This study elaborates on a model system developed in earlier papers to predict the perceived value and use of travel information. The value of travel information is conceptualized as the extent to which the information allows the individual to make better activity-travel scheduling decisions at the beginning of the day and during execution of the schedule. By taking the broader schedule context into account, the model is sensitive to the impact of information and decisions on the full activity-travel pattern. Furthermore, the model includes Bayesian mechanisms to make sure that beliefs about travel times, and other uncertain events, and the credibility of the information source are updated each time information is received and the real travel time is experienced. This paper describes the results of numerical simulations conducted to illustrate the system and to derive theoretical implications from the model. The simulations show that the schedule context, learning, and expected information gain in combination determine the perceived value of information and that none of these factors can be ignored in the derivation of estimates of these values. A theoretical analysis further shows how decision trees can be pruned to reduce a potential problem of combinatorics.
Abstract