Following our discussion on Monday, I have been thinking a lot about one of the comment that Professor Rodrigues mentioned at one point – the best time to be scheduled for sentencing is early in the morning or after lunch. While I had seen lawyers on TV and such trying to get these times, I still was still associating this more with the realm of fiction. It is incredibly alarming that whether or not Judge X has had their pastrami sandwich from Frank’s determines whether or not and to what degree somebody will get probation, fined, or imprisoned. A lunch, or lack thereof, should not be what alters the course of somebody’s life. With regards to decision-making, this is one way where I see how technology such as this could potentially make a difference. To my knowledge, an algorithm can’t get hangry or cranky. However, it is also clear that a biased algorithm with a less than stellar accuracy level cannot be the better option. In my discussion group, we briefly touched on an ethical dilemma that could result from this. What if the algorithm got to a point where it was better and less biased than judges at predicting recidivism rates, yet still not completely accurate? Would it be better to use it or not?
As TJ said in class, it is also crucial that people who are responsible for creating this technology take responsibility and work to make it less biased. I don’t know if this will be possible soon, though; the judicial system, being already extremely racially biased, influences whether or not people can get jobs, which affects the algorithm by tying having a job or not to the risk score created. Thus, the fastest way may not be turning to technology, but instead paying more attention to the people making the decisions.
Kate nicely pointed out the pivotal of the ethical dilemma within using technology for judiciary decisions. Letting artificial intelligence to decide and sentence one is good if and only if the A.I. is perfect and no biased. However, it is not possible because the way that the A.I. deep learns something already involves biased data and information around the world. It is only possible in the utopian society where everything in the earth is fair and not biased, which cannot exist in the real world. Thus, as Kate said, the only possible way to achieve the best output is to give more attention and care more.
I’m wondering if it is possible to create an algorithm without any shortcomings considering that whoever the creator will be will certainly be human. On some level I feel as if an algorithm will be better than the humans as they are now, as long as we keep in mind their imperfection, monitor their decision making processes, and improve things along the way. Seeing as I don’t believe in a perfect world, a mindset of constant improvement is, I think, the best we can hope for.
On the other hand there are a lot of ways in which we approach punishment in the justice system that can use complete reform. That is to say, rather than cook up ways to do the same thing we have been doing but slightly better, perhaps a more pertinent conversation would be how the way we approach justice may be completely misguided. We still have a lot within the justice system on a systematic level that merits large scale reform and this might not be the right time to introduce any sort of automation until those things get sorted out.
The point about whether a more accurate algorithm would be better than judges is something I thought about when I chose this article. It might certainly do a better job of adjudicating cases, but the removal of the human element removes possibilities to view defendants within the circumstances and context of their cases. On the other hand, judges obviously still mess up quite a bit. All of this points to the fact that our criminal justice system is inherently corrupted, and that any element added in the name of efficiency is a bandage at best for something that needs widespread reform.
Charles raises perhaps one of the most important questions of the digital age which is how much can we rely on computers to do a job that is also subsumed with so much human error. There is no answer. One might argue that at this point they both cripple each other and it is likely impossible to ever create something that is able to work without these large blindspots because the people who create them suffer from them as well. Of all the job AI and machine learning might take over in the future, it should likely stay the furthest away from law and justice matters.