Software development for
undergraduate course
Module Topic
The Critical Stability Task (CST) Quick et. al. 2018
The CST is a virtual control task designed to probe the sensory motor control systems of the brain by providing a test bed to record and correlate neural signals with motor movements.
The goal of the task is to control an unstable cursor (shown left), which will drift towards the edge of the screen, by imagining there is a rope around the cursor and "pulling" the cursor back to center. The subject must stabilize the cursor for 6 seconds to have a successful trial, if the cursor reaches the onscreen boundary then the trial ends and is a failure.
The task gets more difficult by increasing the speed of the unstable cursor. In the Batista lab, the speed is increased until subjects can only complete 50% of trials at that difficulty. This is known as the Critical Instability Value (CIV), and successful trials near the CIV are analyzed.
Laboratory subjects (Rhesus monkeys) typically slowly increase their CIV over the course of many months of training. For the methods module we wanted a brief introduction to the CST, therefore we wanted the students to reach their CIV in only a few dozen trials.
Efficiently approach CIV
We experimented with a multitude of techniques to effectively and efficiently approach the student's CIV but determined the best method was using the weighted up-down method (Kaernbach et. al. 1991). To determine a 50% success rate, the difficulty would increase by X% for a success and decrease by X% for a failure.
To determine the CIV, the mode of all the previous trials was taken, as the convergence value would be the most common value.
To determine the optimal step size for X, we chose a very small step size, 0.05, and performed hundreds of trials until the values converged to a difficulty of 4.2. We then iteratively increased the step size X until the subject (a.k.a. me) was not able to reliably converge to a difficulty of 4.2. The chosen step size X was 0.2.
Utilizing this technique we drastically reduced the time needed to reach the CIV from hundreds of trials to 25. (shown by the red bar)
Cursor Speed (Difficulty)
Data Collection
To streamline the process of the student's learning, data collection was made automatic and in an intuitive format for the students to analyze.
The data collection would take approximately 15 minutes so it was designed to pick up where the students left off, even if the program was abruptly stopped.
Conclusion
The module was set to take place the third quarter of the spring semester, just as quarantine took effect. However, this was an entirely electronic module which could be conducted remotely, and the students were able to perform their own data collection smoothly.
Much to my relief, there were no problems, bugs, or other issues. The module is set to run every year for the foreseeable future!
"A job super well done."- Dr. Patrick Loughlin