Practical Application Considerations

In a real car, it might be relevant to train the driver on the meaning of a specific scent, to reinforce a preferred behaviour (e.g. "Take a break"), just like we were trained to associate traffic signs with certain pieces of information [95]. Prior work by Kuang and Zhang [115] suggests that there is a potential of doing so by means of the conditioning, which was proven to work in a smell enhanced visual motion perception study.

Our findings are not intended to present a well established mapping for the design of a semantic messaging system, but rather to highlight the correspondence between the arousal of the scents and the alertness level of the messages. This motivates the application of scents based on how important, relevant, or salient the driving-related message is.

9.6.4 Challenges

Our research provides a necessary starting point to open up a new interaction design space for HCI. Despite promising findings, further research is needed involving an even larger sample size, extended set of scents, and more messages.

Our driving setup did not enable links between the scent-message combinations and the current situation on the road. The effect of the scent might be stronger if its delivery is synchronised with a certain traffic event or a vehicle status update. Improvement possibilities also include replacing the air tank with a compressor (more feasible in a real car [220]).

Working with the sense of smell raises ethical concerns as it involves the handling of chemicals, but also because scents have a strong association with emotions and memories. This emphasises the need to allow for customisation of olfaction-based

interfaces. The same applies to personal and cultural preferences. Further challenges may include smell unfamiliarity, persistence, and "the stimulus problem" [11].

Persistent smells could be eliminated through advancements of the "olfactory white"

in the future [211]. We also need to consider the challenge of delivering the scent without invading the olfactory space of the co-driver and the passengers [49]. Moreover, to account for are potential scents that people seated in the car (i.e. drivers and passengers) bring in with them.

Even though the effect of smell on the driving performance and experience has been studied [15, 170], there is a need to investigate these factors within the scope of conveying driving-related information by scents (also in a real driving setting).

9.7 C ^ONCLUSION

Our findings show that using olfactory stimuli as an alternative interaction modality in the car is not arbitrary and that participants are able to establish a mapping between specific driving-related messages and scents. Based on the induced alertness level of both the message and the scent, we demonstrated that it is possible to establish a new semantic layer of information delivery for the driver. These insights open up new opportunities to further explore the topic of conveying information using smell in the context of driving and beyond.

10 | Towards a Framework for

Validating the Matching Between Notifications and Scents in

Olfactory In-Car Interaction

Dmitrijs Dmitrenko, Emanuela Maggioni, Marianna Obrist

SCHI Lab, Creative Technology Research Group, School of Engineering and Informatics University of Sussex, Chichester I, Falmer, Brighton, BN1 9QJ, UK

In CHI EA ’19: Extended Abstracts of the 2019 CHI Conference on Human Factors in Computing Systems, ACM, 2019.

DOI:https://doi.org/10.1145/3290607.3313001 Abstract

Olfactory notifications have been proven to have a positive impact on drivers. This has motivated the use of scents to convey driving-relevant information. Research has proposed the use of such scents as lemon, peppermint, lavender and rose for in-car notifications. However, there is no framework to identify which scent is the most suitable for every application scenario. In this paper, we propose an approach for validating a matching between scents and driving-relevant notifications. We suggest a study in which the olfactory modality is compared with a puff of clean air, visual, auditory, and tactile stimuli while performing the same driving task. For the data analysis, we suggest recording the lane deviation, speed, time required to recover from the error, as well as the perceived liking and comfort ratings. Our approach aims to help automotive UI designers make better decisions about choosing the most suitable scent, as well as possible alternative modalities.

10.1 I NTRODUCTION

Visual notifications dominate in modern vehicles, however, any distraction of the driver’s visual attention on the road can have fatal consequences [163]. Sound can reduce the

visual load and help the driver perceive the urgency of the warnings [54], but it can also be annoying [14] or even distracting [54]. This has stimulated the exploration of other modalities [153]. Tactile interfaces have been widely studied and have indicated, e.g. a positive effect on users’ attention in safety critical environments [189], faster braking reaction times [127] in simulated driving, while also being less annoying [121]. Olfactory stimulation is, still largely unexplored in automotive contexts, even though it could help drivers process information [184].

Olfactory stimulation is the most challenging communication channel to apply in the car, due to scent lingering and interpersonal differences [49]. It has been proven to have a positive impact on the alertness and mood of the driver [15, 170], drivers’ braking performance [130], and on keeping drowsy drivers awake [220, 66, 157]. Nevertheless, there are still only very few investigations of using smell as a communication channel [48, 46]. However, considering the increased visual load in modern infotainment and driving assistance systems, coupled with advances in scent-delivery [45, 129, 4, 161], we see a great opportunity to rethink the integration of scent into modern vehicles. A valid approach is necessary to decide what scent matches each specific driving-relevant notification. Our paper proposes the first steps on the way of establishing such a validation framework.

The significance and the originality of our approach can be summarised as follows:

• We propose the first approach to help automotive UI designers make better de-cisions about choosing the best scent for a specific driving-relevant notification.

This is especially important considering the recent olfactory interface tendencies in the automotive industry, involving such manufacturers as Mercedes-Benz [35], BMW [20], and Bentley [18].

• Our approach goes beyond the most relevant previous work in this area. We offer a structured method of choosing the scent for an in-car interaction scenario, by not only taking into consideration the knowledge on the effects of scents from psychology and neuroscience, but also by comparing their efficiency opposed to other modalities (i.e. vision, audition, and touch).

To make sure our method is valid, we built up on the work of Politis et al. [165], who compared different types of visual, auditory, and tactile stimulation as stand-alone and combined notifications to convey driving-relevant information. We extend this work by including the olfactory modality.

0 2 4 6 8

Olfaction Off Olfaction On

(a) Driving task (b) Sensory notification

(c) Analysis of the driving behaviour

***

Speeding time (s)

Figure 10.1: Framework for validating the matching between driving-relevant notifications (e.g.

"Slow down") and scents (e.g. Lavender). We propose a framework with the following three steps: (a) Selecting a driving task, (b) Displaying a notification (auditory, visual, tactile, olfactory, or combined) relevant to this driving task (e.g. one sensory modality/combination of modalities per condition), and (c) Analysing the driving behaviour based on what is required by the selected task (e.g. mean speeding time in seconds, in case of a "Slow down" notification, error bars ±SD,

∗ ∗ ∗p< .001).