Ahh, Pittsburgh. One more week of data collection is in the can and it will be the last data collection trip to Pittsburgh. What more data needs to be collected can be done in Philly. In theory, I shouldn’t have to come back to Pitt until defense time. One can dream.

While the data analyses haven’t all been run, the coarse level reveals that there is no difference again. A wonderfully interesting turn of events. However, this stack of data does include the 1back by itself (stay and rotate, I still can’t believe I forgot it in the first round of data collection). Now it is possible to look at the interference on the 1back by the pointing task.

In the micro-minutes between experimenter interventions this week, I’ve been working on revamping the jACT-R launching infrastructure. jACT-R projects are now actually OSGi plugin bundles. This will make deployment significantly easier and allows me to leverage the PDE tools in Eclipse.

I’ve been testing it for the day and it’s looking pretty good. There is an initial speed hit when starting a model for the first time (as the IDE is copying configuration files to the working dir) - but after that it’s acceptable (well, on a MBP, that is). I’ll look into a more aggressive caching solution later.

But the extra overhead nets some nice benefits. First off, the OSGi bundles allow for easy deployment (one could drop a fully functioning jACT-R bundle into an OSGi container and it would start automatically). This doesn’t benefit most modelers, but from an applied perspective this rocks. Next up are the PDE tools. The wizards that can be created are quite useful. It took a few hours to create two wizards to assist modelers in the creation of modules and instruments. Very cool.

It took a whole 30 seconds to import the spatial stuff into the new structure.

If a modeler wants to create a new tool, he can use the wizard to create the skeleton. After he’s fleshed it out, he can test it within that project. To use it in his other projects, he just has to link their dependencies and there it is. When it comes time to share his new toy with others, all he has to do is export the project as a deployable plugin and then people can just drop the resulting jar file into their plugins folder. Slam, bam, thank you mam.

Or, more importantly, one can keep all of the model code, data and configuration within the project. Then to share it with others, he just has to export the project - all the configurations will travel with it. No more struggling to run other peoples’ models.

I’m thinking about looking into setting up a discovery site to make it easier for folks to advertise their instruments, modules, and models.

Oh, and in other news, my cogsci submission has been accepted. I can finally put something new on the vita.