Jeffrey Heer

The data vis project "Mosaic: An Architecture for Linking Databases and Scalable Interactive Visualizations" received a SIGMOD Best Demo Runner Up award at the ACM SIGMOD/PODS International Conference on Management of Data in Berlin, Germany.

Congrats to the Mosaic team on this recognition! 📊🎉

Screenshot from GitHub showing 1k stars for the Mosaic project, an extensible framework for linking database and interactive views.

The Mosaic architecture for database-backed interactive visualization just hit 1,000 stars on GitHub! Thank you to all who have participated and contributed!

Congrats Kaitlyn!!

🙏

GitHub - michelle123lam/lloom: Concept Induction: Analyzing Unstructured Text with High-Level Concepts Using LLooM (CHI 2024 paper). LLooM automatically surfaces high-level concepts to analyze unstruc... Concept Induction: Analyzing Unstructured Text with High-Level Concepts Using LLooM (CHI 2024 paper). LLooM automatically surfaces high-level concepts to analyze unstructured text. - michelle123lam...

Hi Lynn, curious if you’ve tried LLM based approaches (like LLooM: github.com/michelle123l...) and if so how they’ve fared for you

A webpage that reads "This domain is for sale! CatGPT.com. $220,000 USD."

TFW you have a grand idea only to find it's out of reach. So it goes for my chatbot that meows and purrs.

Allen School professors Brian Curless and Jeffrey Heer named ACM Fellows for helping transform how we use computing technologies - Allen School News Each year, the Association for Computing Machinery (ACM) recognizes the top 1 percent of its members who have made notable contributions to the field of computing science and technology as ACM Fellows...

Allen School professors Brian Curless and Jeffrey Heer named ACM Fellows for helping transform how we use computing technologies news.cs.washington.edu/2025/02/14/a...

Allen School professor Amy X. Zhang receives Sloan Research Fellowship for empowering users to make ‘our online spaces as rich and varied as our offline ones’ - Allen School News As the head of the Allen School’s Social Futures Lab, professor Amy X. Zhang’s research draws on the design of offline public institutions and communities to then develop new social computing systems ...

Allen School professor Amy X. Zhang receives Sloan Research Fellowship for empowering users to make ‘our online spaces as rich and varied as our offline ones’ news.cs.washington.edu/2025/02/24/a...

And congrats to you, too!

An invitation into Droste's Lair A swords-and-sorcery programming environment for building and counting mathematical structures

@joshuahhh.com and I invite you to explore our experimental visual programming environment (which takes place in a dungeon)

Features
- dissolve the bounds of decision and explore all possibilities at once
- use sigillic incantations, delve into recursive caverns

vezwork.github.io/drostes-lair-post

Us #ieeevis 2025 OPCs worked hard on this: the new call for papers is now live! ieeevis.org/year/2025/ye...

Submit your best work!

Brief account of changes for 2025:
* Extra week for supplemental material
* Student reviewer program
* Opt-in publication of reviews for accepted papers

Hi, I’m a main author of the Vega/Altair curriculum notebooks. Indeed these focus on tool learning and aren’t intended as a standalone text for a full vis course. We combine it with many other papers, links, and lectures, as in courses.cs.washington.edu/courses/cse5...

Kasper Hornbæk on LinkedIn: Dear HCI community (please share, forward) We are seeking two Outreach… Dear HCI community (please share, forward) We are seeking two Outreach Editors for ACM Transactions on Computer-Human Interaction (TOCHI). The Outreach Editor…

ACM Transactions on Computer-Human Interaction is looking for outreach editors, i.e., junior researchers who wants to rethink and implement how the journal engages online with readers and authors. Deadline for applications is Dec 11. Call at www.linkedin.com/feed/update/... or ask me for more info.

FWIW it took UW IT about a month to complete our subdomain request, but once set up you can administer it yourself.

Updating DNS records for a top-level UW managed subdomain was painless, too. (We did it for @idl.uw.edu.) You just need to add an `_atproto` sub-subdomain, then add the TXT record to that.

Nice! I loved that work and always thought it would be good to do more like that 🤗

Congrats Sanjay!

Overview of the DracoGPT-Rank pipeline. (1) User provides prompt templates for an LLM to rank chart pairs; (2) Draco featurizes charts and produces feature vectors consisting of constraint counts; (3) Draco learns constraint weights over LLM-labeled chart pairs by fitting a RankSVM model; (4) The fitted Draco model can be applied to score charts. Results at each stage of the pipeline afford insight into LLM ranking preferences.

People are now using LLMs to create charts and graphs. How might we assess the quality and consistency of the results? DracoGPT is a method that fits a visualization knowledge base (Draco) to LLM responses, enabling comparison across models, prompts, and results from human subjects experiments.

Authoring and visualizing multiverse analyses with Boba. Users start by annotating a script with analytic decisions (a), from which Boba synthesizes a multiplex of possible analysis variants (b). To interpret the results from all analyses, users start with a graph of analytic decisions (c), where sensitive decisions are highlighted in darker blues. Clicking a decision node allows users to compare point estimates (d, blue dots) and uncertainty distributions (d, gray area) between different alternatives. Users may further drill down to assess the fit quality of individual models (e) by comparing observed data (pink) with model predictions (teal).

Love this work! Also related is the Boba project, led by Yang Liu and @kalealex.bsky.social, which includes a variety of multiverse visualization and assessment methods. idl.uw.edu/papers/boba

1️⃣0️⃣0️⃣0️⃣ 🙏 📈

We could precompute the projection and the visualization would load a lot faster, but I wanted to keep all the spatial processing details in the example so folks can see how it’s done. Internally we’re working on a publishing tool that will optimize load times and precompute preaggregations, too!

GitHub - uwdata/mosaic-framework-example: Using Mosaic and DuckDB within Observable Framework Using Mosaic and DuckDB within Observable Framework - uwdata/mosaic-framework-example

We have an example site deploying Mosaic in Observable Framework here: github.com/uwdata/mosai...

More generally, see links on the project GitHub repo (scroll to the bottom): github.com/uwdata/mosaic

Our research lab is now here on Bluesky! 📊📈

Mosaic serves as an architectural layer that provides (1) access to a backing database and a variety of query optimizations, and (2) a shared selection abstraction that enables integrated filtering across data views (visualizations, tables, input widgets, etc.).

Also, FWIW the Mosaic site is built using VitePress, not Observable Framework. For more explanation of Mosaic, start here: idl.uw.edu/mosaic/what-...

Mosaic is a framework for linking interactive views and backing databases. Examples include input widgets and vgplot, an interactive grammar of graphics built on Observable Plot, but you can also build and integrate custom components. The Mosaic core handles database querying and linked selections.

Haha been there done that… Neutraface was the official Stanford HCI font while I was on the Farm. The UW move was my opportunity to jump ship to Avenir

Incidentally it’s also been the official UW Interactive Data Lab typeface since 2013. (FWIW I’ve always had a soft spot for Futura but it’s a bit… much. Avenir plays it cooler 😎)

Evaluating Human Factors Beyond Lines of Code Software systems researchers want to make human-centered claims, but don’t have the proper tools to do so. That’s how we ended up with the ubiquitous lines-of-code comparison found in e…

my take on bringing HCI to PL

blog.sigplan.org/2024/11/21/e...

(And in case it’s interesting to you: underneath the hood Mosaic automatically pre-aggregates sufficient statistics, binned at the pixel level, to compute updated regressions on the fly without having to touch the original, large dataset.)