Talk at CES on 1 December 2014 at 4:00 pm titled "Measuring information flow in fish-robot interactions" by Sachit Butail from Indraprastha Institute of Information Technology, Delhi (IIITD), India
Robots are controllable machines that can be made to look and move like animals thus providing a viable tool for studying animal behavior. At the same time, a clear measure of their influence on an animal subject is not available. In this talk I will describe how tools from information theory, in particular a quantity called transfer entropy, can be used to measure the directional information flow between animals and robots. We will consider a robotics based experimental setup, in which a zebrafish is observed as it interacts with a robotic replica. Our results show that the transfer entropy is significantly more from the replica towards the focal subject than the other way around, and that this difference is not present when the replica is replaced by a conspecific. These results support the use of transfer entropy as a measure of information flow in social animal behavior, and present an indirect evidence of the effectiveness of robots in animal behavior studies.