The paper explores the possibility of creating a robot partner whose actions could be perceived by humans in a way that is similar to that of a working dog. The study is based on several key assumptions: 1) The moral evaluation of a robot partners behavior depends on the psychology of human perception; 2) the moral level of the control system can modulate the behavior of the artificial agent without compromising the functionality of lower control levels, the stability, and the integrity of the robots behavior; 3) a robot partner should perform the dogs functionality and reflect its unique behavioral traits; 4). a robot partner that exhibits dog traits may be limited in its cognitive abilities. A certain set of behavioral traits associated with collaborative capabilities of a dog was selected for analysis. The most important characteristics of a dog, such as body structure, psychology, and human idealized concepts of dogs traits, were ranked by a number of parameters. During the process of model development, the «body structure» characteristic retained such function as the capacity to reflect the emotional state of an agent. The theoretical foundation for the moral level of the robot’s control system is based on H. Sidgwick’s concept of quantitative egoistic hedonism. The developed model was tested in a series of computational experiments simulating two situations of robot-human interaction, each involving several behavioral scenarios. The scenarios reproduced the behavior of dogs and humans in similar situations. The experiments confirmed the operability of a control system which consisted of three levels: emotional-needs, cognitive, and evaluative (moral). It turned out that a number of behavioral scenarios that were externally represented as the result of complex cognitive processes, were in fact realized by a basic conditioned reflex mechanism (the evaluation of the situation and oscillating behavior scenarios). At the same time, what seemed to be some apparently simple reflex behavior required in fact the involvement of cognitive-level control (the reorienting basic reaction scenario).
Petr Sergeevich Sorokoumov (Thu,) studied this question.