A Difference In Attitude Before And After The First Fatal Accident With An Autonomous Vehicle

Abstract

1. Introduction

Since Nicolas-Joseph Cugnot and later Karl Benz came up with the first motor vehicles, the evolution of this device has been undeniable and what we think was a utopia it’s now a reality. “According to the most general definition, an autonomous vehicle (AV) is a vehicle that uses a combination of sensors, cameras, radar and artificial intelligence (AI) to travel between destinations without a human operator” (Rouse, 2018). Several companies have developed a vehicle in which the role of the driver has disappeared. A car in which all the users will become passengers. In a few years, the autonomous cars will be taking the roads and transporting adults and children in a daily basis.

Nowadays, there are a lot of companies which are developing this technology; Waymo, formerly a Google division, one of the leading companies in the field with its Self-Driving Car Project, has been working for a long time in making autonomous cars safer than human drivers on the road. “The Swedish car manufacturer company, Volvo has started to test 100 of its autonomous cars on public roads driven in normal traffic by regular clients by 2017” (Ilková & Ilka, 2017). However, we know that there are still several aspects in terms of safety that must be resolved.

The use of the autonomous car by individuals is just around the corner, but when will the autonomous cars really be implemented? Are we prepared to accept the circulation of autonomous cars on our roads? There are many doubts related to the security, legal and ethics issues.

The autonomous cars are under skepticism and distrust of the public. This new technology has been conceived to improve productivity, speed up traffic jams and offer users a level of safety that manual cars do not have. Currently, autonomous vehicles are being continuously tested and even used by companies, like Uber, that find in the autonomous vehicle a great ally. However, the automated driving is not yet fully developed. On March 2016, something paralyzed the entire sector and made the public reconsider the safety of these devices: an autonomous vehicle caused the first mortal accident of a pedestrian in Arizona. If we also consider not fully automated cars, this is not even the first fatal accident in which autonomous vehicles are involved. In May 2016, the driver of a Tesla Model S (using Autopilot, not a fully autonomous system) died after he crashed into two other cars.

After these fatal accidents, new methods have emerged; models trying to solve these problems of security and mistrust. Along with these investigations and models there are also moral and ethical dilemmas that try to regulate the actions of both, the automatic car and its driver, in normal situations as well as in risky situations.

The autonomous vehicle seems to have come to stay for two fundamental reasons. It follows the laws (except when there is a failure). And it eliminates the human factor, the misleads, which are present in more than 80% of traffic accidents, according to data from the CEA (European Automobile Commission, 2018).

2. Problem statement

Right now, the actual issue that researches want to solve is the acceptance of the autonomous cars by drivers and pedestrians. A key issue with automated driving at this stage of its development is that it is not yet reliable and safe, especially when combined with conventional vehicles (Martens & van den Beukel, 2013). We are in a very early stage of this technology.

After the fatal incident with the Uber’s autonomous car, our goal is to measure if there has been any change in the attitude that people have toward the autonomous car. We want to investigate if the statements that were made before the accident still hold today.

Due to the fact that there are few researches that deal with the latest developments in terms of safety, in this thesis, we will address the issue of safety, mistrust and lack of acceptance in autonomous cars and how these variables have affected to the public opinion in contrast with last year results.

3. Research questions

Most of the traffic accidents are caused by driver failures. Autonomous vehicles seek to eliminate this human factor and make the world’s roads safer for drivers and pedestrians, as well as make driving a much more comfortable activity. However, in this research, we want to resolve certain issues related to the perspective offered by the autonomous car to the public. For that reason, we want to know, in the first place, if people know that the first fatal accident related to the autonomous car has already occurred. The main goal of the investigation is to know if, after the incident, people have changed their mind.

How customers can solve these problems and start to trust and accept in the AVs? (this is necessary for the develop of the AVs, the trust and the utilization of the AVs by the public to improve the regulations and performance of the cars.)

4. Relevance/Scientific motivation

These driver assistance systems, also known as Automated Driving Systems (ADS), want to eliminate, or at least reduce, the influence of errors humans during driving, which are the cause of most of the traffic accidents.

The motivation of this work is to contribute to the works which has been already developed about the public opinion towards the problems of acceptance and trust in the autonomous car.

5. Structure of the thesis

Present the principal issues about the safety of the autonomous cars.

Present different levels of automation and the Acceptance models

Present the social dilemma and ethical issues.

How different models and researches are trying to shed light to this kind of problems

Present the acceptance and trust levels of the public towards autonomous cars in the previous research.

Present the current results

Compare the results

6. Literature review

1. Road Safety

After having reviewed the literature that is available today about the safety issue and the level of acceptance of the autonomous vehicle.

In the last years, the car has grown by leaps and bounds. This evolution has not focused solely on getting faster and more powerful vehicles, but also safer vehicles. With the increase in the use of cars, one of the biggest problems of society has also emerged, deaths due to traffic accidents.

According to the annual report by the WHO (World Health Organization, 2018) on road safety “deaths from road traffic crashes have increased to 1.35 million a year. That’s nearly 3 700 people dying on the world’s roads every day.” These data are alarming and explain why nowadays the development of safety systems and assistance to the driver are being more and more investigated.

According to the recent reports of the NHTSA (National Center for Statistics and Analysis, 2018), in the first quarter of 2018 7,950 people died in motor vehicle traffic crashes. Even though the percentage of deaths has decreased by 3.6 percent compared with last year’s same period, these figures do not cease to represent a reality. Traffic accidents are due to human errors.

A study by the University of Virginia Tech (Blanco, et al., 2016) sheds some light on this. Through the monitoring of the driving experience of 3,300 real vehicles that covered a total of 54 million kilometers, the scientists compared that information with the results obtained from Google’s autonomous car program. The results were enlightening: human drivers have an average of 4.2 accidents for every 1.6 million kilometers rolled, while autonomous cars reduce that figure by a quarter: to 3.2 accidents for every 1.6 million kilometers.

2. Levels of Automation

I would like to introduce now the different levels of automation that exist in autonomous vehicles. The first organization that introduce a classification was NHTSA (2013), which create a 0 to 4 categories levels of automation; Level 0 corresponds to no driving automation (a conventional car), and level 4 corresponds to a complete automation of driving, without any responsibility of the driver, and without it even necessary that there is a driver in the vehicle.

Also, in the same year a classification was made in Germany, by the BASt, the Federal Highway Research Institute (2013). In this classification the five levels of automated driving are: single driver, driver assisted, partial automation, high automation and complete automation.

Finally, SAE International (2014) introduced 6 levels of driving automation for the autonomous cars. These levels went from 0, where no automation was performed and only the human driver was on charge of the control of the vehicle, then in level 1 there were a driver assistance, but not yet the system takes any control. The next level was number 2, where the system takes partial decisions about the acceleration and deceleration of the car. From level 3 the autonomous car starts to have more autonomy and can perform monitor tasks of the environment. In level 4 the car reaches a high automation and shows some tactical responses. Finally, in level 5 the car is fully automated and there is no need of human intervention.

Nowadays, the standard created by the SAE is accepted, in fact the NHTSA has abandoned its own classification and has adopted SAE’s system in September 2016.

3. Importance of Trust

In the literature there are several definitions of trust, which began to be spread at the end of the 50s. One of the most famous definition of trust is the “expectancy held by an individual that the word, promise or written communication of another can be relied upon” (Rotten, 1967). Several authors had proposed a multidimensional definition of trust, but in my opinion, the one proposed by Ganesan (1994) is very relevant. This definition of trust differentiates between two dimensions; credibility, which consists in trusting in something that you conceived like effective, reliable and benevolence, and behavior, which is associated with the act of trusting in something but also implies vulnerability and uncertainty for those who trust.

Trust is identified as a fundamental factor in relationships with a purpose; when a person or group decides to trust in something, it is because they seek to fulfill an objective and have a positive expectation about the result of the interaction, “the operator’s goals” (Ji & Choi, 2015). Related to this concept, autonomous driving is based on trust and, therefore, on acceptance, since users seek to benefit from its advantages.

The optimal use of new technologies is a necessity, given their influence in daily life. These technologies are getting closer and closer to entering the market, which means that we must know how to manage their use properly, but first we should trust in them. The issue of acceptance in technology has been widely studied and the autonomous car is another reason to revise the different model that allow us to identify the key issues of these devices.

Studying the conceptions about the trust that Mayer, Davis and Schoorman (1995) developed about the three dimensions of trust (ability, benevolence and integrity), we can also presents more significations; functionality, helpfulness and predictability, which correspond to the vehicle’s ability to successfully perform the task, the ability to provide relevant feedbacks and assistance to the driver and the anticipation of possible risks (Ji & Choi, 2015). Linked to these meanings, Ji and Choi (2015) creates again a new dimension in which the confidence in the autonomous car is divided into 3 new categories: system transparency (management of the vehicle) , technical competence (AV capabilities) and situation management (user’s control of the AV).

Finally, some researchers have also added the risk as another important factor that affects the confidence and acceptance of the autonomous vehicle (Mayer, Davis, & Schoorman, 1995). The perceived risk is another variable in the research method that Ji and Choi (2015) presented in their article, in which this variable “has a negative effect in the behavioral intention” and can be reduce by the trust in the AV.

3.1. TAM

The objective of this research is to find out the acceptance level of the users and no users of the autonomous car. There are several models to measure the acceptance of technology; but the first to speak about this topic was Davis (1989), who proposed the Technology Acceptance Model (TAM). This model, which analyzes how users accept and use a technology, has been highly tested in predicting the use of new devices. The model is based on two concepts, perceived usefulness and ease of use. These 2 variables determine the intention to use a technology. According to Davis (1989), the perceived ease of use can be defined as “the degree to which an individual; believes that using a particular system would be free from physical and mental effort” and the perceived usefulness is defined as “the degree to which a person believes that using a particular technology will enhance his or her job performance”. As many researches have already stated, when drivers use an automated vehicle, trust is the most important point of this relationship (Parasuraman, Sheridan, & Wickens, 2008), and because of that, the TAM method is said to be an adequate instrument for the evaluation of the acceptance of DAS (Ghazizadeh, Peng, Lee, & Boyle, 2012). Thanks to these acceptations we can determine that trust is the link between the acceptance of the autonomous car and the willingness of users to actually ‘drive’ them (Parasuraman, et al.,2008).

3.2. Trust Fall Method

According to most people, the trust fall is a social experiment in which you let yourself fall without knowing if someone will hold you or not. The conception of this method is very interesting if we compare it with the trust that we place in technology.

This method measures the willingness of the participants to trust (during an extreme situation) in a system which characteristics and limitations have been explained previously (Miller, et al., 2016). This means that we can know, if the participants would take action in a situation, that although it seems dangerous, should not need the intervention of the driver.

The method was conducted, along with the simulation of the driving environment, also with additional questionnaires, surveys and the creation of “a mental model relative to the trustworthiness of the automated system” (Miller, et al., 2016).

The result of the entire experiment was that the majority of the participant did not trust in the autonomous car’s actions. The results showed that people continued taking control of the vehicle even though they confessed that they expected the car to handle the situation (Miller, et al., 2016).

3.3. The Design

One of the measures taken by the designers of autonomous cars, to increase its acceptance, is to make the vehicle’s interface more reliable for the users. The vehicle provides information to the driver or to the passengers about the elements that are occurring around him. Most of the autonomous cars that are being used and tested currently, they have incorporated an alarm system that emits a warning signal when the driver must take control of the vehicle (Lee & See, 2004).

The design of the autonomous car system has to make the driver trust in it. First, giving information to the driver to take immediate actions and encouraging him or her to the cooperation in a determined situation.

A specific area of design is the Human-Machine Interface (HMI), which focuses on the way people interact with the machines’ UIs. As I have already said, if drivers verify that autonomous cars offer certain advantages, they will be more willing to use them. One factors that should improve on order to achieve this, is the adaptation of the AV interface to all types of users. “Integrating universal design early in the development of new technologies will ensure that self-driving cars will be usable by all people” (Ferati, Murano, & Giannoumis, 2018)

3.4. Ethic and social dilemma

The recent mortal accident by an autonomous car of Uber has demonstrated that AVs can fail in critical situations. These failures of the safety lead to situations in which the car has to choose between two alternatives; to avoid an accident or cause the death of the occupants of the car. In most of these two scenarios, both options will cause deaths, whether they are occupants of the vehicle, pedestrians or people who drive other cars.

Many experts consider that we have an ethical dilemma, since we want that a machine makes a moral decision about who should die, and which individuals should prevail over the others. A group of German experts has developed an ethical code (Federal Ministry of Transport and Digital Infrastructure, 2017) for autonomous cars that open a debate of legal and moral issues. The main conclusion of the report is that “human life always has the highest priority”. And, in the case that the autonomous system has to choose between “killing” a human or another, the artificial intelligence should not take decisions based on the ethical laws but apply the rules of traffic and minimize the damage as much as possible.

According to the research carried out by Bonnefon, Shariff and Rahwan about the social dilemma of the autonomous vehicles (2016), where the participants were subjected to an online survey, the main result was that “it would be more moral for AVs to sacrifice their own passengers when this sacrifice would save a greater number of lives overall” (Bonnefon, Shariff, & Rahwan, 2016). However, if only one pedestrian would be in danger of being kill, the contestants indicated that the AV should sacrifice this individual instead of the life of the passenger or passengers of the AV. The number of lives that the participants decided to save or not is a decision-making fact. Nevertheless, when participants were asked about the purchasing of a AV that would not prioritize the integrity of its passengers, many participants were hesitant. This indicates that although most people think about the common good and the established moral laws, when they compromise their lives or those of their loved ones, they change their opinions completely and prefer that the system prioritize and protect their lives (Bonnefon, Shariff, & Rahwan, 2016).