The day started off with an outstanding key note talk by Stellan Ohlsson from University of Illinois at Chicago. The main focus of his talk was the nine information modes that he has identified. These nine modes are the finite set of mechanics which can be used to transfer information from one person to another. Admittedly the list of nine may not be perfect but it provides an excellent starting point. Next he addresses that as you build your ITS you receive diminishing returns on your investment to a single mechanic. For example if you are trying to perfect when to provide positive feedback, your effort is logarithmic in the pedagogical gains a student receives based on your effort. So rather then focus all your time on a single mechanic it is better to provide multiple mechanics, which will result in improved gains. Naturally he provides some examples which support this claim. Examples include tutors which did comparison studies of whether or not to give positive feedback, negative feedback or both. The evidence suggests you get the best results when providing both types of feedback, and potentially those gains continue for the incorporation of all nine information transfer mechanics.
He also included an excellent analogy to the Wright brothers, and that they systematically attacked the challenge of figuring out how to fly. Rather than randomly investigating the infinite sized search space of vehicles that can fly, they broke the problem down to its components, wings, propulsion and structure. He recommends that we, in the ITS community investigate ITS development in the same manner. Don’t randomly search the space, but identify the base mechanics and work on the investigation of those mechanics.
I find the idea of addressing the ITS investigation as a search space problem a profound idea and extraordinarily helpful as a method for thinking about research.
There were some different talks in the sessions leading up to my own talk, but considering it was my first talk to this particular community and I was introducing a fair amount of work that I had completed I was a bit nervous. As such I sadly didn’t take away as much as I could have, like I will tomorrow, with no talk of my own looming over head. Nevertheless I saw an interesting talk about including social words in dialogue as opposed to strictly information in dialogue between players and virtual characters. A work presented by Amy Ogan, titled: Persistent Effects of Social Instructional Dialog in a Virtual Learning Environment, the effect being if there is social dialogue rather than just informative, players will be more considerate about their actions as they interact with virtual agents. The player talks to three characters, Zahora, Farid, and Hassan, and more or less when using social instructional dialogue players feel more of a connection with the agents. An example of social instructional dialogue would be something like:
My brother and sister work for a shop that my mother runs.
as oppossed to strictly information dialogue like:
Family members in Iraq often work in shops together.
In any case check the proceedings for more detail, a pretty interesting paper and pretty meaningful for when trying to create the sense of immersion in your players in a game, with obvious applications to RPGs, like the Mass Effect series.
After the session I had my own talk which went well enough. I felt that my talk bounced around a bit too much but I somewhat felt it was a byproduct of trying to cover too much information. However, I was given very helpful ideas, Gautam Biswas from Vanderbilt will send me some information and resources about sequence identifying algorithms and I was also kindly directed towards works by Mazza in regards to open learner models. Philip Pavlik also offered a very helpful and lengthy conversation, thanks for your time, about some general ideas to look into with the tool. of particular interest to me is hte idea of applying the concepts from the key note talk to my own work which sounds incredibly interesting. Essentially the idea is to take each dimension of the data and represent it with all the possible visual representations. Then build the entire space of visual representations, and then determine a method for discovering which, one visualization, of the entire space of visualization possibilities is the most effective, or possibly which set. A pretty interesting concept and I look forward to talking to the experts at UNC-Charlotte’s Vis Center about exploring this idea.