[Paper Review] Human decisions in moral dilemmas are largely described by Utilitarianism: virtual car driving study provides guidelines for ADVs

Maximilian Alexander Wächter, Anja K. Faulhaber+7 more|arXiv (Cornell University)|Jun 22, 2017

Psychology of Moral and Emotional Judgment4 references3 citations

TL;DR

This study investigates human moral decision-making in autonomous vehicle (AV) scenarios using a virtual reality driving simulation. Participants consistently chose options minimizing overall harm, strongly aligning with utilitarian principles, even when faced with age, group size, or self-sacrifice variables—supporting utilitarianism as a viable ethical framework for AV decision algorithms.

ABSTRACT

Ethical thought experiments such as the trolley dilemma have been investigated extensively in the past, showing that humans act in a utilitarian way, trying to cause as little overall damage as possible. These trolley dilemmas have gained renewed attention over the past years; especially due to the necessity of implementing moral decisions in autonomous driving vehicles. We conducted a set of experiments in which participants experienced modified trolley dilemmas as the driver in a virtual reality environment. Participants had to make decisionsbetween two discrete options: driving on one of two lanes where different obstacles came into view. Obstacles included a variety of human-like avatars of different ages and group sizes. Furthermore, we tested the influence of a sidewalk as a potential safe harbor and a condition implicating a self-sacrifice. Results showed that subjects, in general, decided in a utilitarian manner, sparing the highest number of avatars possible with a limited influence of the other variables. Our findings support that human behavior is in line with the utilitarian approach to moral decision making. This may serve as a guideline for the implementation of moral decisions in ADVs.

Motivation & Objective

• To investigate how humans make moral decisions in autonomous vehicle (AV) scenarios involving life-or-death trade-offs.
• To assess whether human behavior in such dilemmas aligns with utilitarian ethics, prioritizing the greatest good for the greatest number.
• To evaluate the influence of variables like avatar age, group size, and presence of a safe sidewalk on human decision-making.
• To provide empirical guidelines for implementing ethical decision-making systems in real-world autonomous driving vehicles (ADVs).

Proposed method

• Conducted a virtual reality driving experiment where participants acted as drivers in simulated trolley-like dilemmas.
• Presented participants with two-lane choices, each leading to different numbers of human-like avatars of varying ages and group sizes.
• Incorporated conditions with a sidewalk (potential safe harbor) and self-sacrifice scenarios to test ethical trade-offs.
• Collected and analyzed decisions to determine whether choices followed utilitarian logic (minimizing total harm).
• Used statistical analysis to assess the impact of demographic and environmental variables on decision outcomes.
• Applied behavioral data to infer ethical decision-making patterns and compare them to utilitarian theory.

Experimental results

Research questions

• RQ1 Do human participants in a virtual driving scenario make decisions that prioritize minimizing total harm, consistent with utilitarianism?
• RQ2 How do factors such as the age of avatars, group size, and presence of a sidewalk influence moral decisions in AV dilemmas?
• RQ3 Does the inclusion of self-sacrifice in a scenario alter participants' tendency toward utilitarian decision-making?
• RQ4 To what extent do real-world human decisions in simulated AV dilemmas align with utilitarian ethical frameworks?

Key findings

• Participants overwhelmingly chose the option that spared the highest number of avatars, demonstrating a strong utilitarian decision pattern.
• The presence of a sidewalk or self-sacrifice condition had minimal influence on decision outcomes, indicating that total harm minimization remained the dominant factor.
• Age and group size of avatars had a measurable but secondary effect, with participants generally preferring to save larger groups or younger individuals.
• The data showed consistent utilitarian behavior across diverse scenarios, supporting the use of utilitarian algorithms in ADVs.
• Even in self-sacrifice scenarios, participants did not consistently choose to sacrifice themselves, but still prioritized minimizing overall fatalities.
• Overall, human decisions were largely predictable by utilitarian principles, with only minor deviations influenced by emotional or contextual factors.