Real-world datasets often contain inconsistencies that challenge traditional case-based reasoning models. Building upon the result model, a well-established formal representation of case base reasoning in law, we propose a Bayesian reinterpretation that effectively addresses such inconsistencies. Our Bayesian enhancement quantifies the reliability of precedents and encapsulates principled, explainable predictions even in the presence of conflict, representing a meaningful step forward in using these models to design AI agents.