Probabilistic uncertainty refers to situations where learning can proceed by repeatedly trying actions. Beliefs converge to a common standard (“objective probabilities”) regardless of how sophisticated belief updating takes place. This is also the environment where machine learning is successful. Probabilistic uncertainty is the foundation of much of empirical work on decision-making under uncertainty. Economists study all decision-making under uncertainty as if this is the only type of uncertainty that exists.

Complexity permeates daily life. We study two types of complexity: (i) computational complexity, i.e., situations where it’s hard to figure out what to do, even if one has all the data (even the best electronic computers have a hard time dealing with this type of complexity; unsurprisingly, it has been studied extensively in computer science); (ii) inference complexity, i.e., situations in which it is hard, from observations, to infer what caused the observations. When making decisions under either type of complexity, one cannot generally be confident to be right. We (and others) have recently proposed that both types of uncertainty cause people not to try to maximise rewards/values, but to control surprise. Engineers have long used the idea as one way to ensure robust control.

Here we focus on strategic uncertainty: situations in which even formal theoretical modelling (“game theory”) makes it impossible to predict what one’s opponent/ally will do. Psychologists would argue that such situations require “theory of mind.”

Virtually all economic models of social interaction, including through markets, assumes that outcomes reflect a system that is “in equilibrium.” Equilibrium restrictions are used to interpret historical data from the field and advise policy makers such as central bankers. But absent an economic law of “entropy,” it is by far a foregone conclusion that economic systems equilibrate. Since the 1960s, controlled experiments have been run to evaluate under what conditions equilibrium really emerges, and if not, what the reasons could be. 

Before they got focused on equilibrium, economists (most prominently, Cambridge’s Alfred Marshall) actually explored markets behaviour without assuming that it somehow converges to equilibrium. This endeavour was put on the back burner however, because of lack of tools (mathematical framework and analysis; experimental methodology). We have worked at addressing this situation, engaging in a dialogue between theory and (experimental) data in order to discover basic principles that govern behaviour of markets off equilibrium.

The standard view of financial markets is that they serve an important role in re-allocating risk (across agents and over time). But they appear to be equally important as a tool to gather information (prediction markets), and to spread knowledge — whether that knowledge is pure data points or it helps resolve computational complexity.

Financial markets generate a type of risk which is rather unusual, in that there are lots of extreme events (“outliers”): either the market is very calm, or stormy; there does not seem to be a “normal” level of uncertainty. This raises the questions as to whether humans, or artificial agents (machine learning), are well prepared for it, and if not, what to do about it. 

Experimental research on markets requires software with which one can readily set up multiple simultaneous markets where human participants can exchange tailor-made goods and securities. The software has to be flexible enough so many relevant economics situations can be studied (and taught). Exchange mechanisms have to be realistic, yet amenable to control. Because of the relevance of robot traders in the field, the software should also be able to interface with algorithms that participants write to automate trade or that are made available to participants so they can deploy them to enhance interaction with markets. We do not limit attention to emulating successful field market mechanisms, but explore novel mechanisms, with the goal of resolving tough problems in the field, such as lack of liquidity. 

Summary: This course aims to give students an introduction to some basic concepts in economics, law and ethics. On completion of this course, students should be able to: Reflect on and discuss professional, economic, social, environmental, moral and ethical issues relating to computer science; Define and explain economic and legal terminology and arguments; Apply the philosophies […]

Summary: This class introduces students to general equilibrium theory, which should be understood as “the theory of how markets behave and interact and what the consequences are for the quality (efficiency) of ensuing allocations.”: Applications in finance are focused on, which means that students will gain insight into widely used concepts such as alphas and […]

Summary: By the end of the course, students should be in possession of a good grasp of the elementary tools of descriptive statistics; should understand elementary principles of probability and statistical theory; should be competent in applying basic methods of statistical inference; and should be familiar with the use of spreadsheets to undertake graphical and […]

Summary: The overall aim is to introduce students to the microstructure of modern financial markets in general, and to algorithmic trading in particular. Algorithmic trading refers to the use of robots (automatic order submission computer program) to accomplish a certain trading goal, such as automatic market making, statistical arbitrage, technical analysis, portfolio rebalancing, etc. Students […]

Summary: (Part 1 only). The course focuses on how markets deal with uncertainty. General equilibrium is emphasized, which refers, loosely speaking, to the point at which traders in multiple, simultaneous markets no longer desire further trade. Are the resulting allocations optimal? What if everyone can re-trade in the future after some information is revealed? What […]

Summary: The goal is to better understand human attitudes towards uncertainty in general, and financial risk in particular. The method to get there is to go beyond a pure behaviouralist approach (the tradition of economics), point to the difficulty of deciphering the psychology behind behaviour (the “thinking”/cognition and “feeling”/affect), to eventually land squarely in the […]