Could AI solve the trolley question?

KIM SEUNG-JUNG
The author is a professor of archaeology at the University of Toronto.

Last week, a major car accident killed nine people as a car drove backwards around City Hall Station in Seoul. The driver’s claim of a sudden unintended acceleration is controversial. According to data from the Korea Transportation Safety Authority, an average of 30 sudden acceleration cases take place annually, but none have been recognized since 2017.

As we enter the artificial intelligence (AI) era, we will likely see a situation where self-driving cars become commonplace thanks to technological advances toward the direction of minimizing human errors. Driverless vehicles have already been tested in many parts of the world, including Korea, and they are put to practical use in a limited way. The chance of traffic accidents will soon be significantly reduced, but if an accident occurs, there is no legal ground on who should be held accountable. Moreover, controversy has already erupted over how to program a self-driving AI to prepare for a moment of urgent judgment.

The complex dilemma of whom to protect first — passengers or pedestrians — and whom to choose when deciding the possibility of sacrifice based on the gender and age of pedestrians is a harsh reality faced by AI development engineers. The famous “Trolley Dilemma” is no longer just an ethical experiment.

In addition, solutions to such dilemmas may vary depending on countries and cultural backgrounds. Germany’s “Ethics Guidelines for Self-Driving Vehicles” and guidelines published in Korea stipulate that people should not be discriminated on the grounds of gender, age, race, disability and so on. But this will not be easy to enforce.

AI ultimately doesn’t know what it doesn’t know. It just answers questions. Technological solutions cannot be the perfect answer. The idea of creating a safer society by eliminating valuable human efforts may itself be the problem.