Abstract
Modern financial systems are increasingly shaped by ever-evolving technology. This is primarily showcased through the most popular social media platforms as well as complex trading algorithms. An estimated 60–73% of trades within financial markets are made algorithmically, but with the prolific rise of Artificial Intelligence, this figure is likely far higher. Pair that with the revolutionary upgrade in communication technology that is social media, in which people can discuss the trade of assets at a scale and speed that dwarfs anything preceding it. The result is a totally new financial system. Both markets and regulators have been, and still are, actively unprepared to go through such a drastic change, as financial systems were relatively uniform for much of market history compared to the digital era. It follows that, due to the lack of proper regulation, a number of issues have arisen. I will be focusing on value gained through the manipulation of trading algorithms or the manipulation of confused or unaware individuals. Beyond this, the layers of complexity often obscure the chain of responsibility when events like this happen, leading to the biggest issue: the lack of transparency within the relationship between trading algorithms, social media platforms, and AI-generated media. Because of the lack of transparency in these systems and algorithms, it is very difficult to punish bad actors, thereby leading to more foul play.
To address a similar problem, my technical capstone focuses on the development of a personalized speech deepfake detection web application. This system combines AI-generated audio detection with speaker recognition to help determine two things: whether an audio clip is real, and if it matches an individual’s voice. This is especially important as deepfakes become more accurate and accessible, increasing the risk of scams and impersonation.
It is important to consider the human and social dimensions of this technology, as they are the driving force behind its creation. The end goal is to help protect individuals and institutions from deception caused by falsified media. The broader social consequence of this goal is a more reliable and safe digital environment where people have better tools to evaluate whether the information they encounter is real, or falsely attributed to someone else.
In my STS research, I am analyzing the relationship between social media discourse and algorithmic trading through the lens of Actor-Network Theory (ANT). This framework puts both human and non-human actors on an equal playing field within a particular network. Because my topic highlights the breaking of symmetry between humans and machines, ANT was the clear choice. An overarching theme of my research is that the impacts on this network are never made by an individual actor. Market movements are always driven by some combination of human and machine, likely a staggering amount of both influencing each other. This idea is adopted from a core tenet of ANT, which led me to it as my choice of a primary framework.
I am conducting my research through a combination of case study and literature analysis. Past incidents such as the flash crashes of 2010 and 2015, as well as the GameStop short squeeze, present useful cases to reverse engineer. While it may not be possible to look deeper into the inner workings of the most important trading algorithms, we can see how they function by analyzing market behavior during past incidents in which they were manipulated in some way. Through this analysis, I plan to identify a clearer link between market movement and social media discourse, while also examining key moments where the market was manipulated by actors who understood how to abuse this relationship.
The implications of the synthesis between my STS research and technical project are an exploration of digital transparency and accountability. My research examines how algorithmic trading and social media can intersect in ways that obscure responsibility during destabilizing market events. My technical project addresses a similar issue by creating a tool that helps detect fraudulent audio and verify speaker identity across teams. Together, both projects show how algorithms and artificial intelligence can make responsibility harder to trace, whether in financial markets or digital communication. Both topics highlight the need for better governance policies in an age where technology develops faster than humans can react.
