Artificial Intelligence for Unbiased Acquisition Planning: A Case Study in Strategy and Baseline Development

Abstract

Defense acquisition programs experience persistent cost growth, schedule delays, and performance shortfalls, with research identifying human cognitive biases in early planning as a contributing factor. This capstone project examines whether artificial intelligence (AI) can reduce cognitive bias in the development of acquisition strategies and acquisition program baselines (APB). Using a comparative case study design, this research administered the Joint Common Missile (JCM) program scenario to eight AI models across 240 runs and compared their outputs against 31 human acquisition professionals using statistical analysis and a five-dimension evaluation rubric. Results indicate AI models triggered optimism bias, anchoring, planning fallacy, and confirmation bias at rates equal to or exceeding humans. Ninety-six percent of AI runs selected the single-step strategy ultimately cancelled in 2004, while 77 percent of humans chose incremental approaches matching the program’s successful successor. AI achieved near-zero strategic diversity compared to humans’ 97 percent of maximum entropy. Despite these shortcomings, AI showed potential as a structured analytical baseline generator when properly constrained. This research recommends employing AI as decision support rather than decision maker, designing structured frameworks that force AI to highlight independent estimate variance, and expanding the field of behavioral acquisition to study AI decision-making behavior.