  anglais seulement |
|
|
MAIN MENU
|
LABORATORY MISSION STATEMENTThe goal of our lab is to study human decision making under uncertainty, in particular, when faced with financial risks. In contrast to most research in finance, our way of studying is experimental. We probably have an equal amount of theoretical input and we certainly don't shy away from adding to theory, but we have moved from studying (historical) data from the field to running controlled experiments.
Financial risks are a relatively recent phenomenon, evolutionary speaking. As such, the human brain may not have perfectly adapted protocol to deal with it. In a first stage, we would like to know how the brain perceives risks and how it generates decisions in very controlled, say "ecologically relevant" situations. We already know quite a bit about reward learning, and how simple but robust temporal difference modeling is hardwired in the human brain (and the brains of monkeys, rats and mice, for that matter). Our lab has brought another aspect to the forefront in this regard: risk perception and risk learning. These are necessary for any organism that is risk-sensitive (whether risk-averse of risk-loving) or even just attempts to learn to predict stochastic payoffs as accurately as possible.
Financial risks are known to be different from most naturally occurring risks in at least two respects: (i) they are really hard to predict (although earthquakes are too, but once a large shock occurs, aftershocks are likely...); (ii) their distribution is "lepto-kurtotic," meaning that often not much happens, while "outliers" happen too often compared to, say, the gaussan distribution (in other words, the "typical" deviation from expectation is in fact rare).
In our lab, we generate financial risks ourselves: we set up markets where subjects trade purposely designed securities while we control the information flow. We can then replay these markets to other subjects, and ask from them all kinds of tasks, in order to better understand how the human brain perceives the risks and acts on them. We can study, e.g., the role of emotion in financial decision making, why certain people are better at predicting prices than others (what is trading intuition?), or what part of their choices reflect cognitive (or computational) biases as opposed to preferences.
Hopefully one day we will understand enough about human assessment of financial risks that we can go back and improve the design of our markets, so that humans make better decisions, or that anomalous market phenomena such as bubbles, crashes and market meltdowns can be avoided. That is, our research should lead to better tools that improve not only individual decision making, but also system behavior, i.e., market performance.
There are many facets of human decision making under uncertainty. We already mentioned a few. One particularly intriguing question is: what happens when the brain encounters something that was not expected? That is, a real surprise occurs, after which one has no clue what probabilities to assign to possible outcomes. One of the reasons why this is an interesting question is that classical decision theory has no prescription for what to do in that situation (lest one is a Bayesian who puts a prior on everything, which in itself is known to be a doomed strategy).
A last topic we are at present studying is how perception of rewards and risks are converted into decisions. As anyone knows who has observed humans more closely, behavior can be quite erratic even in situations where, despite uncertainty, optimal strategies are nonrandom and simple to compute. Why is that? Is this because utility is random, as economists would claim? CURRENT STUDIES
BOARDGAME CHOCOLATE BOX fMRI EXPERIMENT
COLLABORATING LABORATORIESProf. John P. O'Doherty, Caltech Prof. Wolfram Schultz, Cambridge Masamichi Sakagami, Tamagawa University, Japan
|
|
|
Site map • © 2010 EPFL , ODY 1.17, Station 5, 1015 Lausanne, tel. +41 21 693 0121 ldmu@epfl.ch |