1 Introduction

One of the most frequent road accident types is rear-end crashes, which account for approximately a third of all crashes around the world [1,2,3,4]. For a rear-end crash to happen, two things are generally required: an inattentive driver, and a lead vehicle in front. While many focus on the first factor by trying to mitigate driver distraction, few address the second one, i.e., the issue of drivers following the lead vehicle too close. Still, there is ample evidence that also addressing the second issue would be worthwhile. Close following or tailgating has been identified as a main or contributing factor for 18–54% of rear-end crashes [5,6,7]. Previous research has also shown that the median time headway is shorter for near crashes compared to baseline braking events [8], and that a safe headway can prevent conflicts from occurring at all [9]. Furthermore, it has been concluded that some drivers frequently keep a continuous unsafe headway, which can lead to a higher risk of rear-end crashes [2, 10]. The general logic seems clear; as time headway increases, drivers get more time to react in case the lead vehicle unexpectedly brakes, which results in a lower likelihood of a crash. The question that then arises is how to reduce the frequency and/or duration of close following for the sake of improving road safety.

From a car manufacturer’s perspective, this can be approached by developing technology that supports drivers during routine driving, ultimately increasing the extent to which safe headway is maintained. Such technology could either inform drivers in situ about risks during driving (e.g., when the following distance is deemed unsafe), or provide continuous longitudinal support by automatically maintaining a safe distance to lead vehicles given driver-initiated activation of the technology.

One concern with the former approach is that a person’s driving style (including time headway maintenance) becomes both habitual and automated through repetition [11,12,13,14,15]. Since habits are the dominant and accessible response in a specific context [16], people with strong habits are less likely to engage in elaborate information processing [17, 18] or to address contradicting information [18]. Put into the current context, previous research has shown that drivers tend to keep rather constant headways independent of situational complexity [9]. As a result, a specific driver could habitually keep a shorter than recommended following distance to lead vehicles, but assuming this driver does not perceive it as a risk or never has experienced a near-crash or crash, there is no apparent reason for this driver to behave differently. Expecting drivers to adjust their behavior by informing them about their headway being too short is, therefore, probably not a viable solution.

The current study focuses on the second option, i.e., to encourage drivers to delegate responsibility for headway maintenance to the Advanced Driver Assistance System (ADAS) Adaptive Cruise Control (ACC). ACC provides support to drivers by adjusting vehicle speed to maintain a preset time headway that by design is within a reasonable safety margin. Indeed, previous studies have shown that when ACC is engaged, time headway increases [19,20,21,22] and consequently the frequency of close following events decreases [20, 23, 24]. Although previous research has highlighted decreased visual attention to the road ahead when drivers are using ACC, the overall safety effect when taking increased time headway and earlier brake onset into account has been deemed positive [22, 25,26,27]. Thus, the utilization of ACC seems to be an effective way to remove one of the key risk factors that contribute to rear-end crashes.

Activating ACC is literally just a press of a button away. Still, several studies have found evidence of deficit knowledge and understanding [28,29,30,31,32,33], as well as low usage levels [30, 32, 34, 35], of ACC among consumers driving cars equipped with the function. Efforts to increase ACC usage may also be affected by the existence of habitual behaviors, in this case due to the potential habit of controlling headway manually. It has, however, been argued that the problem of changing behavior is not primarily about the old habits, but rather about the lack of new ones, which is why interventions need to focus on boosting new behaviors until new habits are developed [36]. The main challenge with the approach taken in this research is therefore to establish effective behavioral interventions resulting in increased use of ACC.

1.1 Nudging as a tool for behavior change

Even though there is a vast number of behavior change theories and techniques developed and used in a variety of domains (for an extensive overview see [37]), the topic of how to increase the use of vehicle functions by affecting customer behaviors remains unexplored.

One attractive general approach was formulated by Thaler and Sunstein in their book “Nudge” [38]. They showed that when choices are to be made, highlighting a desired option while still leaving the actual choice with the individual can be a more effective way of guiding said choice than enforcement of any kind. They call this approach to behavioral modification nudging. On the most abstract level, Thaler & Sunstein describe a nudge as “any aspect of the choice architecture that alters people’s behavior in a predictable way without forbidding any options or significantly changing their economic incentives.” [38] (p. 8)

Nudging is of particular interest for anyone concerned with customer satisfaction. Since it does not force itself on the customer, chances of preserving customer satisfaction are improved. In contrast to other behavior change techniques, nudges also do not require an individual to actively form intentions, plans or make commitments related to the target behavior. Consequently, nudging has been shown to be effective in numerous fields, such as health, finance, and sustainability; for extensive overviews of nudging research and nudge effectiveness, see [39,40,41].

There is limited knowledge about nudge acceptance based on individual factors [42]. Previous research suggests that nudges are most effective if they promote behaviors that align with personal preferences [43] and there is room for improvements of those behaviors [43, 44]. However, individuals that do not hold any strong preference towards the target behavior may be particularly susceptible to nudges, specifically to those that involve social proof [43]. Previous research has also identified an association between overall approval of nudging and gender as well as age [45].

There is limited research on nudging in the driver behavior domain. However, one study found improved driving performance by introducing app-based nudges that highlighted a driver’s current trip performance compared to their best and their average respectively [46]. There are also examples of infrastructure nudges that have successfully increased safe driver behaviors. In one study, observed behavioral changes as an effect of messages displayed adjacent to a road were estimated to result in a reduction of the number of accidents by 45% [47]. A similar study showed a speed reduction by drivers as an effect of introducing both static lights and lights moving toward the drivers at the sides of the road [48]. These studies indicate that there is potential in nudging drivers to change behaviors by introducing a simple modification in the relevant context for the target behavior. This may be possible even if strong habits are present, since according to habit discontinuity hypothesis [49], a temporary disruption by changing a previously stable context has the potential to make relevant information more salient and influential, which may lead to new decisions.

There are of course many ways in which a traffic safety nudge can be implemented. In addition to nudge location (e.g., in-vehicle versus in the driving environment), nudging interventions can also be divided between those that engage system 1 (automatic) thinking and those that engage system 2 (reflective) thinking [50]. Caraban et al. provides a detailed description of 23 different ways to nudge [51], of which two specifically relate to the present research: nudges that provide ambient feedback, and nudges that enable social comparison.

In addition to nudging, another increasingly popular way to enhance engagement and motivation of specific behaviors is through gamification. Since there are similarities between applied nudging and gamification, which is also apparent in one of the nudges developed for the present research (a social comparison nudge), the idea will be briefly introduced below. Gamification has been defined as “the use of game design elements in non-game contexts” [52] (p. 1). Common design elements used include points, badges, and leaderboards [53]. A recent example of gamification applied to the traffic safety domain demonstrated promising results of including gamification in user education for partial and conditional driving automation [54]. For an extensive literature review on the behavioral effects of gamification, see [55].

1.2 Research aims

Based on the need for novel approaches to reduce instances of close following in driving, this research explores the potential of using in-vehicle behavior change interventions to increase the use of ACC. Since there is a lack of previous research efforts on this specific topic, a decision was made to focus on nudging and gamification techniques due to their overall effectiveness in influencing behaviors. Two different nudging applications, designed to reflect drivers’ ACC use in a way that encourages increased usage, were developed. The nudges were installed inside the participants’ cars and their effects on ACC utilization were evaluated in a field trial. In addition to studying the overall effectiveness of the two nudges, the aim was also to understand whether some drivers are more prone to being nudged toward the target behavior regardless of which nudge is being used, and whether switching to another nudging concept would affect all drivers in the same way. Finally, to gain further understanding of the effects of the nudges in relation to driver demographics, the age and gender of the participants were considered during the analysis.

2 Method

This section describes the participants involved in this research (2.1), the materials and procedure used (2.2), the two nudging applications (2.3), the test design (2.4), and how the data collection and analysis (2.5) was carried out.

2.1 Participants

The participants were randomly recruited Volvo Cars employees from Gothenburg, Sweden, all driving Volvo XC60 MY 2020 company cars equipped with data collection units due to being part of an internal test fleet. Recruitment of participants was done in September 2019. Individuals working as test drivers, being involved in research and development of ACC and other ADAS, or not being the main driver of the company car were not eligible to participate in the study. A gender- and aged-balanced test population was aimed for. The final sample included 49 participants (26 females and 23 males), aged between 39 and 62 years (M = 50.4, SD = 6.1). Driving experience ranged from 20 to 44 years (M = 32.0, SD = 6.3). One participant left the company mid-trial and did not participate in the treatment II condition (switching to the second nudging app), reducing the number of participants in that stage to 48.

2.2 Materials and procedure

The study was conducted in accordance with the tenets of the Declaration of Helsinki. Written consent was obtained from all participants before being provided access to their vehicle data and personal information. The collection, storage and processing of the data was achieved strictly according to the European General Data Protection Regulations (GDPR). The data was processed confidentially, and all participant identities were anonymized.

The test participants were provided written participant information, stating that the purpose of the test was to evaluate a new platform for providing feedback on driving behaviors. They were informed that their company car would be fitted with a phone holder and an iPhone 6/6s, on which they would receive visual feedback related to their driving and usage of vehicle functions. The participants were asked to keep the phone turned on and visible when they used the car and to report if any problems occurred. The participants were not incentivized or requested to behave in any certain way except for the request above. ACC was not explicitly mentioned in the information they received to minimize the risk of affecting the drivers’ behavior in any way except through the design of the apps.

The cars were already equipped with data collection units at the time of participant recruitment. At the end of the baseline period, each car had a charging holder and an iPhone installed next to the infotainment head unit, see Fig. 1. The decision to use standalone displays (iPhones) instead of the native infotainment system was made based on technical feasibility. However, the chosen location of the displays aimed to give the impression that the nudges were part of the infotainment system. The phones were configured to exclusively be able to display the ambient design nudge app (treatment I) upon treatment start. At a later stage, the app was remotely updated to switch to the competitive leaderboard nudge (treatment II) simultaneously for all participants. At treatment end, which varied between participants, the iPhones and holders were de-installed from the participants’ vehicles.

Fig. 1
figure 1

Positioning of iPhone to the left of the infotainment head unit in participants’ cars

Full size image

2.3 Nudging measures

The specific nudging measures used in this study were selected based on the outcomes of a workshop involving key stakeholders within Volvo Car Corporation R&D, including human factors and safety specialists. During this workshop, the aim of achieving increased ACC utilization and principles of nudging were introduced. This was followed by a conceptual idea generation activity. The result of the workshop was a set of different concepts, that at a later stage were assessed based on the feasibility of implementation in the current test setup. This resulted in two different ACC nudging apps being developed, one based on an ambient design concept (treatment I) and one on a competitive leaderboard concept (treatment II).

The backend of the apps subscribed to the necessary vehicle signals exposed in real-time by the data collection units. The application graphics were manipulated by backend logic based on engine status, vehicle speed and ACC status. Since ACC is also part of the ADAS Pilot Assist (PA), that provides lane centering in addition to the longitudinal support of ACC, the apps provided identical feedback irrespective of which of the two functions were being used.

2.3.1 Ambient design nudge – treatment I

The ambient design nudge aimed to nudge drivers into using ACC by progressively transforming a visual pattern from chaos to order, but only if the drivers used ACC. To reach the goal state and successfully complete the transformation from chaos to order on any given day, drivers had to use ACC for a minimum of 10 min. This duration was set to be easily achievable by most drivers. For further details on the design process, see [56]. The structure of the ambient design nudge is as follows.

The start view of the app (when the engine is still turned off) shows a yellow ACC symbol, the headline “Adaptive Cruise Control”, and the text “Not started” (Fig. 2a).

When the engine is turned on, the color of the symbol changes to grey and the text is changed to “Not available” (Fig. 2b). This view is shown while vehicle speed is below 15 kph.

When the speed exceeds 15 kph and ACC becomes available, ten grey dots start to move with random speeds on the screen and the text changes to “Available” (Fig. 2c).

When the driver activates ACC, the ACC symbol turns yellow, the dots lower their speeds and turn white, while the text changes to “Active” (Fig. 2d). For each minute of driving with ACC engaged, one dot is centered, turns yellow and slowly circulate around the ACC icon (Fig. 2e and 2g.

If the driver (temporarily) deactivates ACC, the yellow dots will stay in the center while the others will behave as they did before ACC was activated and the text states “Paused” (Fig. 2f).

When the driver has used ACC for 10 min throughout a day, all the dots will center and slowly circulate together in harmony around the ACC icon and the text “Goal reached” is shown (Fig. 2h).

After reaching the goal state, the only visual difference is the color of the ACC icon (yellow when ACC is active and grey when not). The next day the app was reset, and the process repeated.

Fig. 2
figure 2

The ambient design nudge: (a) Start view; (b) ACC not available view; (c) ACC available view; (d) ACC active view; (e) ACC active view (1 min progress); (f) ACC paused view (1 min progress); (g) ACC active view (8 min progress); (h) ACC goal reached view (ACC active)

Full size image

2.3.2 Competitive leaderboard nudge – treatment II

The competitive leaderboard nudge was designed to test another nudging approach – social comparison and competition. This app presented the drivers with a leaderboard, ranking all the participants by their weekly duration of driving with ACC activated. In contrast to the first nudge where the ACC utilization goal was fixed, the goal in this app was variable since the drivers compared their own usage of ACC to that of others. The structure of the competitive leaderboard nudge is as follows.

During driving the visuals show a Volvo XC60 either at standstill (when ACC is not activated – Fig. 3a) or driving (when ACC is active – Fig. 3b) and the number of minutes using ACC during that day.

Whenever the vehicle is at standstill or the engine is turned off, the app displays the leaderboard (Fig. 3c-e). The leaderboard includes the individual driver’s rank, their position trend since the leaderboard was last shown (positive, negative or no change), their weekly ACC minutes, and their daily ACC minutes. In addition to this, the leaderboard shows adjacently ranked drivers as well as the leader. In the bottom of the screen, different motivational text messages are shown based on the driver’s rank. All participants were assigned a fake name (due to GDPR reasons) that was shown in the app.

Every Sunday night the leaderboard reset, and the drivers received an email with their final weekly rank and ACC minutes, as well as the name of the weekly winner and his/hers ACC minutes. The next week, the process repeated.

Fig. 3
figure 3

The competitive leaderboard nudge: (a) Driving without ACC; (b) Standstill (leaderboard) without ACC usage; (c) Driving with ACC active; (d) Standstill (leaderboard) with ACC usage and a negative position trend; (e) Standstill (leaderboard) with ACC usage and a positive position trend

Full size image

2.4 Test design

The test used a within-group design, where baseline and treatment (driving with both nudging apps) data were collected from the same vehicles. All drivers were first exposed to the ambient design nudge (treatment I) followed by the competitive leaderboard nudge (treatment II). From a methodological point of view, it would have been preferable to counterbalance the order of the treatment data collection by introducing the apps in a mixed order to the participants. However, due to several practical constraints, it was not feasible to run a mixed order trial. Hence, a fixed sequence of app exposure (treatment I followed by treatment II) was used for all participants. Given that this implies potential carryover effects of the first nudge to the second, we refrain from doing any statistical significance testing of the effects of the second nudging app in isolation. Instead, descriptive statistics are used to provide insights into the effects of nudge substitution, given known outcomes of a first nudging concept. We do however argue that the two apps can be viewed as one common treatment condition if the app switch is treated as part of the study design, thus justifying statistical testing of the combination of treatment I and II compared to baseline.

2.5 Data collection and analysis

The data collection units were set up to record vehicle signal data including engine status, vehicle speed, and ACC status. Data were collected from every trip, initiated every time the engine was turned on and lasting until it was turned off.

Baseline data collection lasted between October and November 2019 and treatment data collection between December 2019 and July 2020. The baseline data consisted of 16 216 trips (4 333 h of driving), while the treatment data consisted of 40 095 trips, 29 579 for the ambient design concept (7 387 h of driving) and 10 508 for the competitive leaderboard concept (2 525 h of driving).

Each individual driver’s use of ACC was calculated as a percentage, by summing up the duration of time driving with ACC active divided by the total trip time (i.e., the duration between engine on and off, which includes instances of the vehicle being at standstill).

$$ACC\ use =\frac{sum\left(ACC\ active\ time\right)}{sum\left(total\ trip\ time\right)}\times 100$$

As an example, if a driver used ACC for 30 min while the total trip time was 120 min, the ACC use of that driver would be 25%. Calculating usage this way enabled analysis of ACC utilization in granularities ranging from specific trips to longer time periods (e.g., the complete baseline period).

To calculate baseline ACC use and the effects of the nudges in the fleet, the percentage of ACC use over total trip time was first calculated for each driver during baseline and treatment as described above, and then averaged on a fleet level. This was done for both nudging concepts in isolation and combined.

$$fleet\ average\ ACC\ use =\frac{sum\left(ACC\ use\right)}{n\ of\ participants}$$

If three drivers used ACC for 5, 10, and 15% respectively, the resulting average ACC use would be 10%. This would be the result even if e.g., the driver that used ACC during 15% of her total trip time drove twice as much as the other drivers.

Note that all percentages of ACC use presented in the results section of this paper is mainly intended to be used for comparison between baseline and treatment behaviors, they do not in themselves indicate e.g., the proportion of time drivers use ACC compared to when it is possible or feasible to. By calculating usage as done in this paper, it is impossible to reach 100%, and the fleet average ACC use is not affected by the amount of driving data a specific individual generated. All ACC use calculations did also include potential usage of the ADAS PA.

To statistically test the treatment effects on fleet level ACC use, two-tailed paired t-test were conducted. Statistical testing of baseline ACC use and treatment effects between the genders used two-tailed two-sample t-tests. Correlations between age and ACC use in baseline and treatment were assessed by means of Pearson’s R correlation tests. Significance was accepted at the p < .05 level for all statistical tests.

3 Results

This section covers the results of the field trial in chronological order, beginning with relevant baseline results (3.1) and moving on to the effects of the ambient design nudge (3.2), of the competitive leaderboard nudge (3.3), and of both nudges combined (3.4). Finally, the results are summarized in Sect. 3.5.

3.1 Baseline results

This section presents results on baseline driving and ACC use in the fleet.

The average duration of the baseline period was 70.2 days (SD = 16.9) and included an average of 338.9 trips (SD = 108.5) per vehicle.

The average ACC use in the fleet during baseline was 14.2% (SD = 14.0). Between-driver variability was large, with usage ranging from 0 to 59.5% (Fig. 4). As revealed by Fig. 4, there were two participants that already used ACC to a large extent before the nudges were introduced. In contrast, one driver did not use ACC at all during the period, while six participants had usage levels below 1%.

Fig. 4
figure 4

Baseline ACC use per driver, from high (left) to low (right) percentage. Each driver is represented by a bar while the dotted line visualizes the average value

Full size image

No significant demography-based differences in baseline ACC use were identified. The average baseline ACC use was 12.7% (SD = 13.5) among females (n = 26) and 16.0% (SD = 14.7) among males (n = 23). A two-tailed two-sample t-test showed no significant gender-based differences in baseline ACC use (p = .42). Conducting a Pearson’s R correlation test showed no correlation between age and baseline ACC use (p = .98).

3.2 Ambient design nudge (treatment I) results

This section presents the results of introducing the first ambient design nudge (treatment I) on ACC use and statistical comparisons to baseline usage.

The average duration of the treatment I period was 146.8 days (SD = 18.1) and included an average of 616.1 trips (SD = 202.1) per vehicle.

During treatment I, the average ACC use in the fleet reached 20.8% (SD = 13.5) compared to 14.2% in baseline (+ 6.6%), see Fig. 5. Thus, on average the participants increased their relative use of ACC by 46% by being exposed to the ambient design nudge. A two-tailed paired t-test shows that the increase in ACC use in treatment I (M = 20.82, SD = 13.52) compared to baseline (M = 14.24, SD = 14.02) was significant (t(48) = 5.25, p < .05). The effect size was small, with a Cohen’s d of 0.48.

Fig. 5
figure 5

Average ACC use in the fleet during baseline (left, light grey) and treatment I (right, dark grey)

Full size image

On an individual level, the changes between baseline and treatment I ACC usage varied between drivers. Figure 6 presents the difference between treatment I and baseline ACC use for each participant, ordered from positive (left) to negative (right) change in function usage. The maximum increase in ACC usage (+ 33.4%) was found for a driver using ACC during 3.7% of driving in baseline and 37.2% during treatment I. This corresponds to a relative increase of 905%.

Fig. 6
figure 6

Differences (treatment I minus baseline) in ACC use per driver, from high increases (left) to decreases (right). Each driver is represented by a bar while the dotted line visualizes the average difference

Full size image

While a majority (61%) of the drivers increased their use of ACC in treatment I, almost a third (29%) of the drivers had lower ACC use following introduction of the nudge. However, the average decrease was 2.1%, which was lower than the average increase (+ 10%) found. The maximum decrease (-5.3%) was found for a driver using ACC during 14.7% of driving in baseline and 9.4% during treatment I. This corresponds to a relative decrease of 36%.

To understand the impact of baseline behaviors on the observed differences, Fig. 7 shows each individual driver’s use of ACC during baseline (shaded light grey) and treatment I (grey), ordered from high to low baseline usage. By comparing specific utilization of ACC in baseline and treatment I on an individual level (and not only the changes), it is revealed that drivers with higher baseline use did not increase their usage to the same extent as drivers with lower baseline use.

Fig. 7
figure 7

Baseline (light grey shaded area) and treatment I (dark grey bars) ACC use per driver, ordered from high (left) to low (right) baseline use. Each driver is represented by a bar and corresponding line

Full size image

This pattern is also visualized in Fig. 8 where the change in ACC usage between treatment I and baseline, compared to baseline usage per individual driver is shown.

Fig. 8
figure 8

Difference (treatment I minus baseline) in ACC use (y-axis) versus baseline ACC use (x-axis) per driver. Each dot represents an individual driver

Full size image

Pearson’s R correlation test of baseline ACC use and observed difference between treatment I and baseline usage resulted in a weak negative correlation r(47) = − 0.37, p < .05. This provides further signs of a stronger nudging effect on drivers with lower baseline ACC use.

To compare the difference between treatment I and baseline ACC use and gender, a two-tailed two-sample t-test was conducted. There was no significant difference in treatment I vs. baseline ACC use between females (M = 7.94, SD = 9.44) and males (M = 5.02, SD = 7.84); (t(46) = 1.18, p = .24).

In addition, conducting a Pearson’s R correlation test showed no correlation between age and difference between treatment I and baseline ACC use (p = .46).

3.3 Competitive leaderboard nudge (treatment II) results

This section presents the results on ACC use during treatment II (switching to the competitive leaderboard nudge) and descriptive statistics compared to the preceding baseline and treatment I results. Due to potential order effects of all participants previously being exposed to the ambient design nudge, no statistical testing of ACC usage levels during treatment II compared to baseline or treatment I was conducted.

The duration of the treatment II period was 52 days, consisting of an average of 225.5 (SD = 84.0) trips per vehicle. As mentioned earlier, the number of participants during treatment II was reduced to 48, why the results comparing treatment II to baseline and treatment I have different average values in this section.

During treatment II, average ACC use in the fleet reached 30.7% (SD = 19.2) compared to 14.5% in baseline (+ 16.2%) and 20.8% in treatment I (+ 9.9%), see Fig. 9. Thus, on average the participants increased their relative use of ACC during treatment II by 112% compared to baseline, and by 48% compared to treatment I.

Fig. 9
figure 9

Average ACC use in the fleet during baseline (left, light grey), treatment I (middle, medium grey) and treatment II (right, dark grey)

Full size image

To compare individual baseline (shaded light grey), treatment I (grey), and treatment II (dark grey) ACC use, data from each driver ordered from high to low baseline usage is shown in Fig. 10. Most drivers (73%) had higher ACC usage in treatment II compared to treatment I. Again, drivers with higher baseline use did generally not increase their usage during treatment II to the same extent as drivers with lower baseline use. The five drivers with the highest baseline ACC use did however have higher usage levels in treatment II compared to baseline.

Fig. 10
figure 10

Baseline (light grey shaded area), treatment I (grey bars) and treatment II (dark grey bars) ACC use per driver, ordered from high (left) to low (right) baseline use. Each driver is represented by two bars and corresponding line

Full size image

Individuals showing high increases in ACC utilization during treatment I did so during treatment II as well. Drivers with the greatest increases in treatment II had previously increased their use of ACC during treatment I, but to a lower extent. Furthermore, one group of drivers (n = 5) had a decrease in ACC usage compared to baseline during both treatment I and II, while another group (n = 9) showed a decrease in treatment I but an increase during treatment II. The correlation between calculated differences in ACC use during treatment I (x-axis) and treatment II (y-axis) compared to baseline is shown in Fig. 11.

Fig. 11
figure 11

Difference between treatment II and baseline ACC use (y-axis), versus difference between treatment I and baseline ACC use (x-axis) per driver. Each dot represents an individual driver

Full size image

3.4 Combined treatment results

This section presents results on the combination of treatment I and II compared to baseline results.

Summing the data from the 48 participants that were included in both treatment I and II results in an average ACC use of 23.4% during the complete treatment period (treatment I + II), compared to 14.5% (+ 8.9%) in baseline. Thus, on average the participants increased their relative use of ACC by 61% by being exposed to both nudging concepts. A two-tailed paired t-test shows that the increase in ACC use in treatment I + II (M = 23.43, SD = 14.18) compared to baseline (M = 14.51, SD = 14.04) was significant (t(47) = 6.11, p < .05). The effect size was medium, with a Cohen’s d of 0.63.

There was a large spread in the relative changes in ACC usage between drivers also when combining treatment I and II, see Fig. 12. The maximum increase in ACC usage (+ 36.9%) was found for a driver using ACC during 2.4% of driving in baseline and 39.3% during treatment I and II. This corresponds to a relative increase of 1538%.

Fig. 12
figure 12

Differences (treatment I + II minus baseline) in ACC use per driver, from high increases (left) to decreases (right). Each driver is represented by a bar while the dotted line visualizes the average difference

Full size image

A decrease in use of ACC after being exposed to both nudges was found for eight of the drivers (16.7% of the fleet). The average decrease in usage was 2.8%, which is lower than the average increase (+ 11.3%) found. The maximum decrease (-5.8%) was found for a driver using ACC during 14.7% of driving in baseline and 8.9% during treatment I and II. This corresponds to a relative decrease of 39%.

To compare the difference between the combined treatment and baseline ACC use and gender, a two-tailed two-sample t-test was conducted. There was no significant differences in treatment I + II vs. baseline ACC use between females (M = 10.30, SD = 11.32) and males (M = 7.40, SD = 8.58); (t(46) = 1.00, p = .32).

Conducting a Pearson’s R correlation test showed no correlation between age and difference between treatment I + II and baseline ACC use (p = .26).

3.5 Results summary

This section summarizes the acquired statistical testing results.

Statistically significant effects on fleet level ACC usage were found for an ambient design nudge alone and when combined with a competitive leaderboard nudge. A weak negative correlation between baseline ACC use and effects of an ambient design nudge was identified, i.e., lower baseline use correlated with higher increases in usage when being exposed to the nudge. The statistical tests performed found no differences between baseline ACC use or effects of nudging related to neither gender nor age.

Table 1 summarizes the statistical tests comparing ACC use in baseline, in treatment I (ambient design nudge), and in treatment I + II (ambient design nudge and competitive leaderboard nudge combined).

Table 1 Overview of statistical testing performed based on baseline data (left), treatment I vs baseline data (middle), and treatment I+II vs baseline data (right). Results on fleet level ACC use (top), gender-based ACC use (middle), and correlation tests between age and ACC use (bottom) per condition are presented. Significant results are in bold.
Full size table

4 Discussion

The general aim of this research was to explore whether in-vehicle nudging can increase car drivers’ use of ACC. As the results showed, this indeed seems possible. Both tested nudging applications significantly increased average ACC usage in the fleet, with an effect size comparable to nudging interventions in other domains [39,40,41]. Given the importance of safe headway maintenance, which is aided by ACC utilization, nudging techniques therefore seems to be a viable addition to other efforts aiming to reduce the number of rear-end crashes.

This study did not apply any inclusion or exclusion criteria based on previous ACC usage, which enabled the possibility to gain an understanding of the normal (unaffected) levels of ACC usage prior to treatment, as well as how different types of ACC users were affected by the nudges. The extent to which ACC was being used in the fleet varied greatly, and a relationship between baseline ACC use and nudging effects could be identified. Greater increases in ACC usage through nudge exposure were observed among drivers not already using ACC to a large extent. The switch to the second nudging app (the competitive leaderboard) did however contribute to increased usage also among drivers with high baseline ACC usage and who were seemingly unaffected by the first nudge. While most of the participants increased their use of ACC when being exposed to the nudges, a sixth of the drivers did not. No demography-based differences in natural utilization of ACC or effects of the nudges were identified.

That individuals would respond differently to the nudges was an expected finding in line with previous research [43, 44]. Unsurprisingly, drivers already using ACC to a very large extent were not strongly affected by the nudges. As pointed out by [44], for a nudge to have the intended effect there must be a scope for improvement in behavior, and since ACC usage was analyzed based on duration, the room for improvement was simply smaller for individuals who already used ACC extensively during baseline driving. Also, given that the daily goal in the ambient design concept was set to only 10 min it was not likely to trigger a goal pursuit within these individuals as they would be able to reach the goal simply by behaving as usual. Switching to the competitive leaderboard nudge may however have introduced an extrinsic motivational element [57] (comparing with others and winning the weekly competition) that gave at least some of them a strong enough reason to increase their ACC usage even further.

It was more surprising that the nudging apps were able to turn some drivers that rarely or never used ACC during baseline driving into ACC users without any coaching elements being implemented in the applications. This indicates that the mechanisms of the nudges in themselves were effective enough to successfully change behaviors in the intended direction. Still, it can be argued that nudges like these might struggle to succeed if they are targeting individuals that either: are unaware of the function being available in their vehicle, do not know how to use it, lack basic motivation to use it [44], or have a clear preference for the alternative [43] (i.e., manual driving). This implies that the participants in this study either already knew how to use ACC or easily could find out, and that other user categories might need additional education and coaching measures to start to use the function or increase their usage.

A sixth of the drivers (almost a third before switching to the second nudge) did not change their behavior as intended by being exposed to the nudges. In fact, there were even some decreases in ACC usage. While these decreases were not large enough to suspect a backfire effect (which is probable for approximately 15% of nudging interventions according to [41]), this finding highlights the importance of monitoring the behavioral effects of implemented nudges to be able to control for unintended effects. It also implies that there may not be such a thing as a one-nudge-fits-all, which has also been indicated in previous research [51]. The key takeaway is that when nudges do not have the intended or even undesired effects on target behaviors, designers need to be ready to counter this by either modifying the nudge or by cancelling the intervention all together.

The results indicate that the switch of nudge design increased the use of ACC further for most drivers. Similarly, some drivers that were unaffected by the first nudge showed increases in usage when switching to the second app. Furthermore, the switch resulted in some drivers ending up with an increase in ACC utilization over the complete treatment period compared to baseline, despite decreasing their usage while being exposed to the first ambient design concept. One implication of this is that the design and principles of a nudge seems to matter, since changing those aspects can result in different outcomes for the same individual on the same target behavior. Future research should try to corroborate whether this is the case or otherwise study if some individuals simply are more prone to be nudged independently of the exact characteristics of the nudge design. Furthermore, it could be fruitful to take individuals’ perceptions and attitudes toward the intended behavioral change into account and study how that correlates with measured behavioral outcomes.

There are some limitations of the current study that are important to mention. First, the fixed order of the nudges made it impossible to draw any strong conclusions on the isolated effects of the competitive leaderboard nudge. Second, baseline and treatment data were collected over several months due to the extended duration of the field trial. Therefore, seasonal effects on ACC utilization are possible in the different conditions, e.g., the potential impact of changes in weather and frequencies of longer trips. These limitations are recommended to be addressed in future studies.

It is also important to remind about all participants being Volvo Car employees based in Gothenburg, Sweden. Whether this has any implications on attitudes towards functions in Volvo cars or the acceptance of nudges remains unknown. Further, the sample did not represent the whole age span of drivers and did not include any novice drivers. Studying nudging interventions aiming to increase the use of vehicle functions by involving participants representing different customer groups, and with a larger variety in age, driving experience and geographical locations, could reveal further insight into which variables are able to predict behavioral outcomes.

Moreover, driving environments were not considered neither in the nudge designs nor during analysis. Since the extent of driving different individuals do on different road types probably varies, it may affect both overall ACC utilization and perceived usefulness of increasing the use on an individual level. It is also important to be aware of the issue of misuse or overreliance of driver support functions [58, 59] when aiming to influence drivers to increase their use of such functions. Such non-targeted outcomes should however first and foremost be mitigated by solutions that either warn the driver about inappropriate function usage, or even limit functional availability based on e.g., driving context and assessments of driver state. Nevertheless, taking driving context into account is recommended in future research endeavors, both related to the design of nudges (e.g., by only nudge when deemed to be suitable and beneficial) and during analysis of nudge effects.

The nudging interventions designed and tested in the current study also had some potential drawbacks. Introducing a nudge design that involves motion may have a negative effect on driver visual attention. However, the ambient design concept, which included motion and visual feedback during driving, only introduced some distraction in the beginning of the drive according to participants in a user test of a prototype version of the app [56]. While the visual information presented to the drivers by the leaderboard concept was limited whenever the vehicle was not stationary, it had other drawbacks. Ranking drivers based on the actual duration (minutes) of ACC usage is not ideal when drivers use their cars to different extents. Some drivers that do not drive extensively but still use ACC during a large proportion of their driving might lose motivation e.g., by feeling that it is unfair and impossible to win, or possibly perceive that they are being encouraged to increase the amount of driving they do, which was not the intention of the nudge.

The main implication for further research and development is to try to gain an understanding of the user before any measures are taken. Nudges like the ambient design concept should ideally target non-users and low users of ACC to achieve the largest benefit of it, rather than trying to push a user that already uses the function to a high degree even further (at least unless making the daily goal adaptive). Furthermore, since it likely is impossible to predict the behavioral outcomes of all individuals, the intended effect should be monitored when the nudge is in place, combined with a readiness to adapt the nudge based on observed behavioral outcomes. Finally, the long-term effects of nudges should be assessed to determine whether it has succeeded to create a sustained behavioral effect.

5 Conclusions

This research has shown that utilizing in-vehicle applications based on nudging principles is a promising approach for changing driver behaviors for improved road safety. The introduction of nudges significantly increased the use of Adaptive Cruise Control (ACC) on a fleet level, most evident for drivers with lower baseline function usage. By switching to a second nudge, an even higher usage of ACC was observed. The results from this research suggest that nudging provides an underexplored opportunity to increase the use, and therefore realize the latent safety potential, of driver-activated ADAS such as ACC. Important learnings include that understanding the users and monitoring the behavioral outcomes on an individual level is key to be able to adapt a nudge when needed.