Punishing Robots That Do Wrong

Justin Shin
Stanford University

Stevenson 1301
12:00 PM

When we hold people accountable for their actions, we often invoke the possibility of punishment. This is especially true in military traditions of dealing with wrongdoing. If the possibility of punishment bears an important role in accountability, and we cannot punish robots, and we require accountability for robots in particular roles, then robots ought not to occupy those roles. How successful this argument is hinges partly on what we count as a successful instance of punishment. I argue that we can and should punish robots. I also argue that, while this debate has so far been constrained to military robots, it is salient for robots and algorithms in other fields, such as carceral algorithms and surveillance robots.