Greetings from wonderful Boston, home of the CHI 2009 conference! The first day’s events are over now and I’m already suffering conference fatigue! We presented WantKnot at the Student Design Competition, and unfortunately it was not selected to the finals. But congratulations to fellow Michigan projects Treasure Hunter and MIFresh, which did make it! I also thought I’d share my liveblogged thoughts of the talks I went to:
Designing Digital Games for Rural Children: A Study of Traditional Village Games in India
tl;dr version: learn from traditional games & apply their lessons to educational games. Very good idea and results seem to be good.
Quality of schooling is a challenge despite increasing primary school enrollment
Low English knowledge
Provide skills in arithmetic, reading, writing (no magic bullet)
Use e-learning games to teach
More fun & incorporate good educational principles
Demonstrated benefits in urban & rural children
Existing initiatives in using games for education in developing world
OLPC, Azim foundation, MILLEE, shared computer w/ multiple mice
Partnerships between govts and private sector
Western video games do not match rural children’s expectations of games, since children have little exposure
eg Frogger – why do you have to move sideways? you don’t move sideways in real life
Did a contextual inquiry of children’s games
Games are a marker of social identity – embarrassed to play rural games in front of urban people?
23 outdoor games, lots of “tag” games, 5 indoor games
Outdoor games are less expensive and more accessible
What can we learn from these games?
players, resources, goals, actions, rules (Fullerton 2008)
Variable # of players – social elements?
some team cooperation
offensive, defensive roles
primary, secondary roles
player states
resources – players, things around (sticks, stones, marbles), places/territories
goals – eliminate, survive, acquisition, object manipulation (seven stones – assemble a heap of stones)
actions – pursue / guard prisoner, escape from capture, throw / dodge projectile, manipulate object
compare traditional games vs. Western video games (bjork and holopainen)
increasing difficulty level – more enemies, more subgoals
leveraged tree-tree for vocabulary learning / alphabet
Eyespy: Supporting Navigation Through Play
tl;dr version: ESP Game-like app for tagging images for navigation. ESP Game clones are all the rage but I like that it’s situated in a city.
Database of images useful for navigation
Geotagged photos
Mobile phone game “with a purpose”
like ESP game
Annotate maps with photos
Not just a matter of networking human brainpower – design challenges
Tag locations in space (w/ wi-fi triangulation)
Other people confirm locations – if you do, both people get a point
2nd player has to go to the location where the 1st player took the picture to confirm the tag
Limit 5 tags per day
collected 200+ tags
people enjoyed the game – score/competition is a good motivation
connection to others who also play the game – walk in their footsteps
new interactions w/ environment
explore new areas
few tags w/ ppl, transient objects
lots w/ buildings & signs
roads / paths were mostly unconfirmed
riddle tags – text only tags with some kind of clue for someone else to figure out
what makes a good tag?
findable, instructional, recognizable
Design for navigation
orient to other players
Compare against Flickr geotagged photos for the same area
Lots of events / artistic photos, less functional for navigation
Flickr has many more people and transient objects, EyeSpy has more sings, shops, roads, doors.
Test this using walking routes – people can navigate easier using EyeSpy than Flickr
Design for human computation
Motivate participants
How is the game marketed?
Orient participants
Relationship between system and by-products
What Do You See When You’re Surfing? Using Eye Tracking to Predict Salient Regions of Web Pages
tl;dr version: Eye tracking tells us that people look at the top left of webpages first and the right side last, and they pay lots of attention to big images. You don’t say? Yawn.
Eye tracking can show which features on the website are most salient
characterize web page behavior
predict recognizable regions based on the DOM
study had 4 tasks
1. four information foraging tasks
2. two distractor tasks
3. four information foraging tasks w/ similar topics, different questions)
4. page recognition – answer questions about what web page / web site you saw before (how often have you visited?)
How do we measure which elements are salient, based on gaze fixation?
- Fixation duration / impact
impact refers to Gaussian distribution of points around the fixation point (2 degrees, ~66 pixels)
this allows a single gaze to cover more than one DOM element
- # of participants viewing an element
- time to first fixation
Web page browsing habits
no matter what the task, there’s a common orientation phase – look in upper-left corner to get your bearings
information foraging goes all the way down the page, much less on the right
takes much longer to look at the right (banner blindness)
top left corner is looked at first
page recognition occurs entirely above the fold
Predict recognizable regions based on DOM
Size, position and level in DOM are the most expressive DOm features
plain HTML element-related features do not have much predictive value
Checked only features that would be easily accessible by a search engine
built a linear regression model to check predictions for webpage vs. actual data
It feels Better Than Filing: Everyday Work Experiences in an Activity-Based Computing System
tl;dr version: virtual desktops (like the Mac app Spaces) split up by activity, not document or application. Seems useful, but a lot of work to set up.
Activity-based computing, not document- or application-centric
Giornata – extends OS X
Virtual desktop manager, tagging, activity-based collaboration
Runs at all times
Features
Jump to virtual desktop for each activity – spatial organization
Each desktop contains applications, contacts, and files most salient to current activity
Can tag activities at the time or retroactively as you make sense of what you’re doing
Include visualization of contacts for that desktop
show unread emails from each contact
Tested system on 5 knowledge workers (4 academic, 1 industry)
participants used system on daily basis for mean of 55 days
Liked system
~7 activites open at any one time
Only 10 activities were closed – in several cases, a similar one was recreated the same day
Lots of activity switches, ~30 per day
System was used in diverse ways
Activities = mental to-do list (lots of activites, little switching)
or managing self (few activites, lots of switching)
or very few specialized tasks (few activities, little switching)
Tag usage
~1.8 words per tag
lots of descriptions – project / event name
lots of untagged object
individual names
multiple simultaneous activities
on other virtual desktops – “it’s never clear where anything was. if i’m not paying attention, that can still happen”
what do you do with email client? it fits into multiple activities
activities have different states – background, completed, dormant (will have to come back to)
Desktop storage was the big win
few leftover items on desktop
use “slough” desktop to represent stuff yet to be filed
desktop undermines desire to keep it neat and clean
Tagging
tagging seems more long-term than participants wanted
Collaboration
Some users didn’t use OS X address book
made users aware of communication practices in activities
Mapping from activities to colleagues might not be the best model – can we be more flexibile?
Pathfinder – An online collaboration environment for citizen scientists
tl;dr version: website to engage amateur scientists to collect and analyze data, make conclusions and discuss the results. I love the idea, but right now it’s basically just Wikipedia with a few extra tags and no “killer feature”. And who’s going to motivate all these people?
Citizen Science
eg Christmas Bird Count – data set for 300+ papers
Non-scientists as field volunteers – Data collection / engage public in scientific process
Are we only engaging public in part of scientific process? Can we built a tool to help?
Track & contribute data sets (user-generated content), such as commute time
Tracks can also be scraped
Also can discuss / analyze data tracks on a wiki page (w overview / summary)
Include evidence pro&con, to-dos, questions for each topic. Each links to data tracks
Can people use it? Would they be able to use it?
Compare Pathfinder against wiki page – no significant differences in how they created it, but liked reading Pathfinder better. Not faster or more accurate though
More arguments in editing than writing – people deleted what they didn’t like
Can you attribute people’s writing? Can you directly address someone who posted earlier?
Similar to Wikipedia – no one attributes their own work (except on talk pages) – [is the backchannel important?]
Meta-discussions – want to comment on the milestones, but leave article itself alone
DIscussion
How do we motivate people? Especially novices?
Milestones – more than just fixing typos, but not at the level of a full article
Foster a community of practice – the task must be legitimate, but not too overwhelming
Experts / professionals – need more of a reward, including reuse
“My Dating Site Thinks I’m a Loser”: Effects of personal photos & presentation invervals on perceptions of recommender systems
tl;dr version: when you’re aware of how you’re being presented on the Internet (through a profile pic or something similar), you’re less likely to “game” a recommendation system that you’re dissatisfied with. Neat connection.
What happens when recommendation systems go wrong?
“Game” the system – give it weird answers to help fix the answers
Personalization system
Gather info
Build profile
Match user
Present content
Does increased self-awareness affect consistency of response?
If recommendations/profile are still being worked on , does timing matter?
Dating website – present ideal self
Study
always presented poor results
photo/no photo x recommendation every 10/recommendation every 40 questions answered
Participants rated recommendations
Photos – more consistent responses (less gaming), despite liking system less
Intermediate recommendations increased frustration
The Application of Forgiveness in Social System Design
tl;dr version: principles to follow for allowing forgiveness in social systems like e-commerce. Good idea, but unfortunately the presenter had difficulty keeping the crowd engaged.
People do not always do the right thing, but these offenses can be repaired
eg – unintentionally post misleading info in ecommerce
Forgiveness is bounded – not mandatory may not be unconditional
Forgiveness is interpersonal, one-to-one
1. Respect dyadic nature
2. Build flexibility
3. Support motivating factors
4. Design interventions for both pre-emptive and retroactive uses
5. Public awareness
Problem with asynchronous communication
Must motivate users to forgive
Touch and Toys: New Techniques for Interaction with a Remote Group of Robots
tl;dr version: Developed a system to control multiple robots by positioning physical objects on a tabletop computer, or using a touch screen. This was just fun to watch, and I’m a huge fan of direct manipulation. Some slight technology issues but nothing major.
Robots have many degrees of freedom
Unpredictable environment
real-time demands
current methods for navigating robots have a high cognitive load, even when many people are involved
some methods use physical objects, not desktop
control groups of robots
tabletop computing / tangible user interfaces
use chessboard metaphor – meaning is embedded in spatial arrangement of objects & tangible direct manipulation
1)use toys (TUI) & their positions on a table to control robot movements
2)use touch-screen to control
cameras positioned to track tabletop objects and robots
test by asking users to duplicate formations on whiteboard
tested TUI and touch, with up to 3 robots on each
More robots = things take longernon
TUI was ~10% faster w/ 2 robots, nothing else significant
TUI was easier to visualize, in contact w/ robot, focus on avoiding collisions, but tracking wasn’t as accurate and caused problems
touch was simpler w/ less equipment, more precise, less intimidating, but more visual clutter & less intuitive (Rotation / translation)
users preferred TUI, esp w/ more robots
two hands for simple tasks, one hand for complex tasks (generalization)
TUI is easier for two-hand touch
participants took efforts to keep robots from colliding – were involved. It feels more real.
focus on where the participant meets the interface
Users liked TUI more even though they were both about the same
So that’s it for Day 1. I’m not sure if I’ll liveblog again – it was a lot of fun, but the Internet is too funky there to liveblog / livetweet the way that I’d like to. Onto day 2!