The weight of experience : an investigation of probability weighting under decisions from experience

In decisions from experience tasks objective information regarding payoffs and probabilities must be inferred from samples of possible outcomes. A series of recent experiments has revealed that people show deviating choice behaviour in such tasks, indicating underweighting of small probabilities instead of overweighting of small probabilities as in decisions from description. In a range of experiments, the research presented in this thesis provides a new direction by showing that such reversals from overweighting to underweighting in decisions from experience are very robust and can be replicated even if all the existing explanations - sampling error, recency weighting and judgement error - are experimentally controlled for. Furthermore, reversals were replicated within common decision making biases like the common ratio effect. An important, but unexpected, new finding has been the observation of a reversed reflection effect under decisions from experience. This suggests that the difference between choice behaviour may not be restricted to underlying transformations of probabilities, as suggested in the literature. Drawing from an extensive range of model tests and parameter estimations, it is also demonstrated that the differences are reflected in the best fitting parameter values for prospect theory under decisions from experience. However, it is also shown that simple reinforcement models, which provide a more intuitive rationale for experiential choice behaviour, can account for the data just as well, without any assumptions regarding the weighting of probabilities.