Park the Trolley and Let’s Address the Real Ethical Questions of Self-Driving Cars

Why is it that every time there’s a conversation about self-driving cars, some bright spark cites the trolley problem: the adolescent “yeah, but what if an autonomous car has to decide between killing an old woman or a child?” Please. Enough already. That philosophical dilemma, proposed in the 1960s to discuss moral intuition, has absolutely nothing to do with how autonomous driving systems work. Anybody who trots it out is simply revealing how little they understand the technology involved.

In the real world, Waymo, Cruise or Tesla or any other company are not engineering their self-driving vehicles to decide who lives and dies in the event of an accident. Don’t take my word for that, there’s enough peer-reviewed research about it in the links of my article. So while some people discuss irrelevant moral quandaries, self-driving cars go on racking up millions of miles in more than 30 cities around the world, with accident rates dramatically lower than those of human drivers.

Let’s be clear what autonomous vehicles do. They analyze routes, brake, correct and apply dynamic controls, and all with with one goal in mind: to minimize damage and avoid collisions. There are no pre-programmed sacrifices of any kind.

In fact, the specialized literature points out that the basic premise of the trolley problem, an inevitable loss, does not apply to automated driving systems, because most real decisions consist of managing uncertainties, applying braking controls and rethinking direction to avoid collisions whenever possible, not in “selecting” which life to sacrifice.

Academics who have debated this question — experts in ethics and technology — explain that the trolley analogy is deceptively simplistic and does not serve as a framework for thinking about the ethics of autonomous vehicles. The trolley problem is simplistic nonsense based on binary scenarios that have nothing to do with the real world, where cars have sensors and time to calculate solutions that reduce damage without the need for a one-or-the-other decision.

Continuing to discuss the trolley problem simply diverts attention from the real ethical and regulatory challenges of autonomous driving, such as liability under the law, assigning blame for complex failures and unequal access to safer technologies.

Similarly, MIT’s famous Moral Machine experiment, which many people absurdly believe is a “real test” of how to program a Tesla or a Waymo, says nothing about decisions that self-driving cars actually make in real time. It is simply an exercise in public opinion on hypothetical dilemmas, with millions of responses reflecting cultural prejudices and not engineering realities. What it found ranges from cultural biases and prejudices to contradictory intuitions, and is sociology, not engineering. No self-driving car uses anything remotely close to those criteria.

And most importantly, not a single line of code has ever been programmed to decide who lives or who dies. Self-driving cars do what any reasonable human would try to do in an emergency, but with greater anticipation and precision: first of all, they will be travelling at a safe speed; they will then brake, maneuver where possible, and avoid any collisions that can be avoided. That’s all. There are no hierarchies of victims, there are no dramatized ethical dilemmas, there are no moral algorithms.

In conclusion, using the trolley problem to understand the ethics of self-driving cars is counterproductive. Infantile, it distracts us from the real debate and reveals a profound lack of understanding about technology. If we really want to talk about ethics and autonomous cars, let’s leave aside the most worn-out, simplistic and disconnected argument from reality. Seriously: let’s park the trolley once and for all and start thinking a little more rigorously.

—

This post was previously published on Enrique Dans’ blog.

***