make large animated embedding plots, and now I’m excited to share it as an open-source JS/TS framework. Presenting virtually @ieeevis this Wed!

RT @w_epperson: We're still looking for a few more participants for this study! If you use pandas for data analysis sign up to try out our….

RT @w_epperson: We’re looking for participants for our user study! If you use #python and #pandas to analyze data in jupyter then try out o….

Check out our upcoming paper discussing these behavior patterns, written with @adamperer and some amazing folks at UPMC!. "Ignore, Trust, or Negotiate: Understanding Clinician Acceptance of AI-Based Treatment Recommendations in Health Care"

We initially aimed to test the effects of AI explanations, but while clinicians rated explanations as more useful, levels of binary "agreement" with the AI stayed about the same! Instead, individuals tended to follow 1 of 4 reliance patterns (surprisingly unrelated to seniority).

Proud to share our #CHI2023 paper on clinician acceptance of AI treatment recommendations! We explore how in complex decision tasks, clinicians often negotiate intermediate actions using facets of an AI prediction, rather than accepting or rejecting the recommendation outright.

RT @a_a_cabrera: Excited to introduce 💠 𝗭𝗲𝗻𝗼, an ML evaluation framework for any data or model, from classification to image generation. O….

RT @michelle123lam: Do you author ML models in your work? Have you ever struggled to reason over the values encoded in your models? We’d li….

You can (collaboratively) create codes as Google Docs comments, and the tool exports the comments and quotes to a Google Sheet (then to Miro/Mural for affinity diagramming). Obv there are fancier tools for qualitative coding, but Docs is free and easy 🥲.

I'm sure a lot of people are working on qualitative analyses for CHI, so I thought I'd share a super simple Python script I built that makes it possible to use Google Docs for qualitative coding of transcripts:

It’s great to see much-needed parents’ and workers’ viewpoints added to the conversation around algorithmic tools in child welfare. Well worth a read!.

RT @anna_kawakami: Excited to share our paper “Why Do I Care What's Similar?” Probing Challenges in AI-Assisted Child Welfare Decision-Maki….

RT @nsmith_piano: My latest article in @TDataScience is about understanding @Neo4j graph embeddings with Emblaze from @cmudig https://t.co/….

RT @anna_kawakami: I’m really excited to share our work on Improving Human-AI Partnerships in Child Welfare:.Understanding Worker Practices….

This wouldn't be possible without @a_a_cabrera's work on Jupyter/Svelte integration, @leland_mcinnes' AlignedUMAP, my advisor @adamperer, and our awesome expert interview participants!. Code (pip install emblaze): Demo: (3/3).

Model builders currently have few tools to choose the most reliable, responsible embeddings for a task. Emblaze lets you project large embedding sets to 2D, animate smoothly between them, and see recommended clusters for comparison. (2/).

This week at #IUI2022 I'll be giving a presentation and live demo on Emblaze, a neat Jupyter-based tool we've developed to help ML model builders interactively compare embedding spaces. Paper: (1/)