Algorithmic Trading, Artificial Intelligence and the Politics of Cognition

Algorithmic Trading, Artificial Intelligence and the Politics of Cognition

Prof. Dr. Armin Beverungen

Leuphana University of Lüneburg

Division: Humanities / Organization & Management




In this presentation I focus on the changes in algorithmic trading in financial markets brought about by developments in machine learning and artificial intelligence (AI). Financial trading has for a long time been dominated by highly sophisticated forms of data processing and computation in the dominance of the “quants”. Yet over the last two decades high-frequency trading (HFT), as a form of automated, algorithmic trading focused on speed and volume rather than smartness, has dominated the arms race in financial markets. While many hedge funds also specialize in fast trading in microseconds, in contrast to HFT the focus is on smartness rather than merely speed, and on exploiting not so much the plumbing of financial markets in high volume, high speed trading as on exploiting information asymmetries in trade that operates with holding times of hours, days or weeks rather than seconds. Partly as a consequence of the rise of HFT, human consciousness and cognition have increasingly become irrelevant to markets and are ultimately discounted, with a market conceived as a computational device in which thinking is delegated to things. Artificial intelligence of machines is complemented by the artificial ignorance of human traders working in financial markets.


I want to suggest that machine learning and AI are today changing the cognitive parameters of this arms race yet again, shifting the boundaries between “dumb” algorithms in HFT and “smart” algorithms in other forms of algorithmic trading. Whereas HFT is largely focused on data internal and dynamics endemic to financial markets, new forms of algorithmic trading enabled by AI are enlarging the ecology of financial markets through ways in which automated trading draws on a wider set of data such as social data for sentiment analysis. In this way AI challenges existing divisions between algorithmic and high-frequency trading largely concerned with speed, towards forms of trade requiring a wider market analysis usually still conducted by highly skilled human traders. AI is now competing with human traders not only in terms of the speed of operations and the complexities internal to markets, but in interpreting the world beyond the market and how events and developments there affect market dynamics. The work of financial traders here becomes one of working with algorithms as much as it is a matter of being replaced by them.


There are three broad current developments related to AI in algorithmic trading that I want to focus on. First, there is a movement towards automating quantitative trading, that is using computation both for placing orders and for calculating strategies, much like HFT is already automated. To what extend trading here is really automated remains questionable, however, and the industry seems to have recognized the danger of an over-reliance on AI and the need for humans-in-the-loop. Nonetheless, this automation points to a further shift towards a machine-machine ecology in financial markets. Second, the large majority of the large hedge funds today claim to work with AI, and there is a significant amount of exchange between companies and research institutes currently developing AI and hedge funds. These high-profile movements suggest that hedge funds will play a key role in the development and politics of AI in the coming decades, and it suggests that the various kinds of AI for which these researchers have expertise will be deployed extensively in algorithmic trading. Third, there is a significant expansion of the data sources with which algorithmic trading operates and from which it seeks to extract patterns offering trading opportunities, leading to a differentiation of trading strategies. This expansion of the data ecology of algorithmic trading calls for AI for pattern recognition, and it would be impossible for a human cognizer to take all of this information into account, thus challenging further the traditional roles of financial traders while involving a number of other parties, chiefly data scientists, in financial markets.


I want to suggest that to understand the politics of these shifts it is insightful to focus on cognition as a battleground in financial markets, with AI and machine learning leading to a further redistribution and new temporalities of cognition. Importantly, this allows us to understand the make-up of financial markets as constituted by a number of cognizers (both human and machinic), to consider how cognition is distributed between these cognizers (both as conscious and nonconscious cognition), and to explore what kinds of autonomy is given to machines in algorithmic trading. It is perhaps no surprise that AI has been a central aspect of algorithmic trading, and that more recent developments in AI, such as varieties of deep learning, are being adapted in algorithmic trading. What is perhaps more surprising is that conscious, human cognition still plays a central role in situ for all algorithmic trading except its most automated variants.