As autonomous vehicles become more common, the question of how to program artificial intelligence (AI) with ethical decision-making capabilities is becoming increasingly important. Unlike traditional driving, which relies on human judgment, self-driving cars must be programmed to make decisions in situations where moral dilemmas are unavoidable.
Wolf Schäfer, professor emeritus in Stony Brook University’s Department of Technology and Society, highlighted the complexity of this issue. “We are now dealing with problems that become not only internal problems for engineering, but are also now becoming problems that society recognizes as critical questions they would like answers to,” said Schäfer.
The challenge lies in the absence of a universal moral code for machines. While ethical frameworks such as libertarianism, utilitarianism, and Kantianism exist, implementing any one theory into algorithms can seem arbitrary. Differences in cultural values further complicate the creation of a globally acceptable system for vehicle control.
Schäfer has led an Automotive Ethics Vertically Integrated Projects (VIP) team since 2020 to address these challenges. He emphasized the need for changes in engineering education to include AI design and ethics: “We should use this time to plan the rapidly expanding AI sphere,” he said. He pointed out that about 40,000 people die each year in motor vehicle crashes in the United States—a number higher than those from homicides, plane crashes, or natural disasters combined—and nearly 10,000 crash victims visit emergency rooms daily due to car accidents.
To integrate ethical considerations into engineering training, Schäfer began building a lab in 2022 where students use model cars equipped with cameras and sensors on a racetrack. The cars respond to bar codes representing different scenarios—including moral dilemmas—based on student-written code.
“We’ll need to distinguish between moral, immoral and rightful machines,” Schäfer explained. “‘Immoral’ machines would be machines that do things that we consider immoral. ‘Moral’ machines are machines that follow certain ethical rules or conventions. And ‘rightful’ machines would be created by the societal certification of new technology to be allowed on public roads.”
He added that current engineering education does not yet fully address these issues: “It’s not like we have Newtonian or other physical laws… But with ethical rules, there’s variation.” Schäfer believes technical skills alone are insufficient: “We’re running into problems that cannot be solved with just technical skills… AI is too important to be left to computer science alone.”
Schäfer advocates for collaboration between engineering students and scholars from humanities and social sciences: “The VIP program and our particular project point to a solution of engineering plus applied humanities and social sciences,” he said.
Ammar Ali ’26 joined the Automotive Ethics Lab out of interest in autonomous vehicles and personal experience using Tesla’s self-driving features. He leads development of a machine learning model focused on ethical decision-making within simulations at Stony Brook University. Ali noted the multidisciplinary nature of the research: “Besides being able to use the most advanced computing resources of Stony Brook, this research is truly multidisciplinary… it’s something that I view as far beyond the typical undergraduate level.”
Ali also described how working within this professional environment helps prepare students for future careers: “In addition to expanding my technical skills… I also hope to build on my leadership experience. Ultimately, I aim to influence how intelligent systems are built, and the Automotive Ethics Lab is preparing me to do so.”
Schäfer brings his background as a historian of science and technology into his work at Stony Brook University; he notes upcoming changes reflecting these priorities: “This year, the Department of Technology and Society will become the Department of Technology, AI and Society.” According to him: “Cars today are computers on wheels… But these computers on wheels have great computing capabilities… Humans couldn’t react that quickly. But the AI in a car can recognize that it has options in a dilemma situation. And if we have options, those are all either political or moral choices.”
He concluded by emphasizing ongoing uncertainty about programming morality into AI: “2000 years of philosophy have not produced a universal moral theory… Depending on how it’s programmed, not every AI-driven car would make the same decision. We are the ones who will try to get as close as we can in terms of translating philosophical theories into algorithms. Engineering has become too important to be left to the engineers.”
