Fast, Accurate, But Sometimes Too-Compelling Support: The Impact of Imperfectly Automated Cues in an Augmented-Reality Head-Mounted Display on Visual Search Performance
DOI: 10.1109/thms.2023.3302152
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 impact of imperfectly automated augmented reality (AR) cues on visual search performance, addressing the trade-off between efficiency gains and automation bias. While AR head-mounted displays (HMDs) can significantly reduce search time by directing attention, automation errors can lead to human reliance on incorrect information. The research specifically compares three types of HMD cues—an AR arrow pointing to the target (world-referenced), a plan-view minimap highlighting the target (screen-referenced), and a constant icon depicting the target’s appearance (screen-referenced)—to determine their effectiveness under both perfect and imperfect reliability conditions. The experiment involved 56 participants searching a 3D environment containing 48 objects to locate a target viewed prior to each trial. Participants were divided into two groups: one received cues that were 100% correct, while the other received cues that were incorrect on 17% of trials, simulating machine vision errors. The study manipulated cue type and cue location (center vs. downward in the field of view) to assess their influence on search speed and accuracy. The design allowed for the analysis of how different cueing philosophies—world-referenced versus display-referenced—affected human performance when automation was imperfect. Results indicated that location-based cues (arrow and minimap) were more effective than the appearance-based icon cue in reducing search time, both overall and when the cue was correct. The AR arrow cue provided the greatest efficiency benefit due to its direct, world-referenced nature. However, this high effectiveness came with a significant cost: the arrow cue induced the strongest automation bias. When the arrow cue was incorrect, participants were more likely to blindly follow it and miss the true target compared to other cue types. Conversely, the icon cue, while slower, allowed for memory-free confirmation of the target’s identity, resulting in higher accuracy benefits relative to the control condition. The study also found that cue location influenced performance, with central placement generally offering reduced information access effort, though this was offset by visual clutter for complex cues like the minimap. The findings highlight a critical design implication for human-machine systems: the most efficient AR cues may also be the most dangerous when automation fails. The "too-compelling" nature of immersive, world-referenced cues like the AR arrow can amplify automation bias, leading to greater errors during the infrequent occasions when the system is wrong. This suggests that while AR can drastically improve search speed, designers must account for the potential costs of imperfect reliability, particularly with cues that strongly capture attention. The study underscores the need to balance cue effectiveness with safeguards against automation bias, especially in high-stakes environments where missing a target due to reliance on flawed automation can have severe consequences.
Key finding
While world-referenced AR arrow cues provided the greatest performance benefits during correct trials, they also caused the highest rate of automation bias errors during incorrect trials compared to other cue types.
Methodology
lab_experiment
Sample size: 56
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | — | — | — | 1 | 2026-05-07 |
| archive | success | canonical_url | — | — | 6 | 2026-06-06 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-04 |
| chunk | success | chunk | — | — | 1 | 2026-06-04 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-04 |
| enrich | success | — | — | — | 1 | 2026-05-07 |
| promote | success | — | — | — | 1 | 2026-05-07 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.