**4. Results Analysis**

By comparing the perceived travel time of the pre-run questionnaire to the actual travel time collected during the on-road tests, it was demonstrated that the accuracy of participants' perception of travel time ranged from 5% to 55% for all five trips, with an average accuracy of only 38%. Consequently, it would be safe to conclude that the participants had limited knowledge of the route performance prior to the start of the experiment. Based on the results of participants' perception in the post-run questionnaire, the average accuracy increased from 38% to 62% with an increase of 24%. Consequently, it would be interesting to see whether participants behave more rationally with higher perception accuracy.

Figure 1 presents the proportions of logical- and inertial- choices as a function of time, identified as trial number. As expected, the logical choice rates are on average around 10% higher in the "with-information" group than "without-information" group, especially for the first two trials in which the enhancement is up to 15%. This demonstrates that the positive effect of information becomes more evident when travelers have limited knowledge of route performance. Although there are some oscillations at some of the trials, in general, the logical rates between the two groups are getting closer from the beginning to the end. The inertial choice rates are basically lower with the provision of information, implying that it is more likely for travelers to risk switching to the faster route when they are informed. However, regardless of being informed or not being informed, the inertial behavior is not reduced in day-to-day variation, which is different from the results in the simulation study [16]. This may be attributed to the habit or other decision considerations.

**Figure 1.** Logical- and inertial- choice rates over trials.

In reality, the behavioral effect of information varies from person to person. One may probably have more confidence in his/her experiences than the acquired information; or travel time is not his/her top consideration. Accordingly, the insights gained from previous analyses are needed. Nine of the participants in this study attended Tawfik and Rakha's experiment. The choice results of these participants were specifically compared between the two experiments in order to see how the effect of information differentiated individually and how well they learned from the information. Figure 2 compares the behavioral types (introduced in Table 1, which was proposed by [21]) for each of the nine participants between with- and without-information. Only 10 trials were compared because the participants were informed for 10 trials only. The degree of the fluctuation of each line gives an explicit generalization of participants' behavioral aggressiveness. The more fluctuated in the lines, the more aggressively the participants behave. In general, the information significantly changes behavioral types either from risk-seeking to risk-aversion or vice versa. Some of the participants exhibited a high preference for one route when information was not provided and switched frequently when they were informed; whereas some switched more without information and maintained a single route when informed. Overall, the effect of information significantly differs at an individual level.

Figure 3 summarizes the behavioral tendency of participants. According to Figure 3a, participants 1, 2, 3, 5, 8 basically moved their choices towards rationality with the assistance of information, whereas participants 4, 6, 7, 9 behaved more irrationally when they were informed. In Figure 3b, participants 6, 7, 9 instead have higher inertial rates with the provision of information, implying that they behaved even more risk aversive whey they were provided with information.

Based on the results of the post-run questionnaire, participants 6, 7 and 9 mentioned that travel time information had little impact on their route choices. Specifically, participant 6 preferred rural roads due to his preference on route scenery, although travel time was important to him as well. Participant 7 held the point that, instead of travel time, the number of intersections was the overriding factor she considered for route choice decisions. Participant 9 preferred to stick to her current route without any route-switching, which is the first type of the typical behavior shown in Table 1. Noticeably, participant 4 had both logical and inertial rates decreased with the provision of travel time information. That was because travel time was not the only consideration to this participant. Based on the results of

the questionnaire, "avoid traffic lights" was the other equally important factor to him, which highly impacted his choice behavior. Occasionally, participant 4 switched to the slower route instead in order to avoid traffic lights even though he was informed the alternative route was better in terms of travel time, which increased the proportion of compromising behavior (the other type of illogical choice other than inertial choice) and decreased the logical choice rate. In general, travel time may have little effectiveness in enabling drivers to behave logically when drivers do not take travel time as their foremost factor in planning their routes. Additionally, participants 4, 6, 7, 9 are all senior persons from the age group of 55–75 year old. This implies that elder drivers are preferable to make choices based on their preferences or habits rather than received information, which confirms the results of [17].

 **Figure 2.** Participants choice patterns without vs. with route information.

**Figure 3.** Logical- and inertial- choice rates over participants.

In addition to individual traits, trip characteristics may also affect the positive role of information. To study such effects, the choice rates were aggregated by trips. As illustrated in Figure 4, information enhances behavioral rationality only for the first three trips. For trip 4, logical rates decrease while inertial rates increase when information is provided. On trip 5, the choice rates do not change significantly between with- and without- information. According to the route characteristics addressed in Table 2, route 7 and route 8 (on trip 4) are almost identical in travel time, whereas many participants pointed out that they were reluctant to take route 7 even though it occasionally took less travel time since they did not want to risk being caught on campus by pedestrian flows. The provided information was considered to be less reliable for this trip. Interestingly, travel time is very close as well between the two routes on trip 1; however, the effect of information appears to be very positive. That is because there is no distinct advantage for one route over the other on this trip. Although route 1 is on a highway system with a 20 km/h higher speed limit than route 2, there are five more signalized intersections on it. The provided travel time information for this trip was considered reliable by participants. For trip 5, route 10 distinctively outperforms route 9 in terms of travel time, directness, less traffic and fewer intersections. The authors of [5] clearly indicated that drivers were able to precisely perceive the route performance and to make correct decisions on this trip without any assistance of information. Overall, information provides little benefit if one route visibly outperforms the other.

**Figure 4.** Logical- and inertial- choice rates over trips.

Figures 5 and 6 provide a broad view of the effect of different information types on route choice behavior. In Figure 5, the comparative analysis was performed between strict information (average travel time) and range information (variability). According to Figure 5a, strict information results in higher logical rates with lower inertial rates for the first trial, demonstrating that strict information is more effective than range information when drivers lack experience. For the following trials, however, there is no significant distinctiveness between the two scenarios. This may be attributed to the fact that the effect of information type tends to be identical after drivers gain experience. As illustrated in Figure 5b, strict information results in higher logical rates and lower inertial rates on average. Nonetheless, to some of the participants, range information performs better, implying that the responses to different information types, to a large extent, are dependent on individual traits, although strict information overall performs better in this study.

(**a**) Choice rates over trials

(**b**) Choice rates over participants

**Figure 5.** Choice rates with strict information vs. with range information.

**Figure 6.** Choice rates with risky-fast scenario vs. safer-fast scenario.

Figure 6 presents the effect of different range information scenarios. As illustrated in Figure 6a, "Risky-fast" scenario refers to the faster route (lower average travel time) with higher variability while "safer-fast" represents the faster route with lower variability. Interestingly, the risky-fast scenario appears to have higher logical rates and lower inertial rates in the first two trials, whereas the positive effect decreases in the following three trials. This implies that, when drivers have limited knowledge of route performance, the faster route with high variability is more attractive and subject to make drivers take risk in the gain domain. Once drivers gather experience, however, they are reluctant to risk seeking in the gain domain under higher uncertainty; instead, the safer-fast route becomes preferable. This confirms the result of [10,11,14,15]. Figure 6b demonstrates that there is no consensus between participants on which scenario is more effective. Some of the participants have higher logical rates and lower inertial rates for the risky-fast scenario while some exhibit the opposite pattern.
