Blog 3: Exception to Data Driven Rules

This case study examines the role of the data-driven exception in decision making, and the importance of acknowledging any individual as one. Because of unpredictable uncertainty, the study argues that everyone has the right to be a data-driven exception (a.k.a. examined as an individual) and shouldn’t be expected to conform to larger statistical averages or accept the consequences of not adhering.

Case Study:
The Right to Be an Exception to a Data-Driven Rule

Discussion Responses

What is a data-driven rule, and what does it mean to be a data-driven exception? Is an exception the same as an error?

A data-driven rule is a decision made based on acquired data to better determine an outcome usually benefiting the majority, whereas data-driven exceptions are essentially outliers to that information, and are more often than not harmed instead of helped when data-driven decisions are made. I don’t believe that exceptions are at all similar to errors as people should not be expected to conform to societal standards. Rather, I think the true error is in data-driven rules being made without the consideration of exceptions.

In addition to those listed above, what other factors differentiate data-driven decisions from human ones?

Data-driven decisions can only be made based on the information fed into it. Machines also do not contain any sense of nuance, which is to say that so long as it’s deciding based on the exact same data fed into it, it will make the exact same decisions on candidates with the same statistics. This is to say then that human decision making can be roughly but never perfectly predictable versus data driven decisions being easily predictable when ignoring uncertainty. It’s why employers conduct interviews: they’re not going to hire people solely based on their resume—they’d rather get to know applicants as human beings rather than data points. (Well, at least smaller companies do).

Beyond what is discussed above, what are some of the benefits and downsides of individualization?

Individualism has the benefit of acknowledging more exceptions the more specific it gets, though I imagine that the more hyper-specific data gets, the less a “one size fits all” decision can be found. Now, I do think this is partly a good thing as it would mean exceptions are being accounted for, but I can also see an importance to note that it would become very difficult to appropriately make any large scale decisions since no data-driven majority will really exist (since no one is exactly the same). If trained consciously though, data-driven decisions have the potential of ignoring negative cognitive bias present in humans, and being able to identify people impartially.

Why is uncertainty so critical to the right to be an exception? When the stakes are high (e.g., in criminal sentencing), is there any evaluation metric (e.g., accuracy) that can justify the use of a data-driven rule without the consideration of uncertainty?

Uncertainty is critical because no matter how many data points you have of a person, they can never be fully predicted. In regards to criminal sentencing, I believe that the punishment should match the crime only, and not the person’s statistics (unless that happens to involve past criminal history). Like the study mentions, if someone receives a harsh sentence (like the death penalty) resulting from suspecting future criminal behavior, it will never really be known if that criminal behavior will actually occur. Future data-driven decisions would also be skewed by this behavior, as it will default to harsher sentences because those almost never have worse consequences than lenient sentences. Because this accuracy can be so easily skewed by such outcomes, I don’t believe uncertainty can ever be ignored when making data-driven rules.

My Own Discussion Question

In what ways could decision makers be negatively impacted (socially, legally, etc.) by failing to recognize individuals as data-driven exceptions?

I really like this question because it can help people to realize that data-driven decisions can harm both exceptions and decision makers. It also leaves some room for interpretation for readers since it’s only touched on very little in the case study—specifically when talking about criticism around minimum sentencing laws. This leaves lots of room for interpretation and synthesis since I felt some of the other questions limited the scope by saying “beyond what was discussed” when they already discussed a lot.

Final Reflection

This post felt quite a bit different I suppose because it was both a pre-chosen article and also discussing a much broader subject than a single current event. The reading was actually quite interesting despite its length; definitely gives me an idea of what our case studies might look like later in the semester. As for the discussion questions, I didn’t always feel like I had a good real-world example (I thought multiple times of bringing up lawmakers but nowadays it feels like their decisions are purposefly harmful) so I was more broad with some of my answers, but beyond that I feel I was able to answer them all well. Generally this assignment made me feel like I was coming away with more knowledge than applying what I already knew which I liked since I still got to do both in the end.