Practical implications

4.3.1 Validation of a new distraction potential testing method

Broström et al. (2013, 2016) and Ljung Aust et al. (2015), for instance, have argued that a robust distraction testing method is needed to assess more reliably the distraction potential of secondary in-car tasks. One major practical implication of this dissertation is the validation of the new distraction potential testing method.

The used testing method by Kujala and Mäkelä (2015), to the best of our knowledge, is the first distraction potential testing method that assesses driver distraction against a baseline of attentive driving, or more precisely, against spare visual capacity in attentive driving. This means that, when conducting distraction potential testing with the method, it is possible to determine if the driver should be glancing at the forward road scene instead of the tested in-vehicle device or application on a particular route point. Based on this, it is further possible to assess the distraction potential of the tested task by examining whether the glance durations of the driver are timed right in relation to the

46

variable visual demands of the driving scenario. In other words, the method utilizes a baseline for attentive driving in which the glancing behavior of the distraction potential testing can be reflected against. The driving scenario of this method is also more realistic (containing intersections and curves, for example) than the driving scenario suggested in the NHTSA (2013) guidelines, and drivers can self-pace their off-road glance durations in relation to driving demands with speed adjustment, as in real traffic.

Article III (Kujala, Grahn, et al., 2016) presents the first research that applied this method, and provided a baseline for an acceptable in-car task to which more complicated in-car tasks can be compared. An acceptable in-car task is a task that seems to not cause excessive visual distraction. In later studies (Grahn & Kujala, 2018; Kujala, Grahn, et al., 2016; Kujala & Grahn, 2017), we were able to demonstrate that the distraction potential testing method produced similar results across similar tasks. The similarity of the results is an indication of consistency, which gives the results and the overall method more reliability (Lazar et al., 2010).

The used testing method, to the best of our knowledge, is also the first method that takes into account drivers’ individual glancing behaviors or, in other words, takes into account drivers’ individual differences. In articles III to VI, we suggest that, in order to take into consideration the drivers’ individual differences in glancing behavior, the driver sample should be validated measuring drivers’ occlusion distances to ensure that the sample contains drivers who prefer to drive only short distances without visual information to those who prefer to drive long distances without visual information. This idea was further justified in Article II (Grahn & Taipalus, 2021) by suggesting that this kind of procedure improves the robustness of the distraction potential testing. For instance, Ljung Aust et al. (2015) showed that, by manipulating the participant pool, the distraction potential test results following the NHTSA (2013) guidelines had "near stochastic outcomes." In addition, Broström et al. (2016) as well as Lee and Lee (2017) were able to affect the results of the distraction potential testing conducted following the NHTSA (2013) guidelines. In Article II, we demonstrated that, when using the distraction potential testing method by Kujala and Mäkelä (2015), the driver sample does not affect the results of the distraction testing. We suggest that the driver sample validation with occlusion distances is a significant factor in enhancing the robustness of a distraction potential testing method.

In the literature, both neglecting the visual demands of the driving scenario and individual differences in glancing behavior have been raised as major drawbacks in the current distraction potential testing (e.g., Broström et al., 2016;

Kujala et al., 2014; Tivesten & Dozza, 2015). It could be argued, based on the studies included in this dissertation, that with similar distraction potential testing method as the method used here (Kujala & Mäkelä, 2015), these recognized drawbacks can be conquered. Hence, Kujala and Mäkelä’s (2015) testing method has four benefits compared to, for instance, previously introduced NHTSA’s (2013) testing method: 1) it provides a baseline for attentive driving by

47

incorporating the visual demands of the driving scenario which is used to assess the distraction potential of the tested task, 2) the driving scenario is more realistic, containing intersections and curves, 3) drivers can self-pace their off-road glance durations in relation to the driving demands with speed adjustment, as in real traffic, and finally, 4) it takes into account that drivers’ have individual glancing behaviors by controlling that there are diverse drivers in the participant sample.

Hence, this practical contribution answers research question 2 (How can driver inattention be measured more reliably and with better validity?) by presenting the idea of measuring inattention caused by the secondary in-car tasks assessing tested tasks’ visual distraction potential against a baseline of attentive driving and taking drivers’ individual glancing behaviors into account.

4.3.2 Context-specific design diminishes visual distraction

Previous research has examined different interaction methods used while driving and conducting a secondary in-car task. Another practical implication of this dissertation is the conclusion that a context-specific user interface design (i.e., a user interface designed specifically for the automotive context) has the potential to diminish drivers’ visual distraction. The interaction methods especially seem to have a large effect on drivers’ visual distraction. Overall, we were able to produce similar results concerning interaction methods as previous literature:

touch screen keyboard is relatively the most distracting interaction method (e.g., Crandall & Chaparro, 2012; McKeever et al., 2013; Reimer, Mehler, & Donmez, 2014), voice-based interaction methods (speech-to-text function and read-aloud function) are less distracting than manual text entry (e.g., Beckers et al., 2017b;

He et al., 2014; He et al., 2015; Tsimhoni et al., 2004), simple swiping gestures are visually low distractive, and a up display is less distracting than head-down display (e.g., Smith et al., 2016). Additionally, we noticed that the head-up display did not cause gaze concentration. Gaze concentration, or a narrowing of the visual scanning behavior, decreases the ability to detect peripheral and central targets (Wang et al., 2014), which, naturally, affects driver’s situation awareness. As a novel discovery, a rather new text entry method called handwriting was found to be less or equally distracting as touch screen keyboard typing. Also, the size of the screen alone had only a minor effect and the orientation of the screen had no effect on visual distraction. These results overall provide more insightful knowledge concerning interaction methods used in user interfaces of different applications. For instance, the knowledge concerning the distraction potential of read-aloud function seems to be especially lacking.

Above all, utilizing multilevel modeling, we were able to conclude that a context-specific design with its multimodal interaction and simplistic design has a diminishing effect on drivers’ visual distraction. This means that if an application is designed in the first place to be visually less distracting, bearing in mind the context it is designed for, it indeed has the potential to be visually less distracting. Hence, we suggest that scientific knowledge regarding human–

technology interaction should be utilized when designing for a safety-critical context. All these practical contributions concerning interaction methods and

48

context-specific design answer to research question 3 (What are the effects of selected in-car task features on drivers’ visual distraction potential?).

4.3.3 Carefully designed subtask boundaries benefit drivers

Furthermore, it is possible to diminish the visual distraction potential of in-car tasks by carefully designing the subtask boundaries utilizing natural breakpoints.

Basically, this is one part of the context-specific design. Earlier studies have indicated that people have a tendency to switch tasks in natural breakpoints (e.g., Janssen et al., 2012; Lee & Lee, 2019), such as dialing a chunk of phone numbers at once. Based on our results, the possibility to break down an in-car task into smaller subtasks decreased in-car glance durations. This enables drivers to better adjust their glancing behavior in relation to the demands of the driving scenario.

This practical contribution answers research question 3 (What are the effects of selected in-car task features on drivers’ visual distraction potential?).

The practical implications presented in Section 4.3 benefit the research community and other instances (such as New Car Assessment Programs, NCAP, Imberger et al., 2020), focusing on traffic safety research by providing a suggestion of how driver inattention can be measured more reliably. These suggestions could also be utilized when developing driver distraction detection algorithms (e.g., Ahlström et al., 2021). In addition, the implications concerning the design of the secondary tasks benefit the automotive industry and designers working within the industry.