Augmenting Large Language Models with Causal Risk Patterns for AI Deployment Risk Assessment

Large language models (LLMs) achieve impressive benchmark performance yet exhibit systematic gaps in realworld deployment contexts. This is particularly concerning for AI governance applications where LLMs might identify risk categories but fail to trace the causal mechanisms through which harms actually unfold. We propose that structuring documented AI risks as explicit causal patterns-sequential mechanisms, feedback loops, and quantified effects from empirical research can scaffold more reliable LLM reasoning through human-AI collaboration. We test this through a controlled study comparing pattern-augmented LLMs against vanilla LLMs using GPT-4o and Grok 4 to analyse five real estate AI deployment cases. Results show substantial improvements: GPT-4o improved in causal depth with Cohen’s d=1.20, while Grok showed even larger effects (d=1.93) alongside improved evidence grounding (d=2.10). Validation using Gemini 2.5 Pro and Deep Seek as judges confirmed findings with moderate-to-high inter-rater agreement (r=0.45 to 0.68). Qualitative analysis reveals explicit mechanistic transfer, with augmented outputs mapping patterns from domains such as aviation automation and reinforcement learning pricing to novel real estate contexts. These findings suggest that structured causal knowledge can enable a new modality of human-AI collaboration for AI risk assessment.