[From Richard Kennaway (2009.06.23.2208 BST)]
There's a program that Bill has mentioned a few times for running a tracking task similar to the one here:
http://www.mindreadings.com/ControlDemo/BasicTrack.html
but which also fits the parameters of a control model to the human performance, and then runs the model on the same data, giving a performance correlating very highly with the human. Is that online anywhere? I want to show it to some people.
···
--
Richard Kennaway, jrk@cmp.uea.ac.uk, http://www.cmp.uea.ac.uk/~jrk/
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.
[From Rick Marken (2009.06.23.1715)]
Richard Kennaway (2009.06.23.2208 BST)
There’s a program that Bill has mentioned a few times for running a tracking task similar to the one here:
http://www.mindreadings.com/ControlDemo/BasicTrack.html
but which also fits the parameters of a control model to the human performance, and then runs the model on the same data, giving a performance correlating very highly with the human. Is that online anywhere? I want to show it to some people.
It’s the TrackAnalyze program in Demo 4 on the CD of Demos that comes along with LCS III. It’s a pursuit tracking task (mean that the target moves so you have to pursue it with the cursor). Bill made me a compensatory version, which is more like my Java demo above. I’m using it to collect data for my paper on “Causality Does Not Imply Correlation”.
Best
Rick
···
–
Richard S. Marken PhD
rsmarken@gmail.com
[From Bill Powers (2009.06.23.1938 MDT)]
Richard Kennaway (2009.06.23.2208 BST) --
JRK: but which also fits the parameters of a control model to the human performance, and then runs the model on the same data, giving a performance correlating very highly with the human. Is that online anywhere? I want to show it to some people.
BP: Rick is right, it's demo 4-1 for LCSIII. You have that, don't you? If not, ask and it will be given.
I think something like this should go in your paper on correlation, too. It would be good to point out that the high correlations you talk about aren't pie in the sky; given the right model, they can be achieved. The basic prediction that works is that people will show characteristics of a control system, rather than a top-down cognitive (compute-action-and-execute) system or a bottom-up (stimulus-response) system. What doesn't work well enough is the kind of model in which some circumstance or stimulus or other independent variable causes some particular kind of behavior to become more likely -- in other words, 99% of psychological studies. "The effect of A on B."
Best,
Bill P.