What Ethical Models for Autonomous Vehicles Don’t Address – And How They Could Be Better

There’s a fairly large flaw in the way that programmers are currently addressing ethical concerns related to artificial intelligence (AI) and autonomous vehicles (AVs). Namely, existing approaches don’t account for the fact that people might try to use the AVs to do something bad.

For example, let’s say that there is an autonomous vehicle with no passengers and it is about to crash into a car containing five people. It can avoid the collision by swerving out of the road, but it would then hit a pedestrian.

Most discussions of ethics in this scenario focus on whether the autonomous vehicle’s AI should be selfish (protecting the vehicle and its cargo) or utilitarian (choosing the action that harms the fewest people). But that either/or approach to ethics can raise problems of its own.

“Current approaches to ethics and autonomous vehicles are a dangerous oversimplification – moral judgment is more complex than that,” says Veljko Dubljević, an assistant professor in the Science, Technology & Society (STS) program at North Carolina State University and author of a paper outlining this problem and a possible path forward. “For example, what if the five people in the car are terrorists? And what if they are deliberately taking advantage of the AI’s programming to kill the nearby pedestrian or hurt other people? Then you might want the autonomous vehicle to hit the car with five passengers.

“In other words, the simplistic approach currently being used to address ethical considerations in AI and autonomous vehicles doesn’t account for malicious intent. And it should.”

As an alternative, Dubljević proposes using the so-called Agent–Deed–Consequence (ADC) model as a framework that AIs could use to make moral judgments. The ADC model judges the morality of a decision based on three variables.

First, is the agent’s intent good or bad? Second, is the deed or action itself good or bad? Lastly, is the outcome or consequence good or bad? This approach allows for considerable nuance.

For example, most people would agree that running a red light is bad. But what if you run a red light in order to get out of the way of a speeding ambulance? And what if running the red light means that you avoided a collision with that ambulance?

“The ADC model would allow us to get closer to the flexibility and stability that we see in human moral judgment, but that does not yet exist in AI,” says Dubljević. “Here’s what I mean by stable and flexible. Human moral judgment is stable because most people would agree that lying is morally bad. But it’s flexible because most people would also agree that people who lied to Nazis in order to protect Jews were doing something morally good.

“But while the ADC model gives us a path forward, more research is needed,” Dubljević says. “I have led experimental work on how both philosophers and lay people approach moral judgment, and the results were valuable. However, that work gave people information in writing. More studies of human moral judgment are needed that rely on more immediate means of communication, such as virtual reality, if we want to confirm our earlier findings and implement them in AVs. Also, vigorous testing with driving simulation studies should be done before any putatively ‘ethical’ AVs start sharing the road with humans on a regular basis. Vehicle terror attacks have, unfortunately, become more common, and we need to be sure that AV technology will not be misused for nefarious purposes.”

The paper, “Toward Implementing the ADC Model of Moral Judgment in Autonomous Vehicles,” is published in the journal Science and Engineering Ethics.

-shipman-

Note to Editors: The study abstract follows.

“Toward Implementing the ADC Model of Moral Judgment in Autonomous Vehicles”

Author: Veljko Dubljević, North Carolina State University

Published: July 6, Science and Engineering Ethics

DOI: 10.1007/s11948-020-00242-0

Abstract: Autonomous vehicles (AVs) — and accidents they are involved in — attest to the urgent need to consider the ethics of artificial intelligence (AI). The question dominating the discussion so far has been whether we want AVs to behave in a ‘selfish’ or utilitarian manner. Rather than considering modeling self-driving cars on a single moral system like utilitarianism, one possible way to approach programming for AI would be to reflect recent work in neuroethics. The agent–deed–consequence (ADC) model provides a promising descriptive and normative account while also lending itself well to implementation in AI. The ADC model explains moral judgments by breaking them down into positive or negative intuitive evaluations of the agent, deed, and consequence in any given situation. These intuitive evaluations combine to produce a positive or negative judgment of moral acceptability. For example, the overall judgment of moral acceptability in a situation in which someone committed a deed that is judged as negative (e.g., breaking a law) would be mitigated if the agent had good intentions and the action had a good consequence. This explains the considerable flexibility and stability of human moral judgment that has yet to be replicated in AI. This paper examines the advantages and disadvantages of implementing the ADC model and how the model could inform future work on ethics of AI in general.

This post was originally published in NC State News.