Overall effects of non-driving related activities’ characteristics on takeover performance in the context of SAE Level 3: A meta-analysis

Shi, Elisabeth; Bengler, Klaus · 2022 · Crossref

DOI: 10.54941/ahfe1002435

archive: archived pipeline: cataloged verified

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

Summary

This meta-analysis investigates how the characteristics of non-driving related tasks (NDRTs) affect driver takeover performance in SAE Level 3 automated driving systems. The study is motivated by the need to understand traffic safety implications when users engage in secondary activities during conditional automation, where the system expects the human to resume control upon request. Building on previous meta-analyses that focused on situational variables or general engagement, this research applies psychological theories of task switching and modality shifting to differentiate NDRTs based on their physical and cognitive demands. The authors hypothesize that specific task attributes—such as the need to free hands, visual obstruction, and cognitive similarity to driving—systematically influence takeover time. The researchers conducted a systematic search of IEEE Xplore, Web of Science, and APA PsycArticles, supplemented by screening references from the German Ko-HAF project. Inclusion criteria required full-text quantitative studies in English or German involving SAE Level 3 automation in passenger cars, where a manual driving phase followed the automated phase, and takeover time was measured. Effect sizes (Cohen’s d) were calculated for differences in takeover time between conditions. To assess cognitive similarity, NDRTs were coded based on five cognitive processing modules derived from Baddeley’s working memory model: phonological loop, visual system, spatial system, central executive, and deliberate long-term memory retrieval. The meta-analysis employed robust variance estimation methods using R to handle complex data structures with multiple effect sizes from overlapping samples, avoiding the information loss associated with averaging effects per study. The results indicate that engagement in active NDRTs significantly increases takeover time compared to passive monitoring (d = .663, p < .01). Among active tasks, the need to physically put away the task or free hands had the strongest effect on prolonging takeover time (d = .625, p < .01). Visual tasks also resulted in longer takeover times compared to non-visual tasks (d = .326, p < .05). Furthermore, greater dissimilarity between the cognitive processing modules required by the NDRT and those required for driving was associated with increased takeover time (d = .096, p < .05). These findings support the theoretical framework that task switching costs and modality shifting effects explain variance in takeover performance. The study concludes that task switching and modality shifting theories provide a valid basis for differentiating NDRT impacts on safety in Level 3 automation. Practically, the strong effect of physical disengagement suggests that system designers should minimize the need for users to handle remote devices or create mechanisms for quick task storage. The authors note that while these findings highlight systematic underestimations of takeover time when only monitoring tasks are considered, the analysis is limited to takeover duration and does not assess takeover quality. Consequently, inferences cannot be drawn regarding other automation levels or the safety outcomes of the takeover maneuver itself.

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-nvidia on 2026-08-10; verification: verified.

Topics

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

• takeover transitions
• manual
• temporal
• visual manual
• automation
• hands on hands off engagement

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: behavioral performance data
• Methodological Resource: measurement protocol
• Theoretical Contribution: conceptual framework