Detecting unacceptable behavior of an autonomous vehicle using electroencephalography

Bertheau, Maren A. K.; Herrmann, Christoph S. · 2025 · Crossref

DOI: 10.1038/s41598-025-18305-2

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the feasibility of using electroencephalography (EEG) to detect when an autonomous vehicle (AV) behaves in ways that are unacceptable or incongruent with a user’s expectations. The motivation stems from the potential for AVs to exhibit behavior that is legally compliant but socially or psychologically unacceptable, which can induce stress, anxiety, and distrust. Because questionnaires are impractical for assessing user perception during time-critical driving maneuvers, the authors explored event-related potentials (ERPs) as an unobtrusive, high-temporal-resolution physiological measure. Specifically, the research focused on left-turn maneuvers through oncoming traffic, a high-risk scenario where human risk tolerance often diverges from AV decision-making. The experimental design involved 33 participants observing simulated AV left-turn maneuvers in a laboratory setting. Participants first indicated whether they would turn or wait for a gap in oncoming traffic. The AV then either executed the turn or waited, creating conditions where the AV’s behavior was either congruent or incongruent with the participant’s decision. EEG data were recorded to analyze specific ERP components: N1, P2, N2, and P3. The study hypothesized that incongruent trials would elicit greater amplitudes in these components compared to congruent trials, reflecting increased cognitive control or conflict resolution. Behavioral results confirmed that participants rated AV behavior as significantly more acceptable when it was congruent with their own decisions. In the EEG analysis, the only significant effect was observed in the N2 component (251 to 431 ms), which showed increased amplitude during incongruent trials compared to congruent ones. No significant differences were found for the N1, P2, or P3 components. The N2 effect was fronto-central in topography and correlated positively with subjective acceptability ratings; participants who rated incongruent trials as less acceptable exhibited a more negative N2 amplitude. This suggests the N2 reflects the cognitive processing of unexpected or incongruent AV behavior, even when the visual field remained static (i.e., the AV waited rather than turned). The findings indicate that ERP-based devices, particularly those monitoring the N2 component, could potentially identify critical or unacceptable situations in human-AV interactions in real-time. This offers a pathway for AVs to adjust their behavior or provide explanations dynamically. However, the authors note that further research is required to translate these laboratory findings into practical applications, such as testing in field-like environments with mobile EEG devices and developing single-trial classifiers for ambiguous traffic situations. The study establishes a foundational link between psychophysiological markers and AV acceptability, highlighting the N2 as a promising indicator for future brain-computer interface applications in autonomous driving.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• automation surprise

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data, behavioral performance data
• Methodological Resource: tool software