Response: Socially perceptive robots

November 26, 2012

Social perceptive robots: Challenges and concerns by Ginevra Castellano and Christopher Peters is a short summary on the current state of socially perceptive autonomous robots in the field of childcare. While movies and television shows suggest these kinds of artificially intelligent agents are just around the corner, Castellano and Peters are quick to dismiss such thinking, claiming that there are a number of flaws in existing socially perceptive robotic systems.

Unbeknownst to me before reading this paper, there are actually many current attempts at producing autonomous agents for use in childcare. In retrospect, this seems to be an understandably desirable innovation — daycare, especially for low-income parents, is quite expensive, and having one parent stay home to take care of their children is not always a viable option. Unfortunately, Castellano and Peters are quick to point out a number of problems and necessary considerations in currently developing systems that will deter such robots from acting as a lone babysitter anytime soon. Among these include: the design of robots to prevent injury to the user; the ability to respond to subtle and unexpected behaviours (often in noisy or uncontrolled environments) appropriately; the present focus of such systems being on analyzing basic and acted emotions rather than the subtleties mentioned previously; accounting for individual and cultural differences; and recognizing when a child might be lying.

Castellano and Peters finish their summary by referencing the concept of ethics and safety guidelines for such childcare robots, which would need to be inherently programmed into an agent of this kind on top of the aforementioned considerations. This might be a difficult concept to swallow for anxious socially perceptive AI researchers, as such a process would require extensive debate, legislation, and involvement from all regions planning on adopting autonomous childcare robots as apart of their social systems. Not only will the technology, which is still not nearly ready for market, take several years to perfect, but there are a notably high number of post-considerations to research and discuss before the technology can even come into the hands of parents. I think I can safely say that this is a technology that is not just around the corner.