When people judge what happens on the road, they may be responding to more than whether a driver followed the rules or avoided a crash. New experimental research suggests that perceptions of the driver’s character can exert an especially strong influence on moral evaluation, even in ordinary, low-stakes traffic situations.
The study, published in Frontiers in Psychology, tested a framework that separates moral judgment into three components: the Agent, referring to the perceived character of the person acting; the Deed, referring to whether the action follows relevant rules; and the Consequences, referring to whether the outcome is favourable or harmful. Rather than relying only on dramatic trolley-style dilemmas, the researchers examined everyday traffic scenarios and compared text, conventional audiovisual material and immersive 360-degree simulations.
Testing everyday road morality
The researchers conducted two studies. In the first, nearly 4,000 professional ethicists were contacted and 796 began the research. After exclusions for incomplete responses, very low response variability and automated bot screening, 457 participants remained for analysis. Of these, 226 evaluated audiovisual scenarios and 231 evaluated textual versions.
Study 1 used a between-subjects design that varied presentation mode alongside positive or negative versions of the driver’s character, the deed and the consequence. Participants also completed a measure of their preferences for virtue ethics, deontology and consequentialism. This allowed the researchers to ask not only which features influenced moral judgment, but whether formally trained ethicists with different theoretical commitments evaluated the situations differently.
The second study moved into a more immersive environment. A total of 206 university students initially participated, with 146 retained after the predefined exclusions. Participants watched traffic situations in a 360-degree audiovisual gallery using eight high-definition projectors. The experiment used five scenarios, including encounters involving a pet, an intersection, a cow crossing the road, an ambulance on a narrow mountain road and a crowd managed by a fire marshal.
Researchers created 30 vignette variants across these scenarios. Each participant evaluated 15 experimental vignettes after a training example. Unlike the first study, this was a within-subject design, meaning participants encountered multiple combinations of driver character, rule compliance and consequences. The researchers analysed repeated ratings using linear mixed-effects models.
Character produced the largest differences
Across the first study, positively characterised drivers received substantially higher moral ratings than negatively characterised drivers. The estimated difference was 2.39 points, with a large effect size of Cohen’s d = 1.21. Positive consequences also increased moral evaluations, but by a smaller 1.05 points, corresponding to d = 0.53.
Rule compliance was more complicated. Across Study 1 as a whole, the difference between rule-following and rule-violating deeds was only 0.14 points and was not statistically significant. Presentation format, however, changed the pattern sharply. In audiovisual scenarios, rule-following actions were rated 1.66 points more favourably. In textual scenarios the direction reversed, with rule-following actions rated 1.37 points less favourably than violations.
Presentation format also changed the influence of character and outcomes. The moral-rating gap between positively and negatively characterised drivers was 3.72 points in audiovisual scenarios but 1.06 points in textual scenarios. Conversely, favourable rather than unfavourable consequences produced a 1.97-point difference in the textual condition, while the corresponding audiovisual difference was only about 0.13 points and was not distinguishable from zero.
These interactions suggest that the medium through which a moral event is experienced can alter which information becomes most salient. Visual depictions may make personality and behaviour particularly vivid, while written descriptions may give readers more opportunity to focus on how events ultimately turn out.
Immersion reinforced the importance of character
The immersive second study produced a clearer overall pattern. Participants rated positively portrayed drivers much more favourably than negatively portrayed drivers. For social acceptability, the estimated means were 6.58 versus 3.75, a difference of 2.83 points and an effect size of d = 1.16. For personal moral judgment, the corresponding means were 7.08 and 3.81, a difference of 3.27 points, again with d = 1.16.
Rule-following also mattered in the immersive environment, although less strongly. Social acceptability averaged 5.61 for rule-following deeds and 4.71 for violations, a difference of 0.90 points with d = 0.32. Personal ratings differed by 0.88 points, with d = 0.31.
Outcomes produced a similarly modest effect. Scenarios with positive consequences received social acceptability ratings 0.83 points higher than those with negative consequences, corresponding to d = 0.29. For personal acceptability, the difference was 0.73 points with d = 0.26.
The contrast is important. In the immersive experiment, all three dimensions influenced moral evaluation, but the effect associated with the driver’s portrayed character was several times larger than the effects associated with rule compliance or consequences.
Moral theories did not neatly divide participants
A further question was whether people who preferred different moral frameworks would systematically prioritise different information. Someone oriented toward virtue ethics might be expected to focus on character, a deontologist on rules and a consequentialist on outcomes.
The data did not support such a simple division. In Study 1, the researchers found no significant interactions showing that participants’ preferred moral framework broadly changed the effects of Agent, Deed or Consequence. The professional ethicists therefore showed notable convergence despite their differing theoretical commitments.
Study 2 reached a similar conclusion at the overall interaction level. For personal moral judgment, the interaction between vignette and moral framework was not statistically significant, and a Bayesian comparison strongly favoured the model without that interaction. Some planned within-framework contrasts were significant, including strong differentiation between positive and negative characters among participants classified as virtue ethicists. However, because the overall moderation test was not significant, the authors caution against interpreting these subgroup patterns as evidence that moral theory reliably changes the effects.
Why ordinary traffic ethics matters
Autonomous-vehicle ethics is often discussed through extreme situations in which a machine must choose between competing harms. Such cases are philosophically useful, but they are unlikely to represent most of the moral impressions people form about automated vehicles in daily life.
The present results suggest that road users may judge a vehicle not only by whether it obeys a formal rule or produces a good outcome, but also by whether its behaviour appears considerate, aggressive, careful or callous. For automated driving, qualities that people ordinarily describe as character may be expressed through observable driving styles such as acceleration, braking, yielding and courtesy at intersections.
That does not mean an autonomous system literally possesses virtues or vices. Instead, the research raises the possibility that human observers interpret patterns of driving behaviour through character-like categories. If later studies confirm this in first-person driving or passenger settings, designers may need to consider social interpretations of driving style alongside safety, legality and efficiency.
Important limitations
The study measured third-person moral judgments rather than actual driving decisions. Participants evaluated scenarios as observers, so the findings cannot establish how people would behave when personally responsible for a vehicle or exposed to real danger.
The scenarios were deliberately low stakes and did not involve lethal collisions. The relative importance of character, rules and consequences could change substantially when severe injury or death is possible. The researchers also note that some character manipulations were intentionally strong, which may have amplified the Agent effect relative to the other components.
Sample composition creates another boundary. The participant groups were primarily drawn from Western, educated, industrialised, rich and democratic settings. Traffic norms, attitudes toward authority and expectations of automated systems differ across societies, so the findings should not be treated as universal.
Finally, the research did not directly test autonomous vehicles. Its implications for automated driving are therefore prospective. The central empirical finding concerns how people morally evaluate human drivers in controlled traffic scenarios. Translating that result into vehicle design will require experiments in which people interact directly with automated systems.
A broader view of machine behaviour
The research adds a useful dimension to debates that often frame machine ethics as a contest between following rules and maximising outcomes. Human moral evaluation appears capable of incorporating both, while also responding strongly to what behaviour implies about the actor behind it.
In everyday traffic, that third dimension may be especially visible. The immersive experiment showed that favourable character cues produced much larger shifts in moral ratings than either rule compliance or collision outcomes. The next challenge is to determine whether this pattern survives when participants become drivers or passengers, when stakes rise and when cultural settings become more diverse.
Source Information
Study: Cecchini, D., Pflanzer, M., Cacace, S. & Dubljević, V. (2026). Goodness behind a wheel: moral judgment in realistic traffic simulations featuring Agents, Deeds, and Consequences.
Journal: Frontiers in Psychology, Theoretical and Philosophical Psychology.
Published: 28 September 2026.
DOI: 10.3389/fpsyg.2026.1885843.
