For full text search please use the '?' prefix. e.g. ? Onboarding

Tuesday - Day 3

Tuesday

  1. Awards (Career + Test of Time Papers)
  2. Best paper session
  3. Keynote: Show it AND tell it
  4. Industry

Highlights

Test of Time

  • Read the enterprise data analysis paper (2012)
  • Try SpaceTree - hint at subtrees with glyph (2002)
  • Design studies methodology: "goal = transferability not converge same solution"
  • "necessity is the mother of adoption": morse-smale computation

Best paper awards

  • Affective viz paper: a framework for how we think about whether viz are working
  • Multiple Forecast viz: effect of color on trust + ensemble ML models + isolating effects of vis choices
  • See the D3 evolution paper to learn about API design effects on wild in Stackoverflow

Keynote (Marti Hearst) - ADD MORE TEXT TO GRAPHS

  • "Language should be considered co-equal with viz
  • Check her Wikipedia: Flamenco project, faceted search

Industry

  • 6 feet of rain graphics : how much attention does your viewer have?
  • When research roadmap is long, don't let up on coming up with littler milestones, not losing connection to business
  • Ethical responsibilities as consumer of OSS / publication

Raw Notes

Plenary day: when event is single-track

Awards

  • For each awrd, name who is on the committee _ how they reached consensus

Papers to check out later

- Anton's friend Fritz Lekschas of Harvard
- Cindy Xiong: Perecntual and Cognitive Affordances

People Awards

  • Top new researchers
    • Danielle: Genomic colors tidied https://danielleszafir.com/
    • Arvind S. Road can be long, Heer focused on impact problems. Lyra was rejected a bunch of times

Top technical achievements

  • Valerio Pascucci: Topology from playing orgami as a kid (great picture slides)
  • Shixia Liu: Machine learning and text visualization at Tsinghua: great diagrams

Lifetime achievement

  • Colin Ware: Ocean and Kelp vis

Vis Service

  • Gautam has attended since 2004

Thesis section papers

Test of Time awards

  • 2012: Sean Kandel, Heer, etc: Enterprise Dataviz - foundational interview study
  • 2002: Spacetree: Jesse Grosjean in Bangor Main makes a notetaking outlining software
  • 2012: Design study methodology @tamara.munzner (Private)
    • Saw "codeswarm " viz... dozens of papers have spawned it since
    • Transfer not reproduce
  • 1997: Core graphics problem; ROAM (triangle binary tree) for performant terrain rendering
    • Tarek and maps team may like this
  • 2007: Fluid flow vis - handle vector fields that are changing over time
  • 2008: Scivis; "Morse smale" : "necessity is the mother of adoption"
    • one of the funniest + most engaging presntations -> How an algorithm scaled up to bigger volumes

Paper Awards

  • Reminded how te tracks are dynamically made every year

  • Check out the honorable mentions too

  • Affective Intent in Data Visualization ( @elsie.lee-robbins (Private) , '17)

    • https://arxiv.org/pdf/2208.04078.pdf
    • Used the Periscopic gun deaths
    • "All vis have affect" (we care about reader response), whether or not we explicitly tried for it
      • not just transmitting a fact
    • Used DVS designers + twitter to eval... develop verbs for evaluating whether vis is "working"

  • Multiple Forecast Vis (Covid) (Lace Padilla)

    • Show deaths coming up: do you trust, and can you guess which way data goes next?
    • Excellent example of how to design a study to isolate effects
    • Use ecological validity over aiming for lab purity
    • Showing people more models may make them trust, but more trust is not same as draw data-backed conclusion
    • Color influenced trust: grayscale seemed more credible

  • Uncertainty in high dim data (Dave Hagel) - Pokemon diagrams :)

  • Vis challenges on stack overflow ( @leilani.battle (Private) ):

    • https://ieeevis.b-cdn.net/vis_2022/pdfs/v-short-1097.pdf
    • See more on d3 template design that followed up on Wednesday
    • Takeaways
      • How to get people to encouarge code that can be modular and reused?
      • Beginners / people short on time struggle to figureo ut how to carve out an example
      • People use d3 from 6 languages, 55 packages, 36 viz types but 4 main types (bar, line, network, map)
      • Lots of meaningful qualitative data snippets
      • Future... think about d3 evolution vresions effect (mixed on whether modularity helped)

Keynote

@marti.hearst (Private)

  • meta: used DALL-E images for section anchors (hundreds of TV scenes, memorable section anchor)

  • Show it or tell it: language is not given same weight in viz frameworks as graphic choices, but it should

  • Scott McCloud- pose question of when picture or text take the lead, other has to fill gap.

  • Annotation core problem

    • How much text should chart have (we get it's important)
    • What should it say, where should it go
    • (I could add more viz to page)
    • Actually wasn't able to trigger a too much text effect as long as it's all relevant.
    • Some people actually liked pure text more than a graph

  • Comparisons

    • Comparing text rankings, comparing pictures...
    • It's relative (e.g. how tall vs not tall)
    • https://a.cl.ly/12uzbYdq Chatbot- ppl feel mixed about seeing graph in texting uI
    • Scrollytelling case... unclear if ppl were maximizing the benefits of text/graphics independently

  • Brain

    • When we have text literacy, it's near automatic, does odd brian thing
    • Emoji, non text looking shadows may interfere with fluency

  • Closing... language as UI

  • Working with prompt engineering

  • Funny closing: minions talk to end :)

Q&A

  • Captions: autogenerate vs handcraft?
    • Autogenerate is OK for predicatble, for complex you need COdex-like human in loop for decent results
  • Language / reading uncomfortable in VR
  • Dyslexia, eng not first language

Industry/Sponsor section

  • Northern Data: "map not the territory"
    • Keep computers mining crypto cool by storing them in arctic circle
    • Use viz to help researchers with AI to keep this stuff efficient
    • Photorealistic data: weather chanel 6 feat of water: convey a feeling
    • Help technician learn what datacenter to plan better for remote maintenance (3d artists)
  • Tableau - Vidya (recognized from Lee Wilkinson memorial meetup)
    • Cool overview of new algos for interacting/exploring with data, going beyond the desktop (slackbot data query)
    • Help people build dashboards
  • Autodesk Bon Adriel just joined
  • KAUST: Saudi Arabia sponsor, drone video
  • Oklahoma: big push into data sci investments
  • Industry Panel
    • Skills student missing
      • Tableau: data is often messed up
      • Oklahoma -> haven't worked in real workflows, gut check fishy data (40 inches rain), collaborative setting
    • Advocating R&D vs business value
      • Tableau: 3-5 year time horizon for big plans (beyond that is too far), work on PM resaerch q's when they can. Share what you're doing often. Has mandate to think bigger than the main product lifecycle. (Me... it's on us to pitch the bigger roadmap and how the local OKRs are chipping towards it)
      • Autodesk -> 5-10 year horizon, but also for slower moving industries (with physical component) like architecture/manufacturing. 50,50 small blocks too
      • Northern: if it fits with mission, basically convince 1 person
    • Reconcile business vs research value
      • Tableau: probably won't spawn whole new project area- but look at peer acquisitions (slack, etc) to see how to link to ecosystem
      • Autodesk: rather than brand new page/app -> try to slot into existing workflow. Don't give them more work
    • Data scale
      • Small data, make synthetic data with properties that you need to stress
      • Oklahoma: make sure they really understand the question, do small prototypes and expand from there. E.g. 5 mo before 5 years.
      • Much bette rpartnership if people come in wanting to use what you have (vs needing to be persuaded)
  • Audience Question:

Hack Ideas

  • Vis for notetaking tools / personal knowledge graphs / entity annotations (timelinejs, dendron, datasette)
  • ASCII art / sparkfont, ascii graph, etc - blur text <> picture
  • Site/plugin that helps make multi paper or overlays
Backlinks