Abstract for: Refining a Systems Thinking Rubric for Serious Games

Effective assessment of systems thinking remains a major challenge in serious game environments, where subjective judgments often obscure how participants frame problems, interpret structure, and reason about actions in complex situations. This work seeks to refine a systems thinking rubric that can generate clear, interpretable, and repeatable assessments of participant reasoning, supporting broader methodological advances in evaluating higher-order thinking within dynamic, serious games settings. We refine a systems thinking rubric through a qualitative, multi‑lens assessment process using written, open-ended responses from a system dynamics‑based serious game. A Multi‑Perspective AI Agent (MAIA) framework provides an explainable analytic platform to identify ambiguous constructs, evaluate interpretability, and examine disciplinary differences in reasoning. This iterative process integrates qualitative review, rubric content analysis, and inter-rater reliability considerations to strengthen conceptual clarity and support future embedded assessment design. Preliminary progress includes development and initial use of nine MAIA agents for multi-perspective critiques of the systems thinking rubric. Early analyses have identified ambiguous constructs, overlapping terminology, and the need for clearer, more precise performance-level descriptions. Initial scoring of sample responses and multi-lens critique have informed iterative revisions that improve rubric interpretability and prepare it for the next phase of inter-rater reliability assessment. This work advances a systems thinking rubric using an AI‑supported approach to provide stronger evidence of content constructs and inter-rater reliability. The serious game offers a consistent environment for eliciting comparable reasoning, enabling iterative rubric refinement. While the primary contribution is the development of a clearer, more interpretable rubric that supports future embedded assessment and evaluation in serious game contexts, MAIA is introduced as an explainable‑AI analytic framework. Explainable‑AI supported rubric critique and analysis