Quantitative approaches to multi-alternative choice

The six papers that make up this thesis describe the empirical and theoretical benefits of quantitative approaches to examining decisions between multiple alternatives. In this thesis I explore two aspects of multi-alternative choice: a classic cognitive question and a modern applied question. In the cognitive section I examine Hick's Law (Hick, 1952), a function that relates expected response time to the number of choice options. Hick's Law - for response times - has been consistently supported across many paradigms and decades of research, however divergent patterns have been observed in response accuracy rates. Sometimes accuracy rates are independent of choice set size, while other times they decrease with the number of choice alternatives. The first four chapters of this thesis develop and test a novel experimental factor that reconciles the discrepant patterns in accuracy data - the choice context. I show that the context of a decision leads to differences in empirical response accuracy rates, which can be explained with a quantitative mechanism that assumes decision makers optimise time-on-task, conditional on a goal accuracy rate. The mechanism is based on decision makers' perception of time, and I rule out a potential confound based on decision difficulty, and a specific process-level implementation of the quantitative mechanism. It is concluded that context effects in multi-alternative choice are the result of discrete changes to response caution parameters in each experimental condition rather than a continuous, gradual change to response thresholds across conditions. In the applied section of this thesis I examine consumer judgments in the context of decision tasks used to elicit attitudes and preferences. These tasks yield rich estimates of the `preference strength' or `utility' that describe value-based judgments, conventionally estimated using multinomial logit (MNL) models. MNL models provide compact descriptions of data but are silent on the cognitive processes involved in these complex, multi-attribute choices. In the final sections of this thesis I develop and test a process-based quantitative model from the sequential sampling class of response time models as an account of best-worst choices. I show that the quantitative model provides utility estimates that mirror those of MNL models, and makes unique predictions for response times in consumer judgments that are subsequently used to further test and refine the model. I conclude with a quantitative model of consumer judgments that accounts for best-worst choices and accepting and rejecting in perceptual and value-based consumer-style judgments.