It’s been a long day so far, I took the red eye from Seattle last night so I’ve been up for a total of 28.5 hours so far. Apologies if not everything is coherent.
I’m not sure how many people reading this are familiar with the CHI conference so here is a quick overview. Computer Human Interaction (CHI) is the premier academic conference every year for academic Human Computer Interaction researchers and for professional usability experts, user experience people and interaction designers who are doing cutting edge stuff. Every year, something like 3000 or so people descend for 4 days of often overwhelming talks, panels, demonstrations, video presentations, networking and, ideally, plenty of drinking. It’s considered somewhat of a taste maker and arbiter within this community. No one person can give a completely impartial view of CHI, most of the time, there’s literally about 14 different things going on at once. The most you can do is go to as much of the stuff that interests you as possible and hook into interesting conversation in the crowd.
Still, CHI is somewhat remote from the workaday world of people who are building just another web application or mobile app. The stuff talked about at CHI is often highly abstract and, quite frankly useless to professional developers and designers. Despite that, I love it. It’s a bunch of overwhelming smart people with an astonishingly diverse array of interests all talking about the stuff they’re passionate about with a mix of brilliant insight and hilarious naivety.
The keynote was given by Judy Olson on Social Ergonomics and I had really high hopes for it as it was right in my area. I, personally, found it unneccesarily shallow but I kept on hearing from other people that they loved it so perhaps my perception was due to an overly familiar view of the work.
Judy talks about the idea of ergonomics and how it defines the relationship between the body and space and then she brings this metaphor to the social sphere. Stuff like proxemics tells us how to behave based on the distance people stand away from us, subtle social cues help us guide conversation and we judge attention based on gaze among other things. Pretty standard stuff so far and, while interesting, nothing overly challenging or difficult to work with. She then goes on to detail how such cues break down when we mess with the physics of real space: satellite delays in speech, video conferencing in which the other participant is both smaller and at weird angle and poor audio causing people to lean into a teleconference.
My biggest disappointment with this talk, and note that this is purely my personal opinion and I think that the research she does is great, is that:
a) I thought ergonomics was an overly restrictive metaphor in this case and resulted in a rather literal translation from real to virtual space. Everything is conceptualised within the standard 3D framework that the human body is used to operating in. I think such an overly reductive approach misses the richness that comes from highly abstract forms of online communication. What are the ergonomics of a social graph or a asynchronous conversation over forums? I’m somewhat of a collector of metaphors for social design and social ergonomics strikes me as a rather poor one.
b) I felt that her work focused overly on rich, mediated face to face or simulated face to face interaction which is a very researchy kind of area but one I believe to be largely irrelevant when it comes to producing impactful designs in the field.
In the early days of HCI, there was this tendency to try and get computers to replicate the full richness of face to face interaction and, with it, came an implicit assumption that face to face interaction was the gold standard when it came to social behaviour. I’m not accusing Judy of that now, I think the field has moved somewhat past that phase now and tries to exploit the unique benefits of technology, but there is still a prejudice towards heavyweight, overly mediated interactions.
I think one subtle bias for this was that such systems were easy to evaluate and, thus, easy to publish for. You just use real face to face interaction as the control and then you can test how good your new system is compared to the control. Simple numbers, simple paper. But as I pointed out in my blog post on virtual worlds, the truth is that real life interaction really isn’t all that great and the true power of virtual interaction is to be blatantly better than real life interaction in certain ways and suffer from being poorer in others. The two modes of interaction are apples and oranges but to admit this means that you can’t perform simplistic analysis.
If we were to look at the true success stories of technology mediated interaction, they would be voice, text messaging, email, IM, forums, blogs, social networks etc. We love these forms precisely because they aren’t face to face interaction. You can perform them across space, perform them across time, strip the social nuance of phrases, deliberately add ambiguity, reply in your underwear, take time in constructing your thoughts and present a fictionalized version of yourself. All of this stuff is horrific if abused but awesome if used right.
Open source projects are probably one of the most well studied remote collaboration tasks and, in almost every single real life open source deployment, people reported that they loved having a constantly on ambient skype channel between sites as it greatly increased team cohesion but every time they experimented with video, they would shut it off in about 2 or 3 weeks. Now, the true HCI zealot would argue that this is because the technology and design of such systems is not advanced enough but I am skeptical about such faith.
Yes, video chat has it’s place and serious, rich co-located interaction of the style Judy Olson (and many other HCI researchers I know) have a place in communication but I think they are and will always remain a niche place and I would much rather the CHI community focus more on the complex social nuances of these seemingly simple and boring communiction mediums instead of building yet another technologically complicated remote co-located workspace that assumes the only reason we’re not remote co-locating is due to technology.
Phew, so that was Judy Olson’s talk. I have to say despite everything else, CHI is taking social stuff noticeably more seriously this year than it was last and I’m glad they’re finally on the social wave… 5 years after it had crested.
Creativity, challenges & opportunities in social computing
Call me old fashioned but I expect panels to actually panel for a significant period of time, not just present. Sure, give a 5 minute overview of your work but the reason we put you up there was to provide the spontaneous discovery of information that only back and forth conversation can provide. Of the 4 panel participants, only one actually stuck out in my mind and that was the Scratch project at MIT. I’ve not had time to explore it but I would definitely love to come back to this at some point and give it a thorough exploration.
One thing that was noted in the questioning was that the panel really didn’t talk about the “dark side” of social computing and I feel like the panel continually soft balled their way through that point, even while claiming to take it seriously.
This was a great session with some really interesting material. The most outstanding talk of the bunch had to be Eric Gilbert’s on revisiting The Strength of Weak Ties work that the Grannaanbanana guy did way back int the 70’s and repackaging it smartly with interesting new data. Social scientists have been theorizing for many decades on what makes some relationships “strong” while others casual and have come up with several factors they think influence it. What Gilbert has done was ask 35 participants on facebook to rate the strength of their tie with each and every one of their facebook friends and see what facebook data can estimate the strength of such ties. He nicely seperates out bulk statistics such as how many blog posts and how many positive emotion words into larger, sociologically relevant categories like intimacy and reciprocity.
He then proceeds to perform a tight piece of statistical analysis which does a careful job of sifting through the data. Of the results I remember, R^2 was 0.5 and MAE was 10% but the most interesting one was that if the task was to simple many a binary distinction between strong and weak ties, the accuracy was 87%. What I wish he had talked about more was the outlier cases which completely confounded the regression model. He presented two but I think a few more would have really put that 87% in context. If a false match is 100 times worse than a correct match, then a 13% error rate stops looking so good.
Jilin Chen’s talk on the performance of recommendation systems was a neat little piece of experimentation but I thought it unfortunately just subtly missed the point. Chen’s work was on replicating the “People you may know” feature from facebook on the IBM internal social network site, beehive (I really have to talk more about IBM’s efforts in the social computing space because it’s really one of the greatest stories never told within the web 2.0 world). He experimented with 4 different algorithms for suggesting people and, my interpretation of his conclusion seemed to be, they all seemed roughly the same in quality but biased differently between recommending old and recommending new friends and that their fancy system resulted in the most friends added.
What seemed like the tragic waste for this study was the implicit assumption that more friends = better friends. I know from my personal experience that when facebook added their “people you may know” feature, I was horrified because the top 5 people on my list were people I was well aware were on facebook but I had no interest in friending. Sure enough, within the next 2 weeks, 3 of those 5 people had sent me friend requests and I was forced to deny a request for the first time in my facebook career. A look at quality rather than quantity of friend requests would have done this paper a lot of good.
The 3rd and 4th papers in this session were both notes so lightning 10 minute talks and I was honestly baffled by some of the results given.
A quick summary of “My Dating Site Thinks I’m a Loser” was that they wanted to evaluate how much people gamed a reccomender system that gave them bad results. They had two conditions: showing the user a picture of themselves or not and giving one reccomendation per 10 questions or 4 recommendations at the end. Oddly enough, putting in a picture turned out to be a major confounding factor. With the picture in, people gamed less for the 10 minute condition but without the picture, people gamed more. (as an aside, there’s a fascinating visual language to psychology 2×2 graphs which I only realised today. I’ll have to write about it sometime as people who see a a lot of psych studies can see two straight lines on a graph and instantly infer what and how is interesting about it). I have no idea what to make of this information, why would looking at your own picture lead to such a radical shift in behaviour?
Finally, was a paper on designing for forgiveness which I found odd and challenging to get a handle on. I understand that forgiveness is an interesting social topic and one I should be in theory interested in but it was presented in such an abstract manner and without any examples that I found myself unable to judge whether it was even an interesting take on the topic or not. Oh well, I guess I’ll read the paper for this one when I have the time.
When I was looking through the program for the first two sessions, there were so many interesting talks I couldn’t make. There was fascinating stuff on privacy, navigation, security, trust and a whole bunch more that I so dearly wanted to go to but were in conflict with more important talks. I think the one I most wanted to see was a panel on how user experience can learn from food design, being a major major foodie and one who’s plugged into the food discourse, it pained me to miss it. But the third session, there was not a single paper I was dying to see. So I mainly spent that time getting some relaxed networking in, free from the vicious time constraints of 30 minute breaks.
Phew, blogging a conference takes a lot of time and I’m starting to understand why people prefer to twitter conferences instead. Still, I think it’s important to make the material in the CHI community relevant to a larger audience and also I think to provide one person’s interpretion on what can often seem like intimidating topics to the uninitiated. So often, I see people approach talks uncritically because they assume they know so little and the presenter so much. But once you’re willing to hold to your own opinions and defend them respectfully, regardless of the “eminence” of the author, you start to see the literature in a different light.