How do people make difficult decisions in situations involving substantial risk and uncertainty? In this study, we presented a difficult medical decision to three expert physicians in a combined “thinking aloud” and “cross examination” experiment. Verbatim transcripts were analyzed using script analysis to observe the process of constructing and making the decision, and using referring phrase analysis to determine the representation of knowledge of likelihoods. These analyses are compared with a formal decision analysis of the same problem to highlight similarities (...) and differences. The process of making the decision resembles an incremental, sequential‐refinement planning algorithm, where a complex decision is broken into a sequence of choices to be made with a simplified description of the alternatives. This strategy results in certain kinds of relevant information being underweighted in the final decision. Knowledge of likelihood appears to be represented as symbolic descriptions capturing categorical and ordinal relations with “landrhark” likelihoods, only some of which are described numerically. Numerical probabilities, capable of being combined and compared arithmetically, were not observed. These observations suggest an explanation for the heuristics and biases in human decision making under uncertainty in terms of the processes that manipulate symbolic descriptions of likelihoods and construct plans of action for situations involving risk and uncertainty. (shrink)

Part I Introduction The specific problem adressed in this thesis is: what is the rational use of theory and experiment in the process of scientific discovery, in theory and in the practice of drug research for Parkinson’s disease? The thesis aims to answer the following specific questions: what is: 1) the structure of a theory?; 2) the process of scientific reasoning?; 3) the route between theory and experiment? In the first part I further discuss issues about rationality in science as (...) introduction to part II, and I present an overview of my case-study of neuropharmacology, for which I interviewed researchers from the Groningen Pharmacy Department, as an introduction to part III. Part II Discovery In this part I discuss three theoretical models of scientific discovery according to studies in the fields of Logic, Cognition, and Computation. In those fields the structure of a theory is respectively explicated as: a set of sentences; a set of associated memory chunks; and as a computer program that can generate the observed data. Rationality in discovery is characterized by: finding axioms that imply observation sentences; heuristic search for a hypothesis, as part of problem solving, by applying memory chunks and production rules that represent skill; and finding the shortest program that generates the data, respectively. I further argue that reasoning in discovery includes logical fallacies, which are neccesary to introduce new hypotheses. I also argue that, while human subjects often make errors in hypothesis evaluation tasks from a logical perspective, these evaluations are rational given a probabilistic interpretation. Part III Neuropharmacology In this last part I discusses my case-study and a model of discovery in a practice of drug research for Parkinson’s disease. I discuss the dopamine theory of Parkinson’s disease and model its structure as a qualitative differential equation. Then I discuss the use and reasons for particular experiments to both test a drug and explore the function of the brain. I describe different kinds of problems in drug research leading to a discovery. Based on that description I distinguish three kinds of reasoning tasks in discovery, inference to: the best explanation, the best prediction and the best intervention. I further demonstrate how a part of reasoning in neuropharmacology can be computationally modeled as qualitative reasoning, and aided by a computer supported discovery system. (shrink)

Humans have long been characterized as poor probabilistic reasoners when presented with explicit numerical information. Bayesian word problems provide a well-known example of this, where even highly educated and cognitively skilled individuals fail to adhere to mathematical norms. It is widely agreed that natural frequencies can facilitate Bayesian reasoning relative to normalized formats (e.g. probabilities, percentages), both by clarifying logical set-subset relations and by simplifying numerical calculations. Nevertheless, between-study performance on “transparent” Bayesian problems varies widely, and generally remains rather unimpressive. (...) We suggest there has been an over-focus on this representational facilitator (i.e. transparent problem structures) at the expense of the specific logical and numerical processing requirements and the corresponding individual abilities and skills necessary for providing Bayesian-like output given specific verbal and numerical input. We further suggest that understanding this task-individual pair could benefit from considerations from the literature on mathematical cognition, which emphasizes text comprehension and problem solving, along with contributions of online executive working memory, metacognitive regulation, and relevant stored knowledge and skills. We conclude by offering avenues for future research aimed at identifying the stages in problem solving at which correct versus incorrect reasoners depart, and how individual difference might influence this time point. (shrink)

This article presents a computational model of the learning of diagnostic knowledge, based on observations of human operators engaged in real-world troubleshooting tasks. We present a model of problem solving and learning in which the reasoner introspects about its own performance on the problem-solving task, identifies what it needs to learn to improve its performance, formulates learning goals to acquire the required knowledge, and pursues its learning goals using multiple learning strategies. The model is implemented in a computer system which (...) provides a case study based on observations of troubleshooting operators and protocol analysis of the data gathered in the test area of an operational electronics manufacturing plant. The model not only addresses issues in human learning, but, in addition, is computationally justified as a uniform, extensible framework for multistrategy learning. (shrink)