Anyone reading this blog will notice that I haven’t been sick much lately, so I haven’t been posting anything. Today I’m taking the time to remind my future self what I was thinking about and doing way back in June.
Sangeeta is getting ready to take Saranya to Nepal for a month. She’s got a week left to get everything ready, her car is in the shop with a major intermittent coolant leak that they can’t seem to find, and she just got addicted to Korean soap operas so she spends all day sitting in front of the computer watching them and reading the subtitles.
World of Warcraft is holding less and less of my attention. I still log on every day to do my jewelcrafting daily quest, and sometimes I take a break and do some Argent Tournament stuff. Oh, and I have calendar reminders to check and renew my Mysterious Eggs on 5 different characters. I can do all this while only actually paying attention to the game for as little as 30 minutes per day. The rest of my time I’m either programming, reading about programming, reading literature on human memory models, or studying math/statistics.
I’m the only employee of Insight Learning Technology, Inc. who’s not on vacation. I’m using the time to take cars back and forth for repairs, and to rework, refactor, and modernize all my PLM server code. I had to fake object orientation before, but PHP5 lets me write code the way I’m used to thinking about it. We’ve got at least two big projects coming up this summer, so I’m scrambling to get some of this background work done and tested before I have to focus on deliverables. The hope is that all this will make later projects easier.
I’ve been studying Psychological journals to see what other people have been doing in the field. Phil Pavlik and John Anderson have a nice model that predicts forgetting and recall time, and I think I’d like to adopt a similar model. Our system has a couple of arbitrary parameters, and I need to figure out a system for making them less arbitrary.
Whenever we create a new module for adaptive training, we have to decide what sort of performance reflects sufficient learning that the learner will be able to correctly answer the item (or an item from the same category) after a delay. We also have to determine the parameters that tell us approximately how long to wait after an item is presented before we show it again. Right now, these parameters are arbitrary and independent, but I think we need to come up with a system for not only generating these parameters automatically, but for relating them theoretically. That’s a path we’ve been loathe to tread, but access to funding for research in the field lies down that road, and we need to show our feet thereupon before the monetary gates will be opened to us.
Then there’s math and statistics. I’ve been looking at performance data from an earlier experiment, and trying to find a pattern of accuracy following particular patterns of problem presentation. I guess I need to learn some data mining and regression techniques to figure out the relationships. My lack of statistics background is holding me back, so I think I’m going to try to sit in on some classes next year.
If I still have a job…
