Jarrett Interaction Design

• SINQ – engaging kids in scientific inquiry
• example: “Why is the poop on fire?”- smoking pile of manure at urban farming project
• SINQ – online community for guiding kids through scientific inquiry: pose a question, explore resources and answers, form hypotheses, define variables, create a project to investigate, provide feedback on projects
• scientific inquiry has multiple entry points, multiple kids can enter and participate at different points (social process), distribute the work of inquiry, let kids choose tasks, combine into a whole
• iterative participatory design of SINQ with kids; kids want more color, media, game-like elements, more explanation of the process

• how to analyze networks changing over time; show and analyze the changes (node/link changes, growth/contraction, clustering, peaks, comparison); many tools (ManyNets, NetEvViz (NodeXL), TempoVis), apply to real problem applications (Nation of Neighbors, Encyclopedia of Life)
• TempoVis + EOL: monthly growth; node = person (237), link = commented on species (571); discover interesting cluster in a particular month; some curators are network hubs, others aren’t; one student group very active in one month without curator involvement

Lessons learned from the Nation of Neighbors

• Nation of Neighbors – online neighborhood crime watch; collaborate with law enforcement
• more that 500 neighborhoods, 12 of which include law enforcement
• collaborated with community managers and social scientists to study what made some communities more successful that others; what does success mean? who are the influential members? how does it change over time?
• success metrics: active months, number of members, total activity, interaction intensity; analyzed using ManyNets
• successful communities have more reports in their activities; 22% anonymous
• influential leaders send many invitations; recruiting; leaders contribute to growth of community as well as growth of activity
• over time, successful communities have fewer invitations but growing and stable reports activities
• recommend: promote leadership among citizens, integrate with popular social media, use visual analytics to understand community dynamics

• node-link visualizations break down at a couple thousand nodes, or particularly dense relationships
• also want to reduce barrier of entry for novice users and make it simpler for everyone
• better layouts, alternate visualizations, graph summarization
• repeating patterns (motifs) exist, motifs often dominate, motif members can be functionally equivalent
• fan motif: fan-shaped glyph; parallel motif: connecting relationship glyph
• glyphs must be representative (shape, size, color), easily distinguishable, easily comparable, allow overlaps
• interaction with glyphs to provide more detail and actions
• network diagram -> categorical coloring -> glyph substitution -> color for value
• implemented and tested in NodeXL; users successfully discover relationships they couldn’t see without it
• challenges: only two motifs so far; glyph design has tradeoffs

Lots of activity this morning around the demos and posters. I took a cruise through, but had other work to do, unfortunately.

As usual, the HCIL Symposium has provided a series of really interesting research projects with engaging presentations and a very warm, welcoming vibe. I’m looking forward to tomorrow.

There are many things I saw today that I am going to take away and see if I could apply them to my work. List comparison, pair comparison, event flow alignment are always interesting to ponder as I’ve worked in a variety of information spaces with visualization and dashboards. The other stuff, while it may not be as directly and immediately applicable, gets my creative and thinking juices flowing. Good things will come of this.

I’m taking notes differently this time. I used to take handwritten notes and process them a bit more while I posted them later. I don’t have time for that anymore, so doing it live was my option. I also added the Instagram vector. What I found difficult was using Twitter hashtags consistently across three different streams: my site (a blog), Instagram, and Twitter. I cross post to Twitter from both my site and Instagram, and I sometimes forgot the hashtag, sometimes misspelled it, and the 140 character limit really is a problem with long research presentation names. I had to mangle my blog post names to ensure the hashtag would appear on Twitter.

• signed up early for Twitter, but found she didn’t use it… until she found a Washington Capitals community; back-channel communication with a group; now Twitter is useful
• if you back-channel, you can reveal the results to those who can’t participate at the time; so, how to prevent “seeing the result”, yet still participate in Twitter for other things
• preventing spoilers requires blocking 100% of tweets related to an event; but this is not a sane computational goal
• 93 million tweet database covering 24, Glee, and the NFL NFC championship game; randomly select 10,000 from each show; went through each of the 30,000 to code (friends and mechanical turk); 75 about 24, 150+ for glee, 1200 for NFL
• A Naive Test: blacklisting; automatically pulled strings from databases like IMDB and wikipedia to create the blacklist; did really well, generally 3-4%; missed some really hard ones that are killers – like a score; false positives – block lots of safe tweets – 30-70% of non-problem tweets were blocked; probably 30% is okay, but 70% is too high
• areas of research: social computing; computational linguistics; UI issues
• applications: events; parental control; unwanted content (e.g. politics)

• how to get casual users to translate for us; for International Children’s Digital Library (5,000 visits a day)
• MonoTrans: take a phrase in a source language -> machine translation -> native destination language speaker for corrections -> machine translation -> source speaker uses explanations with pictures or otherwise -> go back and forth and select best translation; 68% of sentences are high quality (Google translate only 10%) with test participants
• need to engage casual users; current UI has too many sentences and too many tasks
• MonoTrans Widgets: one sentence at a time, one task at a time, embedded in ICDL to ask for engagement; tasks choices – make corrections, point out errors, choose best translation, explain translation
• issues in the real world: many more English speakers, so most translation is between English and other languages, rather than between other languages, so gave priority to rarer languages
• since August 2011, 11 languages in 6 language pairs; translation quality much higher than Google translate with casual web users
• original UI is better for dedicated users, more context, more choice; new UI is better for casual users, less overhead, less commitment, minimal context

• collection of “middle-state” artifacts; artifacts created during the process and practice of design
• green-wiring: modifying a printed circuit board with wires; like the red pen of a literary manuscript
• Jennifer Freyd – dynamic mental representations – e.g. artificial character drawings; depending on how character is drawn, distorted characters would be recognized if the process of character production they were taught more closely matched the distorted character
• history-enriched digital objects: age, wear, and patina digital objects to indicate usage over time
• diggers and spelunkers: exploring the edges of world-of-warcraft to find unfinished and abandoned areas and then document them; understand process and temporal information about game development
• development of a poem: is it really gestational? collection shows a wide range of metaphors, not just gestation; e.g. textile weaving, warp is foundation, weft is the design
• frustration is inherent in the creative process; many things in collection are about failures, false starts, or abandoned elements
• artists often try to provoke and even alienate their user; different than designing for use; e.g. responding via physical postcard to tweets (Mark Sample)
• collection may provoke new ideas, tools, methods, and environments
• “anticipatory design strategies” -> design fiction

• textbook illustrations are important to learning, but are largely inaccessible to blind students
• in Maryland, schools are required by law to make reasonable efforts to provide all instruction materials in accessible media
• one textbook translated to Braille costs about $30,000
• using 3D printer, transform 2D illustrations into 3D tactile representation; cost of printer has gone from $10,000 5 years ago to $1,000 for an assembled kit today
• still have to write 3D model to print
• goal: take a photo, print a model from the photo directly
• process: photo -> image processing -> assign meaning to features -> generate program to -> generate printer code -> printer
• some issues to explore: flat vs. multi-height, text and callouts, overlapping lines, how much detail
• testing with Maryland School for the Blind