Learning and Cognition in Financial Markets: A Paradigm Shift for Agent-Based Models

Abstract

The history of research in finance and economics has been widely impacted by the field of Agent-based Computational Economics (ACE). While at the same time being popular among natural science researchers for its proximity to the successful methods of physics and chemistry for example, the field of ACE has also received critics by a part of the social science community for its lack of empiricism. Yet recent trends have shifted the weights of these general arguments and potentially given ACE a whole new range of realism. At the base of these trends are found two present-day major scientific breakthroughs: the steady shift of psychology towards a hard science due to the advances of neuropsychology, and the progress of reinforcement learning due to increasing computational power and big data. We outline here the main lines of a computational research study where each agent would trade by reinforcement learning.

Mesoscale effects of trader learning behaviors in financial markets: A multi-agent reinforcement learning study

“…Our approach: We base our work on our reinforcement learning MAS stock market simulator-SYMBA (SYstème Multi-agents Boursier Artificiel). IN SYMBA all of the agents are autonomous and are endowed with reinforcement learning [52], by which they forecast stock prices and send individual transaction orders to a centralised order book. In a previous publication [53], we detailed the cautious calibration procedure of SYMBA to real stock market data, also see Supplementary Material.…”

Order book regulatory impact on stock market quality: a multi-agent reinforcement learning perspective

“…However, the ongoing role played by machine learning and artificial intelligence in finance [30,33,42,23] changes the epistemological weight of this consideration. Notably, one should especially highlight how reinforcement learning [16,45], with its numerous links to decision theory and the neurosciences [24,28,35,43,22], could impact the use of ABM as statistical inference tools [37], not unlike what has been done with the recent, AI-augmented, order book models [48,10,46].…”