Is incarceration after the crime too late? Why not before?

[Jon]

I want you to imagine a future where a super-smart artificial intelligence can predict with 98% accuracy that someone will commit a crime and what that crime will be. Let's for now just assume this is possible.

Should we incarcerate this individual, even though they have not yet committed the crime? What is your instinctive response? (I know there was a Tom Cruise movie about this quite some time ago.)

Now ask yourself this. What if it predicted that your next-door-neighbour was going to kill you in the next 7 days? Would your answer be the same?

 

[The_Doc_Man]

That 2% is enough for doubt, but there is also the issue of the civil rights that would have be usurped to be able to specifically predict the actions of an individual. I might consider using that 98% as a reason to talk to the person contemplating this action as a way of letting them know that their intent has been detected. Intervention, not incarceration, should be the first line of positive action. Because I think that inevitability of punishment is adequate as a deterrent in most cases.

There was a short story some years ago, I believe called "The Circuit Riders" (and I don't recall the author). In it, your hypothetical AI was a machine that could detect strong emotion. A good operator was able to detect that one sudden burst of strong, violent emotion was caused by a bad call at a baseball game. In the story as it developed, they found a low-level but persistent and growing signal that they traced to a guy who WAS going to do harm. But they merely took him away to short-term incarceration to let him think about things.

As it relates to your proposal, I would also point out that a 98% AI success rate probably would only be for premeditated crimes because there is such a thing as a crime of passion that is totally unpredictable.

 

[Jon]

I read somewhere that "beyond a reasonable doubt" refers to a 95% chance of being right. That is why I chose 98%. So it would fall within the category of "beyond a reasonable doubt." Since we incriminate based on that, why not in this case?

 

[The_Doc_Man]

In the USA, we do not use a 95% standard. We do not generally attempt to put any specific number in place. My issue is that in this hypothetical world, we need to assure that there is a chance for redemption or reconsideration. Otherwise, the first target to be a victim of extreme violence would be the AI.