JT Blog

GWAP launch

Luis and Co. have launched their Games With a Purpose site. In addition to old favorites like the ESP game, there are some newer ones there related to common-sense knowledge (Verbosity) and music tagging (Tag a Tune). Happy addiction.

How much is your attention worth?

I was pointed to Seriosity today when Jenn sent this article along. The premise, rooted in research on models of attention from MSR among other places, lets users attach a sort-of attentional currency to their outgoing messages as well as to the messages they read. In this way an economy develops that (hopefully) results in a reliable indicator of which messages are important and what is required to get certain messages read quickly.

I think it would be a lot of fun to study how these kinds of economies develop within an organization, how people make these estimates (e.g., what factors go into the amount of currency attached to a particular message, such as importance, immediacy, etc.), and how long it takes to stabilize. More importantly, it will be interesting to see whether people will incorporate the additional work of using Seriosity currency (called Serios) into their email habits.

Publish or...don't

Good post on why publishing is important for us industry folk.

Metcalfe at Mot

Bob Metcalfe spoke at Motorola a few days ago, it was the first time I'd seen him in person. He's quite a funny guy, and even managed to hold my attention when he got a little technical in the "how ethernet started and evolved" portion of the talk. I think I enjoyed the "invention is a flower, innovation is a weed" portion the most. He was frank in noting that all his financial/business success is not due to the fact that he invented ethernet, it was because he sold it to anyone who would listen. I think he was in town because a number of the startups he advises have some interest in the mobile phone market. His new focus seems to be on clean renewable energy, let's hope he sees more success.

The wait is over

After numerous failures, I have finally managed to squeeze my way into the CHI'07 full paper proceedings. It was a fairly laborious process involving the following steps:

• Write paper, revise at least 10 times.
• Submit paper.
• Receive reviews, write rebuttal, revise rebuttal at least 5 times.
• Receive conditional accept, revise paper about 3 more times.
• Receive final accept, revise once more for publisher.

Actually, with each year the process is growing more similar to the journal submission process, which quite honestly is probably a good thing. It hopefully improves the overall quality of the conference and (if the conference program committee members do their jobs well) provides some accountability to the whole process.

Anyway, here's me in the program. I will have a copy of the paper on my website once it's approved by the publisher. See you in April.

Google's user happiness

I had the good fortune of attending a talk by Google new-hire Dan Russell (formerly of IBM Research) at Georgia Tech on Monday. The topic was "What do Google Searchers Think About?" and centered on the "User Happiness" group at Google. This group attempts to explain the user rationale behind their search queries by examining at a micro level how users interact with search, and at a large-scale level by analyzing the millions upon millions of queries that users generate every day.

Dan's opening comments included the classic HCI mantra that a designer's intuitions are usually wrong, and that the designer, the designer's mom, and the designers friends are not typical users. Key to the group's efforts are the search for intuitive transfer of skills between google applications through evolving design guidelines for both interactions (e.g., Google maps and google earth) and visual design (blue indicates a link). To do this, they run large-scale analyses on queries and click data, conduct qualitative diary and ethnographic studies, and perform usability lab studies that gather eye-track and think-aloud data.

What do they find? Here's a smattering of some of the results presented:

• a large number of the top query terms are the names of other sites like ebay, amazon,etc. What this means is that alot of people actually use Google to take them elsewhere rather than typing the simple URL into their browser.
• people are quite bad at quoting their search terms, with only 16% of them using quotes properly.
• people are similarly bad at using more advanced boolean queries. "Booleans and people don't mix."
• the click-through rate of the 1st, 2nd, 3rd, etc. results presented by Google follows a Zipf distribution, meaning that the first link is clicked on roughly twice as much as the second, three times as much as the third, etc.
• users do very little scrolling, and rarely look at subsequent pages of results after the first page.
• while use patterns vary from user to user, individuals have stereotypical interaction patterns with Google that hold up pretty well.
• queries can give insight into temporal (elections), seasonal (flu season), and geographic (a team's World Cup progress) phenomena.
• interruptions play a large role in information-finding tasks, as people get easily distracted from whatever goal they were initially seeking.

These results help to calibrate a designer's thinking when considering the actions a "typical user" might make when using a new potential product. I came out with the impression that what they are really looking for are the simplest UI tweaks that can make the difference between being merely usable and being utterly intuitive. It also causes designers to reconsider what success really means in the context of search, whether it's the time to get a result (not all queries necessarily come a specific information goal), or subjective satisfaction with the search experience. It also raises some interesting research questions with respect to what can be inferred or even predicted about what is going on in the world by analyzing the kinds of queries flowing into Google's engine.

Along the way there were some interesting anecdotes about the early days of user testing at Google that mirrored a talk I heard from Marissa Meyer at NPUC 2003. For example, the copyright notice on Google's main page was added to let users know that the page had finished loading. There were also some slides on the massively parallel systems, simple storage structures, and relatively simple API that allows access to massive amounts of log data with minimal coding effort.

AJAX and calendars

ZDnet has had an AJAX-related article/blog post seemingly every day for the past few weeks. This time, David Berlind discusses the threat posed by this collection of technologies to the current calendar standards now set by Outlook. There are a slew of these systems out now, and Spongecell (cited by the article linked above), is one of the cooler ones, integrating email and SMS for easier phone calendaring.

So what will it take for such systems to beat the competition? Personally, I think it's in large part a question of enterprise versus personal use. Gmail made inroads into MS Office's user base because Outlook and Outlook Express are too heavy-duty (and expensive) for the personal email needs of a lot of people. Integration with calendar and todo isn't as essential. Calendars do not have the indispensability of email for most, and viable systems will need the meeting scheduling, calendar browsing, and email integration valued most by enterprise consumers. In this respect, I still think Google has a leg up on the competition, though their calendar app is still unreleased as of this writing. Two interesting areas apart from enterprise use include domestic (family) calendaring, and public event calendars, both of which could see a dark-horse startup establishing a respectable level of adoption.

Note that adoption by a critical mass of users may not apply here for a given application. The iCalendar standard is supported by nearly every system out today, and an ability to import/export this format may be all that is needed to enable sharing among heterogeneous apps.

AJAX certainly makes it easier to build shared calendars, and iCalendar makes it easy to support calendaring on lots of applications. Indeed, my Augur system would have been immeasurably easier to build had we employed it. But in the end, it's still about identifying your users and giving them what they need.

Update (2/3/06): A friend pointed me to this post about a new calendar called 30 Boxes that sounds reasonably cool. Nice integration with Gmail and upcoming.org, easy sharing and meeting arranging, not bad. Read the comments, and you'll see that people are still looking for workplace support (PalmOS sync/Activesync, Outlook support, etc.)

Metadata is people

I had the good fortune of attending the thesis defense of Luis von Ahn last week, titled "Human Computation". Luis specializes in inventing clever ways to get lots of people to do tasks that computers can't do yet. He is a co-inventor of Captchas, the ubiquitous warped text you must type before you can get an account on many websites, and the ESP game, an addictive game that has the side effect of generating tags for huge numbers of images on the web. The basic premise of the talk was that by designing the right fun games for people to play, you can get them to do alot of the hard work currently beyond the abilities of machine intelligence. You don't have to pay them, you don't have to recruit them; in fact, many will play for hours a day free of charge. Since those two projects, he has branched out into a game called Peekaboom, which identifies the locations of objects in an image, and a new game called Verbosity (under submission) that attempts to use similar principles to collect common-sense facts. If successful, this project could have a huge impact on AI.

I think the most important thing to take from the talk, however, is Luis's claim that human intelligence can be applied this way to virtually any problem. While there are certainly the problems of designing the right game to encourage participation and whether the principle holds for specialized knowledge that fewer humans possess, this is the kind of work that makes everyone in the room think about how they could apply it to their own research. The potential benefits to both researchers and practitioners are significant.

See also: Amazon's Mechanical Turk and NASA's Clickworkers project, which also leverage human computation but rely on a kind of scientific altruism to obtain participants.

Brain games

Good essay here on the recent craze in "brain training" games for adults in Japan. Apparently, fear of old-age dementia has driven the sales for “DS Training for Adults: Work your Brain” through the roof.

Aside from using the very real fear of aging to sell to an untapped market of older adults, this game shows the possibilities for incorporating mental acuity into technological solutions for an aging population. While improving acuity is still an active area of research (the Nintendo DS brain trainer was developed by a Japanese university professor), evaluating this acuity in an unintrusive, possibly fun way may provide an important piece of context to share with family and other loved ones.

HCI Conference calendar

Thought I'd mention that I maintain a conference calendar on my former lab's swiki. I'd at some point like to transform it into an RSS feed, but right now it's pretty low-tech.

A nice calendar is also available at interaction-design.org (with iCalendar file links - awesome), so my calendar may be on the verge of obsolescence.