Taking on the responsibilities of a judge felt daunting, even if it was only for a simulation. Knowing that the fate of one defendant could affect the lives of many people meant I had to make informed decisions with only undefined risk scores and vague statements. Upon receiving my first case, many questions appeared in my mind before I could make a choice: What were the definitions of “low,” “medium,” and “high” risk individually and in relation to each other? How were these “facts” obtained? Was the prosecution’s statement reliable? Why did they recommend a specific outcome, aside from the projected risks? I often found that those succinct statements ended up hindering me more than helping. I also wondered if the obscured picture of the defendant was necessary or if it was there to sniff out potential bias – while no defining facial features could be made out, the skin tone was still implied. Given the lack of answers to my questions, I had no choice but to blindly trust the information and apply my own subjectivity in order to push through this Herculean task.
The “jail capacity” and “fear” features immersed me in the situation, added pressure, and prevented the “cheese strategy” of detaining everyone. I did my best to ignore the pictures and the age of the defendant as I felt they would not meaningfully contribute to my decision-making process. The seriousness of the crime was taken into account. For example, I mentally categorized drug trafficking higher than motor theft (easier said than done). If I felt that the defendant’s statement provided insight into their character such as an indication of family to take care of, then I also considered that key piece of information only in combination with lower risk scores. These guidelines were loose and still resulted in no-shows or second crimes from certain individuals even if the predicted risk was on the lower end of the scale, but they at least helped me not to dwell on each case for too long. This was the goal because I wanted to reach the end and see how each decision played out. All in all, I wish I had been given clear definitions and rationale for the “low,” “medium,” and “high” designations. It also would have been nice to have more depth in the prosecution statements instead of “We recommend detention / release.”
Going through this simulation made me wonder what consequences have already occurred in the real world when decisions are only informed by AI algorithms where some creators may not even know the full capabilities of them. How many innocents have been implicated, while true criminals have gotten away scot-free? How many lives have been lost because they had an emotional connection to chatbots, leading them to make irreversible decisions? When used in the right way and in conjunction with other tools, these resources can provide assistance, especially with the more mundane and repetitive tasks in our lives. But when they are used as a crutch, critical thinking may go out the window and the extreme risk of harm to oneself or others comes in its place.
