Neural processing of context and information: Implications for behavioural modelling

Many disciplines focus on the exploration of human choices, hence considerable progress has been made with respect to understanding how people make decisions within different fields. Nonetheless, cross-disciplinary efforts are still limited. This is especially apparent for the fields of choice modelling and neuroscience, mainly due to their contrasting focus. In particular, choice modelling seeks to understand why specific decisions are made, through capturing the differential influence of explanatory variables on that decision process. On the other hand, neuroscience is focused on a direct measurement of the biological activity of the brain under specific circumstances to infer the neurological foundations of the observed behaviour. For decades, these two disciplines have been developing in parallel, separated by the lack of practical and theoretical grounds to build on, and also separated by major differences in the type of data used. While choice modelling has looked at complex multi-alternative, multi-attribute settings, using either revealed or stated preference data, neuroscience has focussed on simple tasks that are repeated a very large number of times in a laboratory setting. Technological advancements such as virtual reality have recently allowed for more dynamic and complex situations to be reproduced in experimental (lab-based) settings, while further developments on non-intrusive sensors have made it possible to collect physiological and neural data in a way that is more comfortable for the participant and allows for more flexible experimental design. The emergence of these novel opportunities gave the basis for the work gathered in this thesis which adopted an integrative approach, combining choice modelling with virtual reality data collection and neurological measurement, with an applied focus on cycling behaviour. The conducted studies demonstrate the feasibility of such complex data collection efforts and evaluate the impact of the experimental design in virtual reality on the elicited behaviour and neural data. With the employed choice models, we demonstrate the differences in behaviour and neural reaction as a result of the adoption of immersive and non-immersive visual stimuli and changes in the riskiness of simulated road scenarios. Furthermore, a statistical analysis of the cycling behaviour and neural data, when two different input devices are employed, yields intuitive findings providing practical implications for researchers who plan to use virtual reality in their future research. In the final study of this thesis a hybrid choice model framework is proposed to simultaneously model cycling behaviour and brainwaves data. It shows that neural inputs can successfully be used as indicators for a latent construct in a hybrid model structure to capture risk, serving as an alternative to traditional measures e.g. attitudinal scales. This work not only demonstrates how to operationalise such modelling efforts but the addition of a neural perspective allows us to improve the understanding of cycling behaviour achieved with the existing models. Taken together, the findings presented in this thesis allow us to gain a better understanding of determinants which influence cyclists’ choices in risky road situations, which we would not be able to explore in the real world due to safety concerns. Finally, they provide evidence of the potential for collaborative research between choice modelling and neuroscience to encourage more studies in this new direction that would stimulate the development of new modelling structures incorporating biometric data, enable more extensive exploration of different stages of the choice process, and consequently lead to more informed decision-making.