Automated decision aids: When are they advisors and when do they take control of human decision making?
DOI: 10.1037/xap0000463
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 cognitive mechanisms underlying how humans integrate automated decision aids into their decision-making processes, specifically examining how automation reliability influences whether aids are treated as advisors or autonomous triggers. Motivated by the risks of automation misuse and disuse in safety-critical industries like aviation, the authors sought to determine if higher reliability leads operators to grant automation more autonomous control over decisions. Previous research had only examined scenarios where automation reliability matched human performance, finding that humans primarily used inhibition (slowing accumulation toward incorrect choices) rather than excitation (directly triggering decisions). This study expanded that framework by manipulating reliability levels to observe shifts in cognitive strategy. The researchers employed a simulated air traffic control conflict detection task with 24 participants. Each participant completed three conditions: a manual baseline (no aid), a low-reliability condition (75% accurate aid), and a high-reliability condition (95% accurate aid). Participants decided whether aircraft pairs would violate minimum separation standards. The study utilized an evidence accumulation model to analyze choices and response times, distinguishing between two integration mechanisms: excitation, where automation inputs directly drive evidence accumulation toward a decision (autonomous triggering), and inhibition, where automation inputs slow accumulation toward opposing decisions (advisory influence). Subjective trust in the automation was also measured after each block. Results indicated that decision aid reliability significantly affected both behavioral performance and underlying cognitive processes. Higher accuracy was observed when the aid was correct, while errors increased when the aid was incorrect, compared to the manual condition. These effects were more pronounced in the high-reliability condition. Crucially, the computational modeling revealed distinct strategies based on reliability. In the low-reliability condition, participants primarily used inhibition, treating the aid as an advisor that influenced but did not trigger decisions. In contrast, in the high-reliability condition, participants exhibited significant excitation, directly accumulating evidence based on the aid’s advice, consistent with granting the automation autonomous influence. Individual differences in the degree of excitation correlated with subjective trust ratings, suggesting a cognitive mechanism linking trust to autonomous automation use. The findings demonstrate that humans adapt their cognitive strategies based on perceived automation reliability. Low-reliability aids are integrated via inhibition, preserving human oversight, while high-reliability aids are integrated via excitation, allowing them to autonomously trigger responses. This shift explains why high-reliability automation is more susceptible to misuse when it fails, as operators defer information processing to the system. The study provides a formal cognitive model for understanding human-automation teaming, highlighting that trust manifests not just as an attitude but as a specific change in information processing architecture. These insights are critical for designing systems that mitigate automation complacency and misuse in high-stakes environments.
Key finding
Participants primarily treated low-reliability decision aids as advisors requiring independent verification, but allowed high-reliability decision aids to autonomously trigger their decisions through direct evidence accumulation.
Methodology
simulation_modeling
Sample size: 24
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. Discovered via author_sweep_intake on 2026-05-28.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 2 | 2026-05-28 |
| archive | success | canonical_url | — | — | 11 | 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 | 2026-05-28 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| 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.