Modelling cognitive tasks as PID control

[Seanny123]

In "" the researchers model a few tasks as PID control problems. Mostly just problems where a person tries to predict where a dot will show up on a circle or a line.

Experimental tasks.

A) On each trial of Experiment 1, participants selected a horizontal location with a joystick, and then were shown the correct location. Figure adapted from McGuire et al., 2014

B) Participants in Experiment 2 selected a location on the circle with their mouse, and then were shown the correct location.

C) Experiment 3 was identical to Experiment 2 but participants were rewarded based on their accuracy, according to one of four reward-error functions. They were informed of the current reward mode during Fixation, and during Feedback they received the reward corresponding to their accuracy on that trial (conditional on their reward mode).

To make these models work, they have to do a lot of awkward fitting. It would be interesting it to see if basing the PID controller on biology makes it easier to fit.