Bayesian Updating in Experiment: Good News

This paper explores small decision problems experimentally. Conducted is the current experiment in which agent's payoff distribution is limited to either high (favourable) distribution ("Good News") or low (unfavourable) distribution ("Bad News"). We conduct calibration of numerical optimal solution to search behaviour by Bayesian updating and agents' tendency in the laboratory experiment in small feedback-based problems. One assumption on an rational agent is that an agent is to behave to maximise his expected payoff. Results of the current experiment, however, show subjects' seemingly puzzled tendency inconsistent with the assumption above. The law of small numbers is observed in the experiment. The law of small numbers tells us that an agent will gather too little data and will overgeneralise from small samples to distributions. Agent's overgeneralisation of distribution may lead him to behave not to maximise expected payoff.

SSRN Electronic Journal, 2003

We examine decision-making under risk and uncertainty in a laboratory experiment. The heart of our design examines how one’s propensity to use Bayes’ rule is affected by whether this rule is aligned with reinforcement or clashes with it. In some cases, we create environments where Bayesian updating after a successful outcome should lead a decision-maker to make a change, while

Behaviormetrika, 2007

This is an experimental economics research on human behaviour in "small decision making". I set up ambiguity treatments in which there are two states of nature: a favourable state and an unfavourable state, but only one of them obtains on any given trial. The decision makers' basic task is a binary choice between a risky option with higher expected value and a riskless option with lower expected value. The one-person game is iterated hundreds of times. Experimental results are reported with several findings, such as underweighting of rare events and deviations from expected value maximisation. Finally, I investigate the imperfect Bayesian decision makers observed in the experiments by exploring to what extent they can update prior probabilities and reflect it in making decisions.

2003

We examine decision-making under risk and uncertainty in a laboratory experiment. The heart of our design is directed at studying how one's propensity to use Bayes' rule is affected by whether this rule is aligned with reinforcement or clashes with it. We create a decision problem where there are two states of the world, two 'urns' containing different proportions of desirable 'balls' in each state, and two draws with replacement from these urns. In some cases, Bayesian updating after a successful outcome should lead to a second draw from the other urn, but should lead to a second draw from the same urn after an unsuccessful outcome. We observe striking patterns of behavior: When reinforcement and Bayesian updating are aligned, nearly all people respond as expected. However, around 50% of all decisions are inconsistent with Bayesian updating when reinforcement clashes with it. We also find a pronounced preference for simplicity, as people tend to make costly initial choices that eliminate uncertainty/complexity in a subsequent decision, thereby minimizing the chances for regret or cognitive dissonance. Follow-up treatments suggest that, to a large extent, errors are driven by the affect experienced from the initial draw, with positive affect being a particularly strong force.

Games and Economic Behavior, 2001

We report on an experiment designed to evaluate the empirical implications of Jordan's model of Bayesian learning in games of incomplete information. A finite example is constructed in which the model generates unique predictions of subjects' choices in nearly all periods. When the "true" game defined by players' private information was one with a unique equilibrium in pure strategies, the experimental subjects' play converged to the equilibrium, as Jordan's theory predicts, even when the subjects had not attained complete information about one another. But when there were two pure strategy equilibria, the theory's predictions were not consistent with observed behavior.

2006

This thesis aims at pursuing an extensive investigation into decision making in small decision-making problems with the tool of experimental economics. A typical small decision-making problem is characterised by three critical features: first, it involves repeated tasks; the decision maker faces the same choice problem many times in similar situations. Second, each single choice is of little consequence in terms of net payoff; the alternatives tend to have similar expected value that may be fairly small. Finally, in choosing among the possible alternatives, the decision maker will have to rely on the immediate and unbiased feedback obtained in similar situations in the past. The importance of shedding light on the economics of small decisions is twofold. The first is that nowadays many common economic activities can be captured by small decision-making problems. Second, although each small decision

2006

Behavioral economics aims to provide more realistic psychological foundations for economic models. Experimental methods can contribute to this effort by providing the ability to identify causal processes and motivations that can be confounded in field settings. The essays in this dissertation examine three critical issues in behavioral economics using lab and field experiments. The first two essays examine two core elements of economic rationality; expected utility theory and Bayesian updating. The essays consider, respectively, ambiguity, and information cascades, in environments in which limitations of the theories can be studied. The third essay examines a contracting game in which other-regarding preferences are explicitly considered. Decision making under ambiguity has been of interest to economists since the 1920's (Knight (1921), Keynes (1921)). It has received renewed attention due to the work

The GENEVA Papers on Risk and Insurance-Theory, 2002

Following a brief review of the main experimental work into the economics of risk and uncertainty, both static and dynamic, this paper reports the results of an experiment testing one of the key assumptions of the theory of dynamic economic behaviour -that people have a plan and implement it. Using a unique design which enables the plan (if one exists) to be revealed by the first move, the experiment was implemented via the Internet on a subset of the University of Tilburg's ongoing family expenditure survey panel. The advantages of using such a set of subjects for the experiment are twofold: the demographic characteristics of the set are known and therefore demographic inferences can be made; the representativeness of the set is known and therefore inferences about populations can be made. The results suggest that at least 36% of the subjects had behaviour inconsistent with the hypothesis under test: that people formulate plans and then implement them. Interestingly demographic variables are unable to explain the consistency or inconsistency of individuals. One conclusion is that subjects simply make errors. An alternative conclusion, consistent with previous experimental research, is that people are unable to predict their own future decisions. The implications for dynamic theory (particularly relating to savings and pensions decisions) are important.

How humans achieve long-term goals in an uncertain environment, via repeated trials and noisy observations, is an important problem in cognitive science. We investigate this behavior in the context of a multi-armed bandit task. We compare human behavior to a variety of models that vary in their representational and computational complexity. Our result shows that subjects' choices, on a trial-totrial basis, are best captured by a "forgetful" Bayesian iterative learning model [21] in combination with a partially myopic decision policy known as Knowledge Gradient [7]. This model accounts for subjects' trial-by-trial choice better than a number of other previously proposed models, including optimal Bayesian learning and risk minimization, ε-greedy and win-stay-lose-shift. It has the added benefit of being closest in performance to the optimal Bayesian model than all the other heuristic models that have the same computational complexity (all are significantly less complex than the optimal model). These results constitute an advancement in the theoretical understanding of how humans negotiate the tension between exploration and exploitation in a noisy, imperfectly known environment.

Experimental Economics - EXP ECON, 2003

We describe Bayes factor, an explicit measure of the strength of the evidence, the extent to which the data increase or decrease the odds a given hypothesis or model is true. Issues and techniques for deriving a Bayes factor are outlined. We illustrate the technique with data from an ultimatum game experiment that looked for an experimenter observation effect. We show that the evidence increases the odds of an effect, but not by enough to convince someone with a skeptical prior.