This multidisciplinary project brings together insights from psychology, philosophy, anthropology, complex systems, behavioural finance, artificial intelligence, and computer science. Inspired by the foundational work of Simon, Popper and contemporary thinkers from fields such as evolutionary biology, we explore the frontiers of scientific modelling in the behavioural social sciences (particularly economics). We investigate the implications when the scientist is an evolving machine with nearly unbounded rationality and augmented cognitive abilities.
The applied science approach uses empirical datasets from macroeconomics and finance. The principal modelling technique is Genetic Programming via Symbolic Regression, which optimizes simultaneously for error reduction and complexity minimization (simple and accurate models survive). In our research, we focus primarily on two capabilities of this technique: the ability to find driving variables, and the possibility to discover trustability metrics and uncertainty ranges by constructing ensembles of models. We also formalize the frameworks required for multi-target optimization, e.g. when crowds have to simultaneously optimize for profits and social impact of their investment activities. The focus of our application of the evolutionary algorithms is on Understanding vs. Prediction: finding The Why.
Analytic Philosophy of Financial Economics via Evolutionary Symbolic Regression
Core questions
Applied Genetic Programming for Macroeconomics and Finance
Using merged financial and alternative data sources (web traffic panel data, ESG metrics, social network graphs, and blockchain activity), we explore both stock markets and alternative assets such as private equity and digital assets. We inquire about issues at the convergence of Economics, Behavioural Science and Philosophy of Science:
It has been established that non-experts (retail investors and internet crowds) can move financial markets. But, are all non-experts equal? Furthermore, can the meta-modelling process that is used to map semi-decentralized market participants' behaviour work as a quantitative estimation of creativity in scientific research?
Co-Evolution of Machine Cognition and Attention Markets
Attention flows are treated as the fabric of the macroeconomic construct, not simply as an "attentional bias" as in traditional behavioural economics. Since the evolutionary algorithms use inferential sensors (e.g., click stream panel data) akin to perception and show creativity akin to imagination, we approach the inquiry from a machine cognition standpoint.
We build on two novel areas of research: cognitive economics, which is defined as the economics of what is in people's minds (Kimball) and the attention market (both in the microeconomics sense, i.e platforms, and the broader attention economy sense). We investigate the feedback loop between algorithms and the crowd's tacit knowledge, implicit confidence, and belief dissensus as precursors for economic activity. The key methodological question is, how to quantify the asymmetry of trust across the parameter space?
The Political Economy of Extreme Risk and Uncertainty
We propose policy applications in emerging socio-economic systems and edge technologies such as evolving trading algorithms, distributed ledgers and quantum computing. In the context of the adoption of those technologies, the world is entering a new uncertainty regime where the probabilities of many outcomes are unknown, and concepts such as Existential Risk and the Precautionary Principle are applicable. However, a sizeable methodological obstacle lingers on: the world is a complex place where there are too many variables with multiple, changing associations and interactions — therefore, it is necessary to simplify the search space. In this respect, we use automated hypothesis generation to enunciate the questions that matter according to specific policy goals and government funding priorities.
Further information is available on the project website.