Notes from 2010 HCIL Symposium

I attended the annual HCIL Symposium yesterday.  This is my second year in attending, and it was worth every minute and every dollar.  Below, I’ve captured some of the tidbits that leaped out of the content for me.  I encourage anyone who is interested in computer science, HCI, social networking, or just having a good, fun, intellectual time to attend next year!

After some introductory material, a few HCIL Hero awards were given out, Ben Bederson presented a keynote on Zoomable User Interfaces, and then thirteen short presentations on HCIL’s current research were given, each for about 15 minutes.  Finally, lab tours, demos, and posters were presented at the lab itself.  This is a fun, high-energy group, and they did a fabulous job presenting very interesting research.  Bravo! Today, workshops and tutorials are being held, but I’m not participating in those.  On Wednesday, they held a pro-bono design service day for local non-profits.  I didn’t know about that soon enough; I plan to participate next year.

Six Attributes of HCIL

Jenny Preece, dean of the School of Information Studies, did a short introductory presentation highlighting the things that define and differentiate HCIL:

1. Exemplary research through hard work.
2. Interdisciplinary approach, across both the campus and the world.
3. Outreach, such as the service day.
4. Amazing food.
5. Supportive; they rejoice in and applaud each other.
6. High energy with extraordinary leadership.

HCIL Hero Awards

Catherine Plaisant presented the first annual HCIL Hero awards, for people involved with HCIL who have made great contributions in the world:

• Ben Shneiderman: Founder of HCIL and the field itself.
• Juan Pablo Hourcade: HCI for Peace; Digital technology for marginalized youth.
• Sharon Laskowski: NIST, VAST competition, and many other contributions to HCI.
• Kiki Schneider: Coordinator of the lab for the last six years; graduating and moving on.

HCIL Service Day

The lab organized a service day of pro bono design work for six area non-profits.  They had 43 volunteers, and it was an amazing success.  Next year, they plan to enlist other organizations and individuals to hold many instances around the world.

Keynote: The Promise of Zoomable User Interfaces

Ben Bederson reflected on his 15 years of research into Zoomable User Interfaces (ZUIs), including early ideas and hopes, successes and challenges, and what aspects of the ideas have “stuck” and become commonplace.

• Google Maps may be the most well-known instance of the idea to date.
• 1994: Spacial organization of related documents. Works great for 100 things, may be okay for 1000, but doesn’t scale to 1 million.  Focus was interacting with the space, not just viewing and printing.
• Now: Huge individual documents (e.g. Google Earth) and small sets like slide decks, iPhone home, etc.  Not dealt with in early ZUI work.
• Microsoft PPTPlex and Prezi are examples of current implementations.
• Early work was all about navigating spaces.  It required special devices with special controls to do well.
• Today, focus is on navigating content, generally in large linear lists, with a small set of useful and easy to use zoom functions (such as the zoom slider in IE7/8 and MS Office or the zoom functions in many View menus).
• Lessons from PhotoMesa: Treemap of lots of photos in groups (folders).  Learned that people aren’t good at scanning unstructured 2D spaces, so most modern photo management software has long lists, which people are good at scanning.  Another deficiency: the layout wasn’t stable as the content changed, so spatial memory was hampered.
• Lessons from SpaceTree/TaxonTree: When showing hierarchical overviews, the children should not get smaller. No way to get a real overview.
• For mobile devices, LaunchTile predated the iPhone by a number of years.  And the iPhone mis-implemented it by zooming from the center rather than from the selected icon.
• See current products like Canvas for OneNote, SeaDragon, and Zoomobi ZoomCanvas.
• Benefits of ZUIs: engaging/fun, feels natural, helps some task performance (harms others), creative potential, overviews, animations.
• Challenges: hard to scale, hard to author, temporal issues, … [moved too fast for me to get them all down]
• ZUI design guidelines: need small representations of each object; keep same aspect ratios when zooming; consistent spatial layout; meaningful layout; scannable layout; don’t do too much – breadth over depth; simple navigation.

Session I: Communities

Self-Promotion in 140 Characters: The Use of Twitter by Congress

Jennifer Golbeck presented research on the use of Twitter by senators and representatives.

• More republicans than democrats are tweeting (125R, 65D, 2I). No one from Maryland.
• Republicans have a tighter follower network amongst themselves (visualized by NodeXL).
• Analyzed approximately 6000 tweets using content analysis techniques.
• Mostly informational posts, then locations/activities.  Mostly sound bites, self-promotion, position statements.
• A few good back and forth debates (e.g. one between a republican and democrat that lasted from noon to 3am). Some direct rep to constituent communication. Some support of transparency (e.g. tweeting from closed-door session).
• Recommendation: Communicate, don’t just broadcast. Good advice for anyone using Twitter, not just Congress.

Analyzing Social Networks with NodeXL

Derek Hansen presenting research on using NodeXL, an Excel plug-in for visualizing networks, to analyze social networks.  The goal of the project was to make Social Network Analysis (SNA) methods easier to use and visualize; most current specialist tools are powerful but complex.

ManyNets: An Interface for Multiple Network Analysis and Visualization

Manuel Freire presented research on a tool to help analyze and compare multiple networks, even thousands of them, or divide individual networks and compare within.  It was a very technical presentation where I didn’t have the grounding in network theory and SNA to grok it all. It seems like a powerful technique in the right hands.

New Design Methods for Children: Layered Elaboration

Greg Walsh presented research on a co-design technique with children where an initial design is done on paper with a group of children, the design is explained to others, then that design is handed to another group that augments the design with a transparency over-top.  This is repeated up to seven times (because the transparencies get cloudy after that).  They have also prototyped a digital version for performing this long-distance.  The method is non-destructive (the new groups augment, but never remove/erase) and asynchronous (the groups can pass it back and forth over time).

This one made me think about how cool it would be if Balsamiq Mockups supported separate layers that individuals could augment and annotate other peoples work over time.  I’m going to suggest it to them for their MyBalsamiq web-hosted product.

Designing Social Musical Technologies at Carnegie Hall

Allison Druin presented research in collaboration with the Weill Music Institute at Carnegie Hall where they are trying to capture the energy, loudness, and creativity of live musical performance in a long-distance collaborative social networking and creation model.  They have worked with projects that are collaborations between NYC and New Delhi, India and NYC and Mexico City, Mexico.  Musicians used co-design techniques to design their own textured lighting for their live performances.  The project is in its early stages, but is exploring how can our visceral experience of audio be incorporated into social media.

Session II: Text and Translation

Human-Computer Collaborative Translation

Chang Hu presented research on using single language speakers in a human-machine collaborative model to provide translation that is higher quality than machine translation but lower cost than expert translators, and also makes it easier to find people to do the translation work, especially between languages where finding someone bilingual would be really difficult.  The process basically involved two single language speakers iteratively improving machine translations with non-textual annotations passed back and forth between the two people to augment the iterative machine translation.

Finding Entries in an On-line Arabic Dictionary

Sarah Wayland presented research on ways to help non-Arabic speakers use an Arabic dictionary so they can communicate with Arabic speakers on social networks.  “Arabic is not English” was the main message; there are many ways in which the language and its written form are extremely different than expected by English speakers, so their ability to use dictionaries is very small.  A finite state machine was developed that helps with many of the common translation problems to help them use a dictionary successfully.  The model can be customized to deal with unique elements of any language.

iOpener Workbench: Tools for Rapid Understanding of Scientific Literature

Cody Dunne presented research on an interactive tool to support quickly generating summary literature survey articles using the text of the articles and their citations themselves.

CrowdFlow: A Human-Computer Hybrid Cloud Computing Model

Alex Quinn presented research on computer-learning systems that incorporate human work and judgment into accomplishing tasks, using Amazon’s Augmented Turk.  Cost, speed, and quality can be adjusted against each other based on needs.  The basic model has a machine generated solution checked by a human to decide if it is worth “fixing” or just “do over”.  Then a human does the “fix” and its quality is checked again.  The example used was finding human forms in surveillance video, which reminded me of Jeff Hawkins work at Numenta stemming from his book On Intelligence.  That vision engine plugged into this model could make a quite powerful human-computer hybrid!

Session III: Search

How Children Search Online at Home

Allison Druin presented research on a year-long contextual study of how children perform search on the web, because children are often frustrated in using current search mechanisms.  They studied 7, 9, and 11 year olds, both boys and girls, and discovered seven different “search roles”: developing, domain-specific, power, non-motivated, distracted, visual, and rule-bound. Individual children demonstrate multiple roles at different times.  There are definite age and gender differentiators in what roles are demonstrated.

Finding Temporal Patterns in Electronic Health Records

Krist Wongsuphasawat presented research on continued development of tools to allow finding patterns in categorical temporal data, building on LifeLines, LifeLines2, and Similan.  One of the interesting elements incorporated into the new LifeFlow tool is the use of Icicle Tree visualizations (invented by Jean Daniel Fekete).  Also, the civil engineering department at UofM is interested in applying these techniques to accident data sets.

Analyzing Trends in Science and Technology Innovation

Ben Shneiderman presented early research on how we might be able to predict the viability and success of scientific and technology ideas via metrics such as publications, citations, etc.  How does the system of publications lead to research and development leading to sales? One case study he covered historical time-lines of treemaps versus cone trees versus hyperbolic trees showing the success of treemaps but no matching success for the others.  Interesting, difficult analysis work, but no strong causal theories have emerged yet.