This continues our series of student reflections and analysis authored by our research team.
On the Topic of Efficiency
“Even Too Much Water Can be Lethal”- Edward Tenner
In a database filled with thousands of cases, and reaching far beyond in the amount of sources to back these up, a system of some sort of efficiency must be in place. The 39 page manual, 24 page codebook, an assembly line to quickly scrape documents, and other processes are ever-changing and updating to best organize and analyze data. As no project is error-free, especially one as new as tPP, this week’s work is being done to ensure each case has reputable sources and match with the designated formatting for files sources.
Processes such as this simply ensure that the project becomes more valid and consistent. However, it also points out room to further clarify the system to ensure efficiency. Taking quite a few hours to look through the individual cases of 2009, I found many cases lacking the requirements of sources outlined in the manual, likely never uploaded. This is a sort of expected problem- not necessarily detrimental to the project. Yet, it raises a few questions of different nature.
What does this say about efficiency of technological systems?
As we have seen the advancement of technology simply explode over the last few decades and journey into the fourth technological revolution, we have to question if there are underlying errors in the system- possibly analogous to our source document errors on the project.
As described by Jesus Mena, the use of neural networks is growing. These networks are another form of technological efficiency, like algorithms for example, that are rooted in a biological action. They reflect the way in which the brain organizes patterns and learning and use this idea to build a system that can organize data and make predictions. Mena describes this connection to the brain, seen below:
While I do not doubt the extensive research that went in to develop this system in conjunction with our understanding of human processes, problems arise as our knowledge expands in this area. Learning, as suggested, is rooted in synaptic changes- also known as plasticity. The brain is incredibly talented at accounting for deficits and adapting need be. Another way this is done is through modulation, which is shorter in time and allows for fast changes to a behavior done by one area. This is where I fear networks may not truly reflect this idea. As Mena suggests, using networks will help in organizing and creating predictive models of criminal feature, for example. Can the network truly change as society does- what if the system is multipart and can modulate? While organization seems to be exemplified, there comes a point where this type of efficiency may show error that will be hard to return from.
Is efficiency always the best thing?
As with everything else, there are critiques to systems incredibly efficient. As Edward Tenner describes in The Efficiency Paradox: What Big Data Can’t Do, “even an excess of water can be lethal.” Tenner describes many concerns of efficiency, but one of interest is the idea of hyperfocus. Hyperfocus has increased our skill to recognize large patterns and pictures, but cannot focus on small details. This is of concern to tPP and criminal proceedings. Is there a point where we generalize too much? An issue with prosecution is found in stereotypes, could this happen if begin to build systems that start to generalize patterns?
Does tPP have appropriate error in process?
We must not forget one of the goals of this project is to provide students with skills and an opportunity for creative growth. Amongst error in provides opportunity for innovation. Finding flaws in the process allows for students to really think about the project and how to build systems. In going through all of the cases from 2009 I was able to think about the process, but I also learned a great deal about a lot of cases I had never been exposed to prior. Priority should absolutely be based in producing results that are valid, but maybe a structure that is only 90% efficient is not a flaw at all.
Mena, Jesus. Investigative Data Mining for Security and Criminal Detection. 1 edition. Boston, MA: Butterworth-Heinemann, 2002. [Excerpt, Chapter 6, “Neural Networks: Classifying Patterns”].
Tenner, Edward. The Efficiency Paradox: What Big Data Can’t Do. New York: Vintage Books, a division of Penguin Random House LLC, 2019.