Elise Payzan

UPDATE: From September 2007 onwards, I am now pursuing my PhD in Neurofinance under the supervision of Peter Bossaerts, at the Ecole Polytechnique Fédérale de Lausanne (EPFL, Switzerland). Peter is transferring from Caltech to lead a new lab and take a joint position at EPFL and at the Swiss Finance Institute. I will also be affiliated to the SFI. No need to say that while it is always sad to leave a place such as the LSE, I am excited and eager to work under the supervision of Peter and join such a leading institution in the neurosciences area (for more info on this, see the Brain Mind Institute).

I will update this website very soon. In the meantime, for more information please look at the website of the newly created Laboratory for Decision Making under Uncertainty.

Hello and welcome on my "home" on the net ! My name is Elise Payzan and I am currently a PhD Student at the Finance Department of the London School of Economics.

My research interests have two intertwined parts: first at some individual level I am interested in individual financial decision-making under uncertainty. Second at an "aggregate" level, I am interested in the price patterns emerging in the financial markets: the idea is to emulate real participants in an artificial market in which agents display adaptive learning patterns, that are empirically grounded.

As for individual financial decision-making under uncertainty, I ask the following questions: how to optimally switch between different learning modes, depending on the uncertainty of a trading environment? Do real trading behaviors suggest that such a meta-arbitrage between different modes does exist? If this is the case, how does the brain implement/approximate it?

Besides, I am also interested in the way traders trade-off exploration and exploitation, and in the role of emotions in a "particular style" of trading under uncertainty, based on implicit learning and tacit knowledge.

I will use methods from computational neuroeconomics and experimental economics to examine these questions.

As for the "aggregate" level, I focus on the nature of the price patterns resulting from the interaction of "agents" of a trading system. I will use methods from agent-based computational finance to investigate this point. The "agents" are modeled in such a way that some important characteristics of learning, metalearning, and trading psychology, are taken into account.

I strongly believe that there is an important bridge between computational neuroeconomics and computational finance: the use of realistic algorithms to rigorously describe the micro behaviors of the agents allows one to be more confident about the "validity" of the emergent patterns of the simulated systems. This "bottom-up" approach is a foundational response to the skeptics.

Individual financial decision-making

Computational neuroeconomics

Learning modules and uncertainty

Using a computational approach, I try to shed some light on important metalearning aspects of the decisions under uncertainty, such as the understanding of the arbitrage between different learning modules, depending on the nature and level of uncertainty. The competing learning modes I identify as relevant in the context of trading are "classical" cognitive ones (statistical research and pattern recognition), "sophisticated" ToM (Theory of Mind, to interpret the markets), and "intuition" ("implicit learning").

The primary objective is to derive realistic algorithms to describe each mode. For example, I conjecture that "intuition" might be described by an algorithm of the type Q-learning, while ToM might be modeled through bayesian mechanisms (with intentions and motivations of others as the set of beliefs), and the "mechanical" rules through delta rules - also called Least Mean Square (LMS) learning rules.
Also, the goal is to look for the optimal arbitrage between them in a simple trading environment.
The next step then is to see the gap between this optimal arbitrage and the real behaviors displayed by real actors. In particular I am interested in the behavior of professional traders vis-à-vis normal persons.

The exploration/exploitation trade-off in trading under uncertainty

Reinforcement learning allows to actively learn about an unknown markovian environment and come to choose appropriate actions. I want to study in the lab various aspects of the trade-off between exploration and exploitation, when the environment under study is an artificial stock market.

First the idea is to calibrate i.e choose the model of probabilistic choice that fit best real trading data. I distinguish three possible competing models: random choice, non-directed exploration (standard "softmax rule"), directed exploration (with exploration bonuses). I conjecture that the second or the third model will be kept as the best fit.

The question then is whether professional traders are more prone than lay investors to adapt (somewhat optimally) the exploration parameter to changes in uncertainty in the environment.

Trading psychology

Emotions, learning and motivation in trading

The impact of emotions on trading performance is commonly seen as a negative:

Indeed an abundant literature in behavioral finance, whose implicit conceptual frame is the Kahneman-Epstein "dual-process" paradigm, insists on disruptive influences: anticipated and anticipatory emotions distort the cognitive evaluation, hence well-documented markets anomalies - such as overreaction to market signals, impulsiveness, excessive cautiousness.

The "ex post" emotions - generated by the realized trading performance - have similar negative effects on future trading: the "disposition effect" and the disturbing effect of these emotions on attention are two of them.

Furthermore the distortion of perception by emotions might refer to "bottom-up" attentional mechanisms too: in their article "The psychophysiology of real-time financial risk processing", published in "The journal of cognitive neuroscience" in 2002, Andrew Lo and Dmitry Repin study the importance of emotion in the decision-making process of 10 professional securities traders by measuring their physiological characteristics (e.g., skin conductance, blood volume pulse, etc.) during live trading sessions. They simultaneously capture real-time prices from which market events can be detected. Significant differences in emotional responses to those market events are systematically related to the traders' level of experience, and the proposed message is that more experienced traders will be more prone to perceive important signals vis-à-vis less relevant ones: they perform better because they have better "attentional skills", due to their higher emotional "coldness".

Still the picture is more complicated: in real trading life uncertainty echoes ambiguity more than risk. The "Iowa Gambling Task" paradigm has demonstrated that when information is missing, emotions play a beneficial role of "somatic markers".

I believe that the examination of a particular trading method, belonging to the family of "implicit learning", offers the opportunity of an integrative view, regarding the role of emotions in financial decision-making. Brett Steenbarger, in a beautiful article called "Learning to trade", which can be found here, explains the origin of the concept of "implicit learning" and its meaning, and suggests that much of the trading expertise is the product of implicit learning, i.e of processes that are not conscious, and not intentional either.

I endeavor to investigate the possibility that emotions play a very specific functional role in the implicit learning of trading. This learning mode, which can be referred to as "intuitive", is characterized by the unconscious learning of rules which are cognitively imperceptible. It requires both maximal concentration and at the same time some metacognitive regulation: top-down control mechanisms must be "shut down" (i.e if one strives to consciously recognize some rule, she will fail to learn). In other words, it is a crucial requirement to be fully receptive to bottom-up signals. Now the interesting thing is that the latter are inherently emotional in financial markets (they are "somatic markers"). Hence there is an optimal management of one's emotions in this specific case, with both a cognitive reappraisal (inhibition) of the "bad" disruptive emotions (they will hinder concentration and thereby disrupt attention, as suggested by A.Lo and D.Repin's previously mentioned article), and an opening to others (the relevant market signals). This implies a sophisticated filtering out of emotions. I would like to examine the neural underpinnings of this filtering.

Aggregate price patterns

Computational finance : Simulation of a trading system

At an "aggregate" level, I am interested in the price patterns emerging in a stock market represented as an agent-based system comprised of interacting traders. Because of the absence of both ubiquitous rationality and a common knowledge principle, the others' behaviors are unpredictable, hence the use of adaptive inductive rules by the agents. To study whether some coordination between agents towards the rational price might still emerge, it is necessary to simulate the trading system. In that purpose, the use of evolutionary algorithms is a tool to mimic the adaptation of the decision rules but also the adaptation of crucial meta-decisions such as the choice of the relevant decision rule among the pool of learning modes, and the degree of exploration. Individual rationality and market pressure dictate which strategies will survive, according to their "fitness" (performance).