Machine learning and social theory: Collective machine behaviour in algorithmic trading

Abstract

This article examines what the rise in machine learning (ML) systems might mean for social theory. Focusing on financial markets, in which algorithmic securities trading founded on ML-based decision-making is gaining traction, I discuss the extent to which established sociological notions remain relevant or demand a reconsideration when applied to an ML context. I argue that ML systems have some capacity for agency and for engaging in forms of collective machine behaviour, in which ML systems interact with other machines. However, ML-based collective machine behaviour is irreducible to human decision-making and thereby challenges established sociological notions of financial markets (including that of embeddedness). I argue that such behaviour can nonetheless be analysed through an adaptation of sociological theories of interaction and collective behaviour.

Alternative data and sentiment analysis: Prospecting non-standard data in machine learning-driven finance

“…It is no longer mainly considered a mere fad, although the label 'alternative data' is still shrouded in hype, and an interesting area for quirky quants' wild experiments. Instead, alternative data are becoming integral to investment processes in many firms, which resonates well with the general data turn in the industry, propelled by decades-long processes of computerization and automation as well as more recent widespread adoptions of artificial intelligence and machine learning (Borch, 2021;Hansen, 2020Hansen, , 2021Hansen and Borch, 2021).…”

“…One of the challenges of using social media data as a predictor of the prices of securities concerns manipulative behaviours. For conventional market data, this type of problem might materialize in so-called ‘spoofing’, where a market participant (or their algorithm) submits a large number of fake orders to give the impression that the market is about to move in a particular direction and then exploits this (on spoofing, see, Borch, 2021; MacKenzie 2019; 2021b). For example, a spoofer may sit on a pile of stocks that they want to sell.…”

AI ethics and systemic risks in finance

“…Herding might also arise because technological know-how is quickly spread across markets-through transfer of employees, reverse engineering and copying of successful algorithms-propagating the use of similar tools. Third, AI applications could lead to stronger interconnectedness of human and algorithmic market participants through new types of contract and relationship [9,72]. For example, WEF [9] envisages the possibility that AI systems autonomously learn to collude with each other.…”

“…More generally, financial AI applications, which are programmed to guess and outsmart each other, are based on mutual observations and are thus prone to locking in their actions, leading to herding, disastrous resonance and tail events. Hence, the interaction order of algorithms, or collective machine behaviour, becomes central to better understanding systemic risks in markets [23,72].…”

“…5 On top of this, the limited understanding of how AI algorithms interact with each other and with human participants makes markets themselves a highly complex black box with rapidly shifting interlinkages and risks. It is difficult to anticipate how AI-based algorithms will act collectively due to feedback loops, 6 emerging (self-organizing) properties and tight couplings [72]. Thus, AI programmers and users might be unsure about the causal contribution of their actions to the negative collective final outcome which is the very subject of moral evaluation.…”