Robot Criminals
papers.ssrn.com
papers.ssrn.com
And, funny enough, humans seem to be pretty good so far at influencing robots.
I could imagine a robo equivalent to the 2008 financial crash, maybe a horrible transit accident or something, but where no single person or corporation is quite criminal enough to hold accountable, but the end result is still bad. In one world, all the people go "Oh hey, we weren't really punished" and do it again. In another world, you put the robot in "jail" and the people go "Oh shit, they took our property. That was expensive. Let's not do that anymore." ... optimistically.
In that case, you have three possibilities:
1. You don't confiscate/destroy the robot. BigCorp goes "hell yeah, cost of committing $crime is 0!" and continues.
2. You confiscate/destroy the robot. BigCorp goes "nice, we can make another for $5 - we'll take another 100,000 thanks" and continues.
3. You confiscate/destroy the robot. BigCorp goes "shit, that robot cost a ton of money - better work out a way to avoid doing that again in the future" and fixes the problem.
The hypothetical "how should a robot car decided which bystander to hit when it looses control" or similar trolley problems are so so so far off in the future. Right now autonomous cars can't even operate in inclement weather.
I do find the concept of "moral reasoning" in robots an interesting one, but again feel the author has skipped into the far further and missed more tractable questions. If moral reasoning were in place today, I think the likely decision of all autonomous cars would be to drive slower. Is saving a few minutes of a trip worth dramatically greater probability of fatality? US DOT found that speed limit increases from 55 to 65 MPH increase the likelihood of fatality by 24% once an accident occurred.
https://safety.fhwa.dot.gov/speedmgt/ref_mats/fhwasa09028/re...