Forecasting Outcomes with Root Cause Evaluation for Fielded Systems (FORCEFIELDS)

Artificial intelligence/machine learning (AI/ML) developers, V&V specialists, and mission mangers face the daunting challenge of overseeing and guiding the development, deployment, testing, use, and inevitable revision and redesign of hundreds of battlefield ML models. The stakes are particularly high, as actual battlefield conditions, tactics, and environments inevitably differ from those these models were tested on, limiting their effectiveness. Understanding where and when ML models are performant is further complicated by their black-box nature. To empower developers and users, Charles River Analytics proposes the development of AI Forecasting Outcomes with Root Cause Evaluation for Fielded Systems (FORCEFIELDS). This innovative software solution is designed to naturally integrate with AI/ML enterprise pipelines (e.g., the Project Linchpin ecosystem) and specifically to detect and mitigate cross-functional common causes of model failures. By analyzing systemic issues across a wide array of ML models, FORCEFIELDS enables swift, informed decision-making that can address multiple models simultaneously rather than on a piecemeal basis. Charles River, with its extensive DoD R&D experience, including AI/ML pipelines in particular, is uniquely positioned to address this critical challenge. We will deliver a practical, fully functional prototype, supported by a comprehensive technology transition and commercialization plan for both DoD and commercial applications.