Critical Making: Moving Data to the Physical

Jessica:

In the introduction and the original conception of this project, the goal was to have a representation of Latinx deaths from police killings across the United States. However, as the math continued and we mapped out all four years in our trial state, California, we quickly realized we needed to change something. At the beginning, we distilled the embodied effects and points/messages we want the viewer to experience and leave learning, and, on some level, some of these points have been non-negotiable. The main one has been the anticipation built in by all four years falling at different intervals. The second has been the largeness, and that police killings are not rare or occasional but greatly affect the community.

With these two in mind, when I fixed the mistake I made in the original four-year data distillation, a problem arose (interdisciplinary work comes with a learning curve). The mistake was that I forgot the element of anticipation: when I broke the four years into age groups and their counts to figure out the quantity and pin size, I didn’t consider that the years would create different groups and that I’d have to break down the data twice. And in the version with error, all the pins fit into California. But then, in its correct form, it did not. At this point, it became a question of scale and time.

So, What Has Changed

After staring at my map, we realized that to maintain both anticipation and scale, as we wanted to, we had to change the map’s scale. We went back and forth on all the different things we could change to make the pins fit, from quantity per pin to having each pin size represent a range of victims; for example, a 7 mm diameter pin (which used to be 4 mm, but this too had to change due to a different mechanical issue) originally represented 1 victim, and we discussed having it now represent a range of 1-5. We also considered exponentially decreasing the number of pins to four per affected U.S. state to represent the years and keep the pins falling at different intervals, along with other ideas.

Each idea came with different things we might lose or gain. And it felt like we were trying to find an ideal effect. I don’t think I have ever had such a serious conversation about pins before this semester. But the conversations were extremely helpful in remembering and reorienting the story.

All to say we finally landed on shifting the representation of Latinx deaths from police killings across the United States to just California. Why California? For multiple reasons. But from here, we did have a conversation about whether, in representing just one state as opposed to the entirety of the United States, we could maintain the scale of deaths that occur in the United States.

Additions With Change

Narrowing the map to California offered unexpected inspiration and additions to the project. Ammon pointed out that focusing on just one state gives us another way to narrow it further, and what we do from here could be more specific than broad data allows. In a national map, or with data in general, you always risk dehumanizing people and reducing them to points on a map. Within the construction of the U.S I didn’t really know how to negotiate that, and it was a question I had been wrestling with for weeks, and in my work more broadly, often. But California has allowed me to have more room to be a bit more creative.

A question I didn’t anticipate, but that came up when I put the data into its first physical form, is what gets lost/added in the visualization process, and when you adapt the model to better fit the required changes, what are you negotiating, shifting, or gaining? These questions, along with all the other conversations, have been teaching me how important the non-negotiables and the story are. Each version tells a different story. The one we make will be based on the final story we want to share.

Ammon

About the Physical Changes

3D modeling has the same issue of scale that we saw in placing the data on the physical map. In the software, you can make tiny things look big. I ran across this issue as I modeled a 4mm diameter pin and the mechanism to keep it in place and then let it fall. I could zoom in, make the design, figure out the mechanics, and readjust measurements. And it all looked great! But then after 3D printing the pins and the mechanisms, I recognized that they were delicately small. The holes for the screws to attach the mechanisms (which I neglected to account for while modeling) were woefully inadequate. Too small and fragile. I would have needed to use a sewing pin instead of a screw. The idea is to have a string pull on these mechanisms to allow the pin to fall; and these tiny mechanisms would not withstand a single tug, let alone repeated pulls.

By reconsidering the scale of the project, we were able to increase the size of the pins to allow physics to work in our favor by making the parts actually usable. And this is all, again, a great lesson in iterative creation. It takes repeated attempts of thinking through ideas, designing, modeling, making, and testing to create something. Some people call this failing forward, or learning from mistakes. I like to call it just plain learning. You can’t learn without repeated attempts. Failure is only when you stop trying. Mistakes are just learning how things work. Reflecting upon this learning process is part of what makes this Critical Making.

Critical Making Fellowship 2026-2027

Jessica writes:

Within the United States, the racialization and criminalization of Black and Brown bodies have been and continue to be a violent process. One shape and form of this violence is police violence, more specifically, police shootings. Different articles, books, scholars, and organizational databases have gone about examining and tracking police killings, including but not limited to: Washington Post’s Fatal Force Database and Mapping Police Violence.

This year-long fellowship and project is a data visualization project using data from the Washington Post’s database, with a focus on the Latiné people killed in police shootings.

The Visualization Itself

The current idea: there will be a cutout of the United States, and using a pin model, four years of Latiné deaths across the states will be physically represented. Breaking the data into grouped years, the model will drop each year’s number of victims, year by year, with different points of information represented by the pins’ form, color, height, and grouped quantity by diameter.

Focus of Anticipation:

The pins will drop by year and hang suspended. Imagine a map base with pins standing on top, but the base will be embedded with gaps/holes that the pins will fall through when released and stay suspended, so they don’t fall completely to the ground but hang on the mirrored side of where they once lived.

Take, for example, the year 2015, which has 176 recorded victims of police killings; the distribution of the 176 persons will simultaneously fall while the victims of the remaining three years remain standing until the year has passed. The pins dropping down at different intervals represent the ways Latiné death is anticipated and almost inevitable. Latiné bodies within the United States anticipate the potentiality of violence, the knowledge that racialized bodies are not and have not been safe within the nation-state.

Along with this lived sense of precarity and anticipation, the hope is that the viewer who sees this visualization and its process will notice and feel the growing sense of anticipation, though it will differ from the embodied one the community itself feels.

A question here: why not just let the pins drop, and why keep them suspended? Beyond the practicality and efficiency of restarting the board to all four years standing, I am working and inspired by Latiné writing and artistic production of death. It may sound odd, but the idea of the suspension comes from a mixture of Mexican and Chicanx celebration of the Day of the Dead and a poem by Jose Olivarez.

The idea is that underneath the main structure and base of this nation, the lives of the dead remain. The nation-state stands on top of these deaths and is built on this history.

Design and Prototyping

What have we tested and thought through so far? With any prototype, the minute you think you narrow down one detail, a thousand more questions arise. Theoretically, to aid in the visualization of different data elements we have discussed what about the pins can be transformed to make different points.

For example, the height of the pin will represent the different age groups largely affected by police killings; the diameter, the quantity. Where one year has 176 victims, imagining filling even one map with 176 pins felt nearly impossible, not even to consider a four-year range. To consider that the diameter of the pin will represent quantity groups.

1 victim – pin of 4 diameters 5 victims – pin of 7 diameters 15 victims – pin of 10 diameters

Along with height and diameter, the pins will have two different color groups for sex: male and female.

Questions We’ve Encountered

As we began designing and isolating the data, each detail of the prototype raised new questions. With any data visualization projects there are certain audience takeaways you try to achieve. With maps, the process raises more and more questions about geographic exactitude versus approximation. Along with that, you ask the prototype which parts of the data you can visualize at the same time versus what gets lost or is prioritized differently along the way.

Ammon starts here:

A Pedagogical Approach

We started thinking about this project by beginning with the end in mind. I have found that bigger projects like this benefit from taking this approach first, as it narrows down the plethora of alternate paths that are available at the start of a project.

We started by thinking about the project being all done. It’s created, installed, and ready for people to interact with it. Now imagine someone has just finished interacting with the piece. The answer to the following two questions guide our design and development of the project as we go forward. What emotion do you want this person to have? What knowledge should they walk away with?

The answer to the first question was: Anticipation. They should feel a bit of anxiety and trepidation as they anticipate which pins will fall next, and how many will fall. Who’s next?

The answer to the second question was: The concentration of deaths occurs in the border states. With those two questions answered we could delve into how best to represent the anticipation and the knowledge. We came up with a few different alternatives but always came back to the idea of pins dropping through the map. As Jess explained above, the suddenness of some of the pins dropping away leads to anticipation for the next set of pins to drop. Where? How many? Who?

Limited linguistic inclusivity in public health communication leaves many South African communities underserved, particularly regarding critical information on zoonotic diseases such as rabies. This pilot study addresses this gap by developing and evaluating AI-driven methods for delivering reliable rabies information to Sepedi speakers, a low-resource language group. The study presents a novel, curated Sepedi dataset of 60 question–answer pairs, created through a systematic pipeline: thematic analysis of authoritative English sources guided the synthetic generation of QA pairs, which were then translated and manually verified by a native-speaking expert. This dataset was used to compare two large language models, GPT-4o and Gemini-1.5 Flash, under both base and fine-tuned conditions. Evaluation used a human-centred rubric assessing fluency, accuracy, and cultural appropriateness. The findings reveal a key nuance in applying LLMs to low-resource domains. The base GPT-4o model, with strong foundational multilingual capabilities, outperformed all other configurations, including its own fine-tuned variant.
In contrast, fine-tuning provided a marked improvement for the less capable base Gemini model. This result indicates that fine-tuning can enhance weaker models; its benefits are not universal and may be outweighed by the strong zero-shot performance of state-of-the-art architectures when training data is scarce. The curated Sepedi rabies QA dataset will be released under an open licence to support future work in low-resource public health communication.