When it comes to autonomous technology of any kind, the first thing that often comes to our minds is our safety, our well-being and our survival. It’s not ridiculous for us to have these concerns and ask the hard questions, such as who is responsible should a death result from an edge case accident, what is an acceptable level of autonomy and what isn’t, and how does this technology come to a decision?
With driverless cars – a prime example of autonomous technology – starting to be deployed on public roads across the world, we must seek answers to these questions sooner rather than later. The ethical dilemmas we face with driverless cars now will be similar to the ethical dilemmas we will need to face later on, the more autonomous technology becomes normalised in our lives. Facing these issues head on now could help us get a head start in the many ethical issues we will need to face as technology becomes ever-more high-tech.
MIT researchers are confronting the ethical dilemmas of driverless cars by playing out hypothetical scenarios when it comes to autonomous cars making decisions on the safety of their passengers and people the car makes contact with on the street, and choosing between the less of two evils. The viewer must judge which decision they would make if placed in a particular intense scenario. This data is then compared with others and made publicly available.
The researchers are gathering many people’s views on what they consider acceptable and not acceptable behaviour of an autonomous car, and which decision is taken when making the impossible choice of sacrificing one life over another’s. As alarming as it is, this research could be used to help data scientists and engineers gain a better understanding of what actions might be taken should a far-fetched accident occur. Of course, avoiding the far-fetched accident in the first place is a bigger priority, but the research is a step towards facing the issue head on rather than believing that engineering alone is going to solve the problem.
Some proposed ideas for minimizing the risk of self-driving car accidents include limiting the speed of autonomous cars beyond that of speed limits in certain densely populated areas and having a designated right of way for these cars. More sophisticated mechanisms for this include using machine learning to continuously asses the risk of an accident and predict the probability of an accident occurring so that action can be taken pre-emptively to avoid such a situation. The Center for Autonomous Research at Stanford (a name suspiciously chosen for its acronym CARS, it seems) is looking into these ideas for “ethical programming”.
Putting in place ethical guidelines for all those involved in the build, implementation and deployment of drivereless cars is another step towards dealing with the ethical dilemmas. For example, the Federal Ministry of Transport and Digital Infrastructure in Germany released ethical guidelines for driverless cars this year. The ministry plans to enforce these guidelines to help ensure driverless cars adhere to certain expectations in behaviours. For example, one guideline prohibits the classification people based on their personal characteristics such as race and gender so that this does not influence decision-making should an accident occur.
Transparency in the design of driverless cars and how algorithms come to a decision is another way to work through the ethical dilemmas of driverless cars and other autonomous technology. This includes consumers of these cars, and the general public, having a right to contribute to the algorithms and models that come to a decision.
A child, for example, might have a stronger weight than a full grown adult when it comes to a car deciding who gets first priority in safety and survival. A pregnant woman, for example, might be given priority over a single man. Humans are the ones who will need to decide what kinds of weights are placed on what kinds of people, and research like MIT’s simulations of hypothetical scenarios is one way of letting the public openly engage in the design and development of these vehicles.
As data scientists, we hold great responsibility when building models that directly impact people’s lives. The algorithms, smarts, rules and logic that we create is not to far off from a doctor working in an emergency situation who has to make critical decisions in a short amount of time. Understanding the ethical concerns of autonomous technology, implementing ways to minimize risk and then programming the hard decisions is by no means a trivial task. For this reason, ethics require proper study and training, and attention to detail.