The Lazy Intuitive Statistician : Influence of Data Representation and Retrieval Processes on Intuitive Statistical Judgment

Sammanfattning: Intuitive statistical judgments are an integral part of people’s everyday life and a long line of research has investigated the extent to which man lives up to the norms of statistical theory when performing such judgments. A recent account of intuitive statistical judgments, summarized in the metaphor of the naïve intuitive statistician (K. Fiedler & P. Juslin, 2006), has suggested that people base judgments on small samples, which they have an ability to veridically record but an inability to evaluate the representativeness of.The present thesis builds on research concerning the naïve intuitive statistician and investigates how representation and memory retrieval of numerical information influences intuitive statistical judgments. Two studies were conducted. Study I introduced two possible accounts of how numerical information is represented and retrieved. The first possibility suggests that information is stored as exemplars and that estimates of statistical properties are calculated on small samples drawn from memory at the time of a query. The second possibility suggests that numerical information is stored as abstract summary statistics calculated at the time of exposure. The distinction was summarized in the metaphor of a lazy vs. an eager intuitive statistician. Study II extended the findings of Study I by investigating how point estimates of unknown quantities are formed from knowledge of statistical properties of a numerical variable. More specifically, a model of naïve point estimation based on the naïve sampling model (P., Juslin, A., Winman, & P., Hansson, 2007) was introduced to predict participants’ distribution of point estimates.In general, the results from both studies support the idea that people spontaneously act as lazy intuitive statisticians that record numerical information in a raw format during exposure and postpone evaluation of statistical properties until they are requested to do so. Under certain fairly predictable and limited circumstances, however, participants were able to form abstract representations of statistical properties. The results of Study II support the predictions by the model of naïve point estimation, including a novel phenomenon where participants give point estimates which they know, when probed otherwise, have a low probability of occurring.The findings of the two studies extend previous research concerning people’s ability to be intuitive statisticians by not only measuring how accurate the knowledge of properties of numerical variable is but by also describing how such knowledge is represented. The model of naïve point estimation contributes to the existing body of research by describing how people perform one type of intuitive statistical inference, point estimation, and shows how statistical properties of the underlying distribution influences the pattern of responses. The model also suggests novel explanations to results showing that people seem to have implicit expectations that distributions are normal.