Abstract
My technical project and STS research examines physical hardware technologies that are relevant in our daily lives. For my technical project, I worked on an autonomous Jenga playing robot using the Dobot Magician robotic arm. The technical project dives into various subsystem hardware and software designs that were developed to make the goals achievable. On the other hand, my STS research examines Palantir and government surveillance on marginalized communities. A commonality between the two projects that I work on is that both are data-driven systems (computer vision and integrated data platforms) and their impacts, whether it's for a playful and educational context to a high-stakes community impacted surveillance. Within the technical project I will describe the design, implementation, and testing that comes from the project. For my STS research, I critically analyze and use Actor-Network Theory to explain the surveillance network and its harms on marginalized communities.
The goal of my technical project is to develop a computer vision-integrated robotic arm system that can autonomously play Jenga, aimed as an educational model for teaching robotics and perception. Jenga is a great way to test and showcase robotic skills because of its fine motor controls, perception, and sequential decision making. It also makes for a fun way to teach important ideas about how robots see, move, and stay safe. The system is built with a small robotic arm, the Dobot Magician, that has four degrees of freedom. Because of its limited reach, special ways were needed to communicate with it and plan its moves, including designing the game so it works well within its space and capabilities. The robotic system contained three primary hardware subsystems including an Intel RealSense D405 camera, a motorized turntable, and an end effector with a force sensitive resistor sensor. The physical hardware subsystems were then centralized into a single control station software application that used a multithreaded Model-View-Controller architecture. My team and I used a lightweight edge based detection system to identify the blocks as well as a custom calibration routine to allow the Dobot to work within the constrained workspace. The overall system worked in a control loop cycle that goes through probing, pushing, pulling, and placing states. Testing showed a strong unit level reliability but significantly lower success in various fully autonomous closed loop games which may come from the mechanical tolerances and cheap force sensing hardwares. Despite such challenges, the robotic system successfully helps demonstrate a full Jenga playing robot.
My STS research analyzes the societal impacts on marginalized groups of government use of Palantir’s platforms for predictive policing and immigration enforcement. Palantir had its initial development come from the war on terror, after the attacks on 9/11. However, Palantir has derailed from its initial counterterrorism efforts, with their technologies expanding into policing and immigration enforcement. This growth is concerning as many minority, specifically immigrant, populations are under attack from Palantir’s technologies and the our government that uses the technologies. Within my research, I use a semi-systematic literature and document review across technology, governance, and sociology, looking at predictive policing, Immigration and Customs Enforcement, and contractual documentations. I look at specific case studies as well from the Los Angeles Police Department and the New Orleans Police Department with their use of Palantir’s technologies. Within my research I use Actor-Network Theory to treat Palantir, government agencies, and marginalized communities as interconnected actors within an entire network. Understanding how each actor is being leveraged and used highlights the core ANT concepts of black boxes and punctualization. Palantir systems integrate diverse data sources that come from all aspects of our daily lives into a singular deep profile. These profiles even extend surveillance to those with no criminal justice contact. In addition, predictive policing and immigration enforcement disproportionately target, Black, Latine, and immigrant populations, reinforcing racial and class inequalities and contributing to family separation, health and educational harms, and psychological hyper-vigilance. The continued use of Palantir’s technologies from our government can lead to real world implications of threats to our very civil liberties. This calls for the need of transparency and accountability within the system and network that is cast upon us.
Working on my technical project that involves systems like computer vision, sensing, and integrated control loops has connected techniques that also show up in surveillance and data integration platforms like Palantir’s Gotham. However, there is a contrast between my low stakes educational Jenga playing robot with that of the high stakes implementation of surveillance technology. My experience in designing my technical system has allowed me to be more attentive to design decisions like data collection, object segmentation, and subsystem integration. My STS research reveals that even larger complex systems, similar decisions can be made to encode bias and further allow this dominance in surveillance. Both projects have allowed me to gain insight on my impact as a future engineer, on both a small or major scale.
