The Role of Artificial Intelligence
By Dr. Manouchehr M. Khorasani
Artificial intelligence plays a crucially important role in our lives, today. There are discussions about having driverless cars (fully automated cars), the automated hiring process for companies via computer programs, and human cloning to name a few technological innovations that will change our lives soon.
There are many other innovations also. Many cities are now under camera surveillance to provide more security and safety to their citizens. There are discussions about abolishing money and replacing it with a digital currency. We already have drones implemented in warfare today, and there are discussions and research about implementing them as robot infantry soldiers, in the future. There are ideas about introducing household robots to take care of our everyday chores. There are discussions about the introduction of driverless cars by 2030, where cars drive autonomously and do not need a human driver.
Does it sound like a science fiction movie? If you think about it for a moment, you realize that it does not, and we are heading fast towards a complete digitalization of our lives. Before condoning or condemning the whole process of digitalization, we should be aware that although digitalization makes our lives easier and more convenient, there are many ethical issues involved that need to be addressed.
State of driverless cars
Many big car manufacturers such as BMW, Mercedes, Audi, General Motors, Toyota, and Ford, and even Google are actively working on building driverless cars or fully automated vehicles. The presence of a human driver is no longer needed. The planning of most car manufactures involves reaching the level of fully automated cars by 2030. But Huawei, a Chinese multinational company is planning to launch driverless cars into the market, by 2025.
There are different definitions of the stages of a driverless car as described by the US Society of Automotive Engineers (SAE). These are as follows:
- Level 0 (No driving automation): Cars at this level have no driving automation functions, and this stage represents a traditional vehicle.
- Level 1 (Driver Assistance): At this stage, the car has some automated functions, such as an automated braking system before hitting an obstacle.
- Level 2 (Partial Automation): At this level, the car can perform automated braking, accelerating, and changing lanes, however, the driver must always be present in a monitoring position to take over control when it is necessary.
- Level 3 (High Automation): At this stage, no driver is needed for continuous monitoring of the system. Under some circumstances, the automated system can operate on its own. But it provides the driver with a warning time of some seconds before they assume back control of the car.
- Level 4 (Full Automation): At this level, the car system performs all driving functions under standard conditions, and it does not need any driver to take control under normal conditions. One can define non-standard conditions, such as bad weather.
- Level 5 (Autonomous driving): The car does not need any driver in any condition as it can perform all driving functions.
Currently, there are no commercially available autonomous cars that go beyond Level 2 or Level 3 automation. Although the introduction of driverless cars will reduce road accidents, we cannot ignore the ethical considerations in this case. But let us discuss the advantages of a driverless car first.
Advantages of driverless cars
Let us think about some advantages of driverless cars. I will highlight the benefits as follows:
- Elimination of human error: Different studies demonstrate that 94 percent of road accidents are due to human error. No matter how experienced drivers are, humans are prone to committing mistakes. The Advanced Driver Assistant Systems (ADAS) are programed with active safety technologies, including steering or braking. It means that in case of a dangerous situation caused by different behaviors such as using cell phones during driving, speeding and aggressive/defensive driving, and driving under the influence of alcohol or drugs using, ADAS takes over the driver’s role. In such circumstances, ADAS helps us reduce the possibilities of an accident in such cases.
- Mobility for the elderly and disabled citizens: Driverless cars will provide more mobility for the elderly and disabled citizens who are not able to drive themselves. According to the Center for Disease Control and Prevention, more than 60 million people live with a disability in the United States alone. Senior adults (mostly above 65 and over) account for 40 percent of those with disabilities. Having a good vision is a prerequisite for driving, and 4.5 percent of adults in the United States have a vision disability. We are not only talking about physical disabilities because circa 11 percent of these people suffer from cognition disability. It includes trouble concentrating and making decisions. There are also circa 2 percent of adults with autism condition based on a report of the Centers for Disease Control and Prevention in 2000. Driverless cars will provide these citizens with more mobility in the future.
- Increase productivity: Using driverless cars, drivers can reallocate their time to other activities instead of piloting the vehicle. Additionally, passengers do not lose time trying to find a parking place.
Ethical issues of autonomous cars
As for the ethical concerns regarding autonomous driving, I would like to introduce the well-known ethical experiment, namely the trolley problem. The trolley dilemma in the ethical research deals with the predicament a switchman is facing. A runaway trolley is heading towards some workers on the tracks. The switchman can pull the lever and lead the trolley towards track 1, where one worker is standing, or he can lead it towards track 2, where three workers are working. In such a case, the doctrine of double effect tells us that it is always better to sacrifice one person to save three lives. In the future, driverless cars will surely face such situations and the algorithms need to face such issues. Imagine the driverless car is driving with three passengers on board. Suddenly, one pedestrian jumps in front of the vehicle. There is not enough distance and time for the car to brake effectively. Thus, the vehicle cannot avoid a collision. It either should drive on the embankment to avoid the collision with the pedestrian running the risk that all car passengers will be killed, or it should run over the pedestrian. We can also reverse the situation, meaning that the car has one passenger and three people try to cross the road. In both cases, a utilitarian approach prescribes that one should maximize the overall benefit of all parties involved. Hence in both cases, no matter which solution one chooses, it leads to a negative effect (double effect).
In such cases, the driverless car should be programmed in a way that the lowest number of human lives would be claimed. It means that we are dealing with the first law for programming a robot. A robot may neither kill people nor allow the death of a person through its passive action (Asimov, 1950, p. 40). With one passenger in the car, the vehicle should drive against the embankment and kill its own passenger/driver to save the lives of three passengers on the road. If the car has three passengers, the death of the pedestrian should be taken into account and accepted. The major problem of utilitarianism is how we should define utility. Avoiding a high number of fatalities is the most obvious criterion.
The problem starts when we change variables of the people standing on the track, such as age (a child is standing on track 1 and 3 senior citizens on track 2), gender, education, etc. It means that the way we define the utility would change the outcome, such as whether the utility of a child’s life should be regarded higher than a pensioner or the utility of a highly-qualified physician should be considered higher than a prisoner, and so on. MIT’s Moral Machine takes different utility criteria into account, such as gender, age, health status, social value, etc. But anyone can see this is not compatible with the principles of human dignity as any preference would violate the moral and constitutional principles stressing the equality of all people.
Therefore, different ethical committees consider any application and programming of utilitarian principles as unethical. Other alternatives exist, such as allowing the driver to decide in a configuration menu to determine the reaction of the vehicle in a dilemma situation when they are in a stress-free situation. The reason is that in a stressful and critical situation, we are normally not capable of making rational decisions. But even this configuration allows the driver/passenger of a driverless car to decide about life and death situations. Another alternative would be to allow a random number generator to decide and break the moral dilemma. But few people would trust a machine to make ethical decisions on their behalf, and hence the automated ethical configuration of driverless cars would decrease the confidence of potential buyers. A 2016 study on autonomous driving in the USA shows that most participants prefer a utilitarian approach to program the algorithm of the autonomous car when the purchasing decisions of others are concerned, but if they want to buy a car, they want the algorithm to protect the car passengers in all cases. This shows again the paradox of the whole issue.
There are many advantages of autonomous driving cars, such as providing mobility to senior citizens, reducing car accidents, and increasing productivity, to name a few. But many ethical issues need to be addressed as even the best autonomous systems could and will cause road accidents. Therefore, moral principles need to be programmed into the algorithm of the car. The challenge is no matter which ethical framework we consider, we will have a problem as there is an incalculable number of scenarios on the road. Therefore, a car that puts the safety of its passengers above everything else is just as unacceptable as a car that sacrifices its passengers to save other road users and pedestrians. Therefore, we need tight regulations, we need well-defined moral and ethical guidelines, and we need a public discussion to address this issue. In unavoidable accident scenarios, any utility definition based on personal features such as age, gender, the physical or mental constitution should be strictly prohibited. As different ethical committees have determined, general programming to decrease personal road injuries may be justifiable.
The role of artificial intelligence is a series of 5 articles:
Part 1: Driverless Cars
Part 2: Automated hiring process
Part 3: Human cloning
Part 4: Surveillance
Part 5: Ethical issues of robots in warfare
Doctor Manouchehr, Moshtagh Khorasani is a lecturer at Frankfurt School of Finance and Management and International School of Management (ISM). He teaches cross-cultural leadership and ethical leadership for MBA and BBA classes.