To investigate the sensitivity of a mixed fish-robot population to parameters governing the stochastic motion of a robotic constituent, this paper explores a robot motion model structure that facilitates the exploration of naturalistic changes in robot activity level. This model is intended for use in experiments involving interactions between a robot and a single Banded Archerfish (T. jaculatrix). The robot motion framework consists of a two-state Markov chain coupled with an autoregressive motion model. The parameters governing the robot's motion are tuned to match those of actual individual archerfish, and can be manipulated to change robot activity level. This framework is being used in ongoing experiments designed to investigate the influence of robot activity level on an archerfish's behavioral response to the robot.