OneSpace - Shared Visual Scenes for Active Free Play

Posted 25 May 2014

For academics, there are several kinds of projects: projects that you have to do, projects that you should do, and projects that you desperately want to do. This one falls into the desperately want to do category. For me, this project hits all the major high points: (a) building systems for kids; (b) trying a novel system, and (c) studying “play”.

Main Idea

The main idea that is captured by the OneSpace system is that video-conferencing for kids sucks because most video conferencing systems are built for a “talking-heads”-style interaction (Cohen, Dillman, MacLeod, Hunter, & Tang, 2014). For kids (say aged 5-10), the focus on “talking” sucks. In OneSpace, the idea is that you transport both people (as their whole bodies) into “another place” altogether, and allow them to play in a shared visual scene (Ledo, Aseniero, Greenberg, Boring, & Tang, 2013; Ledo, Aseniero, Boring, Greenberg, & Tang, 2013; Ledo, Aseniero, Boring, & Tang, 2012). This allows them to play with their whole bodies, enacting scenes with their partners, who are also represented by their bodies.

We are of course not the first to design video conferencing systems for kids (see: Yarosh; Inkpen), but I think our work really drives home the point that for kids to be engaged in a video conferencing session, they need to: (a) be “physically” (as in their bodies) involved, and (b) be really engaged with their partners. The point about bodies and toys in a visual scene has been made before, but merging the two remote sites into one is something that has not been explored in the context of “kid play” before.

In this work, we designed a study where we watched kids (and kid-parent pairs) play — for a while using OneSpace and for a while using a standard video-conferencing environment (like Skype)… 10 minutes each. Then, we examined the nature of the play that was happening. Major conclusions: in OneSpace, kids are more engaged in ‘bodily’ play, and far more engaged with the person on the remote side. We also did some post-study interviews where we asked kids about their experiences using the systems.

One of the things that often comes up in this kind of academic study is “novelty” — that is, “Oh, maybe they liked the OneSpace system because it’s something they’ve never used before, whereas they would have with Skype.” This is generally a valid critique. What was striking was that none of the kids that we had in our study had ever used Skype before either. That is… for them, Skype was also novel!

Lessons on Method

It is very, very hard to recruit children. We went through most of the summer with very low recruitment for children participants. We never figured out exactly why, but some theories: (a) a one hour study is not really worth a parent driving a pair of kids to the University for; (b) we weren’t paying enough. Our original idea was to recruit 20 pairs of kids, where some would be similar aged, some would be siblings, and we would look at different gender pairings, etc. Ultimately, we compromised on our recruitment, and also included child-parent pairs (in fact, I think the bulk of our participants were child-parent pairs).
Video coding is hard work. For us, the purpose was to capture the “flavour” of the interaction in a summary form. We did this by taking the video and splitting it into 5s intervals, and coding for each interval. To support this, we found (and then modified) two schemes — one that characterises the “nature” of play (is it physical, is it make-believe, for example), and one that characterises the level of “engagement” the two people have with one other (are they cooperating, or are they playing independently, for example). In retrospect, I would take a slightly different approach, where we would do a quick once-over of a small handful of videos, and think about what is really important to code rather than to try to capture “everything” that is going on. I kind of think this is a fallacy anyway — a coding necessarily discards a lot of information; the point, to me, is that we should be capturing “the interesting bits” rather than to use a coding scheme a priori that may not articulate the things that really make each scenario unique.
People get caught up on video quality. The OneSpace system, as a prototype, was a system that fundamentally had worse video quality than the Skype system… That is, the image quality was poorer (lower resolution), and I think the framerate was poorer. Asking kids for their reactions to the systems was tricky because the quality was something they could immediately point their finger toward, whereas how the different representations changed their behaviour was apparently less obvious to the participants themselves.

Genesis

The genesis of this project for me stems back from my studies of how people interact with one another given different representations of their bodies in video spaces. The more recent history of this work started with David and Bon Adriel’s work in putting together the OneSpace system (with Sebastian’s help) for a graduate level class in CSCW. This grew in part out of Bon Adriel’s prior work, The Looking Glass, which married people’s video representation with a historically captured depth scene. The technical implementation and our ideas at the time were reported in a CHI 2013 workshop paper and a poster in the same year. We also met Seth at the workshop, and coordinated our research efforts so as to not overlap one another’s efforts.

Subsequent to this, the team changed to allow David and Bon Adriel to focus on their first year grad studies. Kody and Haley took on the project as part of an undergrad course, and here, they began looking at how kids would play using the OneSpace environment. Through the summer, Maayan would join the team, and played a central role in recruiting, running participants, and ultimately completing the video coding.

Maayan also created a video for the work, which is totally awesome.

RICELab Rethinking Interaction Collaboration and Engagement