A Behavioral Approach to Modeling Route Choice Decisions with Real Time Information

A Behavioral Approach to Modeling Route Choice Decisions with Real Time Information


E Ben-Elia, Y Shiftan, Technion - Israel Institute of Technology, IL


A behavioral approach to modelling the combined effects of information and experience on drivers'route choice decisions with real-time information provision, based on an experimental psychology study and advanced discrete choice analysis.


One of the great endeavors taken by transportation engineers to alleviate congestion is the development of advanced transportation information systems (ATIS). Their goal is to provide drivers with accurate real-time traffic information, reduce uncertainty in travel time, increase travel time savings and enable drivers to choose routes efficiently. By providing accurate traffic information drivers could economize on travel time promoting network wide savings and benefits.
However, recent attempts to predict the expected effect of ATIS systems are based on the assumption that the behavior of individual travelers can be approximated with the assertion that they maximize expected utility. That is, individual travelers are assumed to minimize expected trip costs (or trip disutility) choosing the most efficient routes between any given origin and destination, with a system resulting in a stable network equilibrium.
This modeling approach, although elegant and parsimonious, has been criticized for lacking a sound psychological basis ignoring drivers? cognitive limitations. Psychological research suggests that cognitive limitations are important. Experimental decision research highlights robust deviations from the predictions of maximization. The results can be summarized with the assertion that human decision makers exhibit bounded rationality.
It seems that route choice models can be improved by adding more realistic behavioral assumptions. However, recent attempts to use behavioral findings in the context of route choice reveal that this task is not trivial. Whereas almost all studies agree that people deviate from the normative maximization predictions, different generalizations (of previous behavioral research) imply deviations in different directions. This severely restricts the possibility of integrating behavioral insights into a travel behavior model with practical forecasting capabilities a crucial phase in ATIS development and assessment.
Robust derivation of the implications of the behavioral research requires better understanding of behavior under the conditions that characterize drivers? route decisions. To achieve this goal the current research examines experimentally choice behavior in an environment in which the decision makers receive partial real time information about the ranges of travel times (i.e. variability), and can rely also on their personal experience through immediate feedback. The results of the study will be used to refine current transportation models with the addition of more robust assumptions concerning the behavior of the human agents.
We consider a hypothetical example of a driver who faces information about average travel times and travel time ranges (deviation around the mean), regarding the choice between two routes from work to home in three traffic scenarios. This problem seems simple, but it is complex enough to demonstrate the difficulty in applying current behavioral research.
The observation that most people refrain from touching a second time a hot oven door (?hot stove? effect) suggests that driver is likely to exhibit risk aversion i.e. routes with a large range will be less attractive. Conversely, the basic research summarized by Prospect Theory (Kahaneman & Tversky, 1979) suggests risk seeking in the loss domain. Finally, recent study of decisions from experience (Erev & Barron, 2005) suggests that an increase in (travel time) variability moves behavior towards random choice.
An experiment based on the hypothetical example investigates the combined effects of information and experience on route choice decisions in a simulated environment whereby the participants can rely on a description of travel time variability and at the same time can rely also on personal experience through feedback. The experiment consisted of a simple two route network, one route on average faster than the other with three traffic scenarios representing different travel time ranges. Respondents were divided to two groups: with real-time information and without. Both groups received feedback information of their actual travel time. During the experiment, participants chose repeatedly between the routes and across scenarios.
The results show that effect of information is positive and is more evident when participants lack long-term experience on the distributions of travel times. Furthermore, information seems to increase initial risk seeking behavior, reduce initial exploration and contribute to between subject risk-attitudes differences. These findings have implications for cost-effective ATIS design especially in the conditions characterized by non recurrent congestion which are difficult to predict in advance.
Based on this data we estimate an advanced discrete choice model using mixed logit (panel data) to capture the combined effects of information and experience in route choice decisions. We use the insights obtained in the experiment to refine current transportation models of ATIS.


Association for European Transport