An Interaction Between Mental Integration and Levels of Automation: An Analytic Approach Applying Systems Factorial Technology

Automation has become nearly ubiquitous in modern society, supporting human performance in personal and professional tasks. Recent works have begun to focus on human-automation performance in speeded perceptual-cognitive tasks using workload capacity analysis, a novel mathematical analysis to measure efficiency of a human-automation team. The present dissertation examined how level of automation and demand for mental integration of information influence human-automation efficiency. Participants performed a multi-element probabilistic decision task with or without the assistance of an automated aid where level of automation and task demand were manipulated. Experiment 1 employed a mixed design with Automation (present vs. absent) and Task Demand (low/two-gauge readings vs. high/four-gauge readings) as within-subject factors and LOA (low/level 4 vs high/level 7) as a between-subject factor. Overall, participant did not respond in anticipation of the aid differently than they would if the systems performed independently. Some evidence suggests that participants were competing with the high LOA automated aid. Participants responded more quickly after the onset of the aid in the low LOA condition, reflecting automation dependence. Experiment 2 examined the high LOA condition with increased task difficulty. Only anecdotal evidence for limited capacity performance was found, and there were no differences overall across Task Demand conditions. Taken together, these results suggest that increasing the LOA of automated aids, particularly to functionality that execute a response and inform the human, may not continue to improve human-automation team performance as predicted by some current literature. Additionally, conventional performance measures revealed faster but less accurate performance in the high LOA condition and slower and less accurate performance in the high Task Demand condition. These findings provide different insight than the SFT analyses, highlighting the potential benefit of employing Systems Factorial Technology methods in tandem with conventional analyses to understand human-automation interaction. Future directions and applications are provided.