Introduction. The integration of artificial intelligence in complex industrial systems presents significant challenges for human-AI collaboration, particularly in safety-critical domains like aerospace manufacturing. As AI systems become increasingly autonomous in decision-making processes, understanding how to effectively design human-AI interaction becomes crucial for optimizing performance. The aerospace industry, facing unprecedented pressure from growing air traffic demands and production backlogs, represents an ideal context for examining these human factors challenges.
Purpose: This study aimed to investigate user perspectives on AI-based decision support systems in aerospace manufacturing planning and scheduling tasks. Specifically, the research examined how different levels of AI assistance, from manual baseline conditions to local and global interactive support. influence operator performance, workload, and acceptance in real-world industrial settings.
Method: Nine expert planners and schedulers participated in a within-subjects experimental design study. Participants completed workforce allocation tasks under three conditions: manual scheduling (baseline), local interactive AI support, and global interactive AI support. Each participant performed two scenarios per condition (six total), with one simple and one difficult allocation task. Performance metrics, perceived cognitive demand (NASA-TLX), efficiency measures, and sentiment were collected. Testing occurred in participants' actual work environment to ensure ecological validity.
Results: The findings revealed complex patterns in human-AI interaction effectiveness. While no significant differences emerged in perceived performance across conditions, cognitive demand showed notable variations, with global AI support demonstrating slightly higher demands than local support. Efficiency measures indicated that manual scheduling required the most time (approx 2000 seconds), while both AI-supported conditions showed improved time efficiency (local: ~800 seconds, global: ~600 seconds).
Conclusions: The study demonstrates that AI integration in aerospace manufacturing requires careful consideration of human factors beyond pure performance metrics. While AI assistance improved task efficiency, the relationship between automation level and operator acceptance is non-linear.