Em Reishus - Last Words

This project focuses on the juxtaposition of impersonal records of death row inmates with the very emotional contents of their last statements.

The data set used was a list of 533 offenders executed in Texas since 1982. (Unfortunately the list was rather incomplete with many entries missing photographs or information about the crime committed, but we attempted to work around this.) This is a publicly available data set out of which many charts and graphs have already been made, comparing the number of words in each statement, the age or race of the executed offenders, and the category which the statement falls into ("Said Goodbye", "Asked for Forgiveness", etc) among other. However, all the existing visualizations we found felt like further denial of the offenders' humanity. The goal of this piece was to provide a different perspective on the data.

My collaborators (Stefano Cagnato, Julia Oran, and Evan Carter) and I choose to focus heavily on the recorded last statements of these individuals, instead of their demographics and offenses, which had already been heavily investigated and visualized by other sources. We created a chart which measured the similarities between the contents of each statement using R and D3 using topic modeling and word frequency comparisons to find the Euclidean "distance" between each statement. The final product of this is a map which places the number given to each offender by the original data set on a plane in which the distance between two points represents the difference between the statements. The closer the points, the more similar the last statements were.

In the top graph, the colors represent race, as you can see there was nearly no correlation between race and statement similarities so I choose to work further with a version that showed the individual's id in our dataset to aid in identifying their particular last statement. I've highlighted two subgroups of closely related statements which I choose to dive deeper into later on.

Humanizing the Data

I felt the above presentation did not do much to further our aims of humanization. It appeared that a representation of similarity between all available statements does display much humanity, or create much sympathy for the executed offenders.

First, I wanted created a composite image using a short Python script to represent the face of the hypothetical death row offender to showcase their humanness. Unfortunately, because of the inconsistencies within the record keeping and photo-taking, this ended up looking more ghostly than human.

I realized that any sort of composite display of these death row inmates intrinsically dehumanized them by denying their individuality. I then turned back to the graph made with my classmates.

I selected two groups of nearby statements at random to examine more closely. I now had two data sets which I looked at separately and in more detail than the whole, the red set and the green set. With each set, I created a word cloud using R, which (although a somewhat cliche visual) was an easy way to showcase the major topics and themes represented in each set of statements and show how the clusters differed. The top-left red cluster is first, the wider green one is second.

I wanted to explore these subsets further and explore ways to turn these somewhat cheesy word clouds into a more powerful visual. I then created a series of images for each set; I wrote a Python program to resize and overlay each inmate's last statement with their provided image. Examples of the outputs are below, the first three are examples from the red, left cluster and the second three from the green, right cluster.

Finally, I created an installation for each of my chosen sets. I printed a large-scale version of each word cloud and surrounded it with all of the overlayed portraits associated with the cloud. This allowed viewers to see the face of each executed offender and read the exact text of their last statement, while still visually representing the overarching themes of each group's statements. A side of effect of this method of display was that it also showcased the inhumanity of the record keeping in contrast with the heart-wrenching nature of the content and it revealed that some of the executed men did not have on file photographs (which I represented with a blank grey square) or had their photos embedded only in forms alongside their basic information such as age and height. Above all, I hoped to highlight both the individuality of each executed man and the things his last sentiments had in common with the other men depicted. I had both visualizations on display in the Logan Center for the Arts for several weeks (images of this exhibition are below).

The task of creating a data visualization which managed to capture the humanity and emotions of it's subject was harder than I originally assumed. Treating people as data automatically removes some of their individuality. After several false starts, however, I believe I managed to create a powerful visualization that manages to convey both individuality and similarity.