Abstract
The world of financial markets is seeing a dramatic transformation as Artificial Intelligence (AI) finds its place in high-frequency trading (HFT) platforms. These systems use algorithms based on machine learning to learn about pattern data in the market and execute trades at velocities orders of magnitude beyond those achieved by humans. While AI-based trading holds various technical advantages, such as better market liquidity, tighter bid-ask spreads, and the capacity to take advantage within microsecond windows of its position, it brings along equally significant new risks in its wake: increased short-term volatility, opacity in decision making, and an increase in market power concentration among few large players. The fundamental problem tackled in my thesis portfolio deals with advancing AI's technical capabilities for finance while meticulously taking into account the broader social, economic, and regulatory implications of such technology. My research includes an STS research paper created in conjunction with a technical proposal outlining an optimized design for machine learning algorithms for use in HFT environments.
The technical report recommends developing a machine learning system to increase speed and accuracy in high-frequency trading models. The report highlights key issues in algorithmic trading, including overfitting of models, latency, and requiring models to be highly adaptable in changing markets. Drawing upon recent developments in AI, the proposal examined ways in which deep learning, reinforcement learning, and real-time feature engineering can be leveraged. In addition to enhancing speed and scalability, the report recommends hardware upgrades such as GPU acceleration and parallel computing. Some of the other concepts were cross-validation procedures to reduce risk of overfitting and to condition models to perform better in various market scenarios. While still in theory and without any algorithms in practice, the proposal lays down a good foundation for future development and simulation. One of the key points in the report is that such technologies, should they become available for widespread use, would enable smaller players in markets to better compete against larger, better-funded institutions by reducing computing barriers to market participation.
The STS research project is grounded in an underlying technical approach for examining the systemic risk and financial imbalances brought about by AI-powered high-frequency trading (HFT) platforms. Using Actor-Network Theory (ANT) as its conceptual frame, the paper explores the intricate and often non-linear intertwinement of market institutions, human actors, machine learning algorithms, and regulative regimes. Through an exploration of critical events such as the 2010 Flash Crash and the 2013 Twitter-driven decline in the market, the paper explains how intertwined algorithmic frameworks can create feedback loops provoking increased volatility and temporarily reduced liquidity. Such events highlight an underlying susceptibility: when AI frameworks react in unison to signs of market disarray, they also run the risk of draining liquidity or performing similar trades, thereby endangering the infrastructure that grants them an advantage.
The research also explains further its socioeconomic implications involving AI-powered high-frequency trading (HFT). A limited number of large players, such as Citadel Securities and Virtu Financial, have disproportionate volumes of orders under their command by virtue of using superior algorithms, high-quality data feeds, and strategic server proximity. This technology leverages advantages allowing for exploiting temporary inefficiencies unavailable to other market players. In reaction, various regulation mechanisms, such as circuit breakers and the Limit Up-Limit Down (LULD) paradigm, have set an initial level of protection. However, this article argues that such measures remain in their infancy stage while largely reactive in nature. Different regulatory standards in various jurisdictions further fuel non-globalization in regulatory practice. Such differences at different locations create potential for regulatory arbitrage, allowing companies to conduct operations in areas of less restrictive regulation. Simultaneously, retail investors, who often make use of apps for free trades, tend to remain uninformed about routing trades and potential extraction of value by faster intermediaries. Together, the two proposals offer an integrated understanding of the two faces of artificial intelligence in finance, its ability to increase efficiency and promote innovation, as well as its possibilities to widen disparities and add systemic risk. While the technical proposal outlines recommendations for its responsible development of better algorithms, STS research lays out the need for regulation, ethics in design, and openness in governance for these technologies. The portfolio suggests the need for collaboration among technologists and policymakers. Financial innovation should not compromise fairness, stability, or public trust.
In coming days, future scholars have been recommended to research using explainable artificial intelligence methods in high-frequency trading (HFT) platforms to address the lack of understanding in current algorithms. Improved explainability in models would not only increase regulatory supervision, but also increase market clarity. In addition, there is an urgent need for greater overseas cooperation in setting regulatory requirements which can be facilitated by better cooperation among financial supervisory agencies and supranational institutions. Introduction of real-time monitoring mechanisms, increased data transparency, and market structure reform, such as pushing for using time-period batch auctions or implementing "speed bumps," could also mitigate risks related to high-speed trading activities. Financial institutions should serve a greater social purpose than serving only those who possess a technological advantage, requiring an interdisciplinary approach for reconciling these advancements and social responsibility.
