If uncertainty were simply the absence of certainty, the two ratings should mirror each other perfectly. They did not. The study, published in Frontiers in Psychology, found a clear negative relationship between the two feelings, but one that was far too weak to support the idea that they are opposite ends of a single scale.

The question sounds almost trivial, but it matters enormously for how scientists measure the human mind. Bipolar scales quietly assume that certainty and uncertainty are perfect inverses: if you are 80 percent certain, you must be 20 percent uncertain, and you can never be both highly certain and highly uncertain at once. Unipolar scales make no such assumption, allowing the two states to coexist or vary independently.

Until now, there has been no consensus in the human factors literature about which approach is correct, and methodological work has shown that the two types of scales cannot automatically be treated as interchangeable. To settle the question, the team turned to crowdsourcing. Participants were recruited through the platform Appen and paid 0.50 USD to complete the experiment online, an approach that traded the tight control of a laboratory for an unusually broad and diverse sample.

The final valid sample included 705 men, 718 women, and a small number of participants who selected other options or preferred not to respond, with an average age of about 35.6 years. The largest national groups came from Brazil, the United States, India, and the Philippines. Participants from one country, Venezuela, were removed after the researchers identified a pattern of cheating.

Because the goal was to test a measurement relationship rather than estimate prevalence in any particular population, the authors argued that this international heterogeneity actually strengthened the study, making it more likely that any observed relationship would generalize across populations. The experimental stimuli were carefully engineered pieces of virtual driving. The team built ten 10-second videos in the game engine Unity, each filmed from the passenger seat of a Level 5 fully automated car, displayed with a blue overlay on the windows so viewers knew no human driver was in control.

The car traveled in the middle lane of a six-lane road at 30 km/h, except in one scenario where the speed was raised to 60 km/h as a potential provocation. Each video ended just before the car began a lane change to the left, a maneuver chosen because lane changes are among the most dangerous moves in driving. The surrounding traffic varied: in some scenarios the left lane was clearly safe, in others a car was overtaking or approaching, and in the most provocative situations other road users behaved unusually, such as overtaking from the right.

The scenarios drew on established instruments including the Driving Behavior Questionnaire and the Driver Anger Scale, and a five-person pilot study determined that 5-second clips were too short to evoke uncertainty while anything longer than 10 seconds caused viewers to lose focus. Each of the ten videos was shown twice, in randomized order, once so participants could rate their certainty and once so they could rate their uncertainty about the upcoming lane change. The questions were standardized: At the end of the scenario, the fully automated car changes to the left lane.

How certain, or how uncertain, would you feel? Participants answered on a slider running from absolutely not certain or uncertain to absolutely certain or uncertain, with no numerical labels shown on screen to avoid anchoring bias, although the software recorded values from 0 to 100 behind the scenes. The team deliberately asked about perceived psychological certainty and uncertainty, letting participants use their own understanding of those feelings, and participants had to move the slider before they could continue, preventing them from clicking through without giving real responses.

The analysis used a linear mixed-effects model, a statistical approach chosen because each participant rated multiple videos, producing observations that are nested within people and therefore not independent. Certainty scores served as the predictor, uncertainty as the outcome, with random intercepts for participants and a repeated covariance structure across scenarios. The result was unambiguous in direction: certainty had a significant negative effect on uncertainty, with an unstandardized coefficient of B = −0.320, a 95 percent confidence interval of −0.335 to −0.304, and an F-test of F(1, 14353.89) = 1610.83, p < 0.01.

In practical terms, every 10-point increase in certainty was associated with roughly a 3.2-point decrease in uncertainty. But if uncertainty were truly the mirror image of certainty, the coefficient would have been −1, and the confidence interval nowhere near included that value. Higher certainty reliably went with lower uncertainty, yet nowhere near one-for-one.

The descriptive statistics reinforced the point. Participants used the full 0-to-100 range for both ratings in every scenario, with standard deviations hovering around 36 to 39 points, so the finding was not an artifact of people clustering at one end of the scale. Mean certainty ratings ranged from 31.48 in the most provocative scenario, which featured an overtaking car followed by a second car about to overtake, up to 67.30 in a basic traffic scene, while mean uncertainty ranged from 34.43 to 62.43 across the same set.

As a robustness check, the model was rerun using only participants from predominantly English-speaking countries, the United States, the United Kingdom, Canada, Australia, Ireland, and New Zealand, to rule out language proficiency as an explanation. The effect persisted, with B = −0.432, still significant and still far from −1. What does it mean for two feelings to be related but not opposite?

The authors interpret the result as evidence that certainty and uncertainty are partially independent psychological dimensions that can coexist within a single person. A passenger might feel confident about one aspect of a situation, such as the car's general competence, while still feeling doubt about another, such as whether that approaching vehicle has been detected. This coexistence echoes earlier theoretical work distinguishing confidence from certainty as separate cognitive constructs, and it fits the broader definition of psychological uncertainty as a mental experience encompassing cognitive, emotional, motivational, and behavioral components, distinct from trust, which reflects a willingness to accept vulnerability despite uncertainty.

The authors suggest that holding both states simultaneously may even create cognitive dissonance, though testing that possibility lies beyond the present data. The practical implications reach straight into the design of automated vehicles. If certainty and uncertainty are not direct opposites, then an interface feature that successfully reduces passenger uncertainty will not automatically produce an equivalent boost in certainty, and a system that maximizes one may leave the other untouched.

Designers evaluating transparency features, communication strategies, explanations, or in-car displays should therefore measure both constructs separately rather than relying on a single bipolar dial. The same logic applies to researchers studying trust, which is widely recognized as a key determinant of whether people accept automated driving at all; if certainty and uncertainty contribute differently to trust and acceptance, a bipolar measure could obscure distinctions that matter. The authors caution, however, that both constructs were measured here by subjective self-report, that the video content was not systematically manipulated to isolate causal drivers, and that the study was not designed to detect cultural differences despite its international reach.

Future work, the team suggests, could experimentally vary the ambiguity and informational clarity of driving scenarios, add behavioral and physiological measures such as response times, eye-tracking, or neural indicators, and use more advanced models with random slopes to test whether the certainty-uncertainty relationship varies across individuals, perhaps reflecting differences in ambiguity tolerance or metacognitive awareness. For now, the message is deceptively simple: the mind does not run on a single certainty dial.

Subject of Research: The relationship between psychological certainty and uncertainty in passengers of fully automated vehicles

Article Title: Dissociating certainty and uncertainty for passengers of autonomous vehicles: evidence against a perfect inverse relationship

Article References: Franssen, M., van der Spek, E. D., & Martens, M. H. (2026). Dissociating certainty and uncertainty for passengers of autonomous vehicles: evidence against a perfect inverse relationship. Frontiers in Psychology, 17, Article 1917625. https://doi.org/10.3389/fpsyg.2026.1917625

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1917625

Keywords: automated vehicles, certainty, uncertainty, crowdsourcing, passenger experience, human-automation interaction, psychological measurement, linear mixed-effects model, robotaxi, trust, unipolar scales, lane change