Abstract
Both my technical report and my STS research paper examine autonomy in high-stakes environments where technical performance alone is not enough to determine whether a system is successful. My technical report focuses on improving the perception stack for the Cavalier Autonomous Racing team, while my STS research paper focuses on responsibility and trust in pilotless commercial aviation. While these projects deal with different forms of autonomy, they are connected by the same broader issue: autonomous systems depend not only on engineering capability, but also on the surrounding structures that make their decisions usable, trustworthy, and responsible.
In my technical work, I saw how autonomy depends on much more than a single model or algorithm. A perception system may appear to be a technical problem, but its success also depends on data quality, calibration, labeling, evaluation, deployment, and how well its outputs fit into the larger autonomy stack. This shaped how I approached my STS research paper, where I examined pilotless passenger aircraft not simply as a question of whether planes can fly without pilots, but as a question of how responsibility is redistributed when a human operator is removed from the system. In both cases, autonomy creates a larger network of responsibility around technology. The technical system may perform the immediate task, but humans and institutions still remain responsible for how that system is designed, tested, trusted, and used.
The connection between the two projects is that both show how autonomy shifts responsibility across a wider sociotechnical network. In the CAR perception stack, a poor output could come from many different places, including sensor limitations, weak calibration, poor labels, model failure, or integration issues with the rest of the autonomy system. Similarly, in pilotless aviation, an accident would not be easily attributed to one pilot’s decision, because responsibility would be spread across software developers, aircraft manufacturers, operators, regulators, and certification bodies. In both cases, automation does not eliminate human responsibility. It redistributes it.
Working on the technical project helped me better understand the argument I made in my STS paper. Autonomous systems can appear to be purely technical from the outside, but in practice they depend on many hidden decisions about testing, deployment, monitoring, and trust. My work on CAR made this clear because perception performance was not just about improving accuracy, but about building a system that others could understand, evaluate, and continue developing. My STS paper extends that same lesson to commercial aviation, where pilotless aircraft may eventually become technically capable, but will only be socially acceptable if the public can understand who is responsible for their design, operation, certification, and failure. Together, these papers show that the future of autonomy depends not only on making systems work, but on making them accountable within the larger sociotechnical systems they enter.
