Urban transports and the act of driving in condensed urban areas can lead to eective changes in people's emotions. The time consumed to these tasks is also signicant as most people spend a considerable amount of time just moving from one place to the other. Therefore, people are exposed to the risk of others and may pose a risk to other themselves. In order to target, and promote sustainable urban driving this articles proposes the use of gamication techniques and estimates of the users' emotions to inuence not only good behavior but also sane emotional states for drivers. This work presents the integration of emotions into the gamication system for driving assessment.
An ubiquitous society enables dierent solutions to classic problem such as urban transportation. It is in fact common knowledge that urban transports has problems that have fueled research and attention of governance bodies.
One such case is the driving analysis and hazardous behavior from drivers. It is fact the case that most of the fault for trac accidents is in fact human error. This error is often avoidable is there are systems in place to detect the drivers state of mind, attention and driving patterns.
This eld has recently caught the attention of insurance companies [
Another form of intervention is to promote users who follow trac and driving recommendations such as alerts and rest pauses. In the eld of pervasive and ubiquitous computing, it is possible to analyze driving behavior and categorize it in dierent categories such as aggressive or relaxed. These are not traditional assessment and possible by the fusion of data capture either directly by the vehicle or with the help of sensing devices [7].
The action upon users is a problem that needs to be solved. Not all users respond in the same manner to suggestions, recommendations or alerts and a tailor made, multi-purpose approach is needed for these structures to work.
In this article we present an architecture to link gamication to the emotional state of mind of users in order to inuence driving behavior. Dierent strategies are devised to the management of points, levels and achievements. The objective is always to drive the user to become a safer piece in the urban transportation model and avoid potential hazardous consequences from prolong bad habits. 2
Applications of Gamication and Aective Computing Gamication and aective computing do not generally appear linked as subjects of research. Despite their dierences, these two themes can be combined in a strategic manner to steer human behavior as a positive inuence. 2.1
The use of aective computing for urban transport implies the detection of
emotional state or users. The emotional state can be interpreted in more ways
than one. There is the personality of users which inuences how users react to
events, the current mood of the user which implies a mid term emotional states
the current spirits of the user and emotion which is viewed how a user reacts to
an event and may or may not alter his mood state. In terms of computational
representation, the PAD space is a useful mathematical space where emotions can
be represented [
Taking this in consideration, in urban transports there are dierent types of events which correlate in some aspect with the before-mentioned aective attributes. For starters, there is the driving style of an user. This is typically viewed as immutable manner in which a driver operates a vehicle. It can consist of a relation of average speed, break and acceleration times across multiple records of driving from an user in a simplied the approach. Then there are the current driving style for a particular period of recordings such as a trip. And nally, we have direct events which occur while driving which may or may not aect how we drive. Some driving events are known to aect profoundly how we drive, as an example we can think of episodes of road rage where some event from one driver such as lane cutting can trigger aggressive and dangerous behaviors from other drivers.
Directly comparing the previous two paragraphs, there is an obvious link between emotion and driving patterns that can be used to research the emotional link between the act of driving and positive or negative emotions. Taking cues from the driving style of the user we can infer some aspects of the users personality and the current driving style from a trip can be used to estimate users mood. Of course, this mood cannot be directly inferred, but it can be analyzed beforehand and validated by specialists from elds of psychology. For instance, a user with a tendency to drive at higher speeds may be inferred to be calm when he drives at lower speeds, but on the other hand a tendency to drive at lower speeds, when driving more slowly it can be perceived as uncertain or undecided not necessary more calm.
The aective computing here depicted is not all related to the user frame of mind but with our interpretation of how a is emotionally responding to driving. Taking in consideration driving proles inferences regarding driving personalities can be made. 2.2
The eld of gamication is characterized by the use of game elements on non
game applications. This can be translated as to use game components
encountered in traditional games such as points and levels to real applications such
as tracking fuel consumption [
Gamication is often used as a motivational and persuasive tool to inuence behavior in a manner that promotes the objectives of the designer of the application system. When used to a community of users it fosters the competition of users towards a dened objective that is viewed as important.
In the eld of urban transportation gamication can be used as reward tool to inuence drivers to acquire safer driving styles [6]. Coupled with techniques from ubiquitous sensing, it can even target driving styles and emotional state of users, contributing to a safer environment in urban transports. 3
The design of this prototype uses the PHESS Driving application, previously developed for assessing user driving events. The idea of this prototype application is to have a system that reward points to user dynamically according to positive or negative emotion, translated by aggressive and non-aggressive behavior as categorized on the PHESS Driving application [7].
Taking into consideration this system, already implemented to categorize driving styles and a gamication platform, experiments with aective computing are made in this research. Taking cues from aective computing and related projects, means for the translation of driving events to emotional states are implemented. Based on the detected emotions dynamic point systems are used to reward or penalize driving activities in two stages: while driving and after driving. The rst will act as a means to directly inuence driving actions while the second aims to raise driver awareness.
In architecture, sensor data obtained through a mobile application of the PHESS Driving is used to obtain the users driving data, build user driving proles and classify driving events.
The process ow is illustrated in gure 1. After the processing of user driving
data, an approximation of usual driving emotion is made based on the driver
prole which is then updated constantly by the data captured by sensors while
driving. Alert signs are displayed in the mobile application depicting points
awarded, thus giving feedback to the driver. After the driving trip is over, the
driver’s prole is also updated and as a consequence the driver’s initial emotion
for next iteration might be dierent as well.
The computational representation of mood states is done according to the
Pleasure, Arousal, Dominance framework described by Mehabian (PAD) [4] and the
PAD space extended by Gehbard in [
Emotion classication is based on the analysis of the number of breaks and accelerations detected by mobile sensors: aggressive driving styles are connected with high frequency of breaking and accelerating actions; relaxed driving is correlated with stable velocity and low breaking and accelerating actions.
The update of emotions is a path between the PAD space where a current emotion is inuenced by events and direct the state of the user towards a nal emotion as a response to the event. The velocity of emotional change is regulated by a variable emotionalW eight which will be higher if the emotion and the event have both either negative or positive classications. When the classication dier then its value is lower simulation inertia in state change.
Eactual = Eactual + (Efinal
Eactual) emotionalW eight (1)
Following this approach, two nal states are dened representing angry and calm states so that a driver’s emotion may be updated towards one of these states. The equation 1 represents the current driver’s state moving in a vector space at a velocity dened by the emotionalW eigth.
For the purpose of this study, the nal emotions considered are dened in table 1.
This study is limited to a classication of emotion states oscillating between pleased and angry emotions. The classication of additional emotions is being addressed in an extended research from this work. 3.2
The use of emotions in gamication has the objective to become more natural and understandable to the user. Therefore points while driving are only attributed when the dominant emotion is classied as positive by the PAD space. Though there are currently only two nal emotions being addressed, intermediate emotions can still exist by the paths the initial emotion makes in the PAD space.
After the trip, an assessment of the trip is made according to the classication of the average number of accelerations and decelerations per unit of time. Based on these inputs the PHESS Driving system can classify the driving session as aggressive or normal and additional points are generated to the user.
The achievements are automatically generated by rules dened in the gamication platform. Examples of current rules include the appearance of achievements related to positive trips after more than 10 negative driving sessions and reinforcement achievements as streaks achievements for positive driving sessions. 3.3
In order to demonstrate the system, here a trace of the trip from a user in the platform is described. The user in question starts from a negative user prole and has a driving session of about 10 minutes generating about 226 driving events. From the results of this driving session, in gure 2, it demonstrates the emotional path derived and a partial sample of the driving trajectory.
From the emotional path we can derive that the user arrives at a positive stage early in the trip around the second minute, from when points are generated during its trip. At the end of the trip the analyzing the the driving session, more points are awarded due to the fact that this is in fact a positive driving session. Nevertheless, after the update of the user prole, its starting emotion is still negative and therefore, there is a penalization because the user driving session history is still aecting its user prole categorization.
Additional work is being done to add more dynamics to the gamication structure The idea is to pick up these experiments and add more point and achievement strategies to positively inuence the user. Gamication can be a useful tool for user persuasion, using reward methodologies or more advances methods such as information diusion through user social networks. In the case of user driving habits it may inuence how the driver reacts and inuence them to positive attitudes. The use of aective computing, demonstrated in this article aims to make the gamication more social and understandable by the users. Linking the emotions to events and using predominant emotions to steer gamication rewards is an innovative idea for user persuasion.
The aim of this research is to continue the study of emotion perception by short questionnaires to the end of each trip and relate user feedback to intercepted notions from the aective system. Emotion linked to driving events is also being explored, in order to expand the current range and make the emotion space more dynamic. The study of aective persuasion will be monitored to be compared with traditional gamication to access dierences or similarities in user persuasion.
4. Mehrabian, A.: Pleasure-arousal-dominance: A general framework for describing and measuring individual dierences in Temperament. Current Psychology 14(4), 261292 (1996) 5. Nettle, D.: Personality : What makes you the way you are. OUP Oxford (2008), http://books.google.pt/books?id=PimuSGj1U_gC 6. Silva, F., Analide, C.: Gamication and the Improvement of Urban Sustainability.
In: Ambient Intelligence and Smart Environments Series. pp. 446455 (2016) 7. Silva, F., Analide, C.: Ubiquitous driving and community knowledge. Journal of Ambient Intelligence and Humanized Computing pp. 110 (aug 2016), http://link.springer.com/10.1007/s12652016-0397-9 http://link.springer.com/article/10.1007/s12652-016-03979%5Cnhttp://link.springer.com/content/pdf/10.1007%2Fs12652-016-0397-9.pdf