AI-Enabled Policy Content Modeling: A Systems Approach

Public policies around safety and security are often involved in the systems design constraints. Gaps in these policies usually manifest after the system is operational. Systems engineers need to be more involved in policy analysis and exploiting emerging technologies to help automate preexisting policy modeling frameworks. This paper introduces an AI-driven approach to identify policy gaps, particularly in ambiguous and vague texts, automating previously published policy content model (PCM). In our automation methodology, we integrate long short-term memory (LSTM), bidirectional LSTM (BiLSTM), and echo state networks (ESN), in processing the subtle and varied nature of policy texts. An attention mechanism further refines the model’s focus, enhancing accuracy in identifying policy gaps. Dropout, early stopping, and word embeddings are employed to mitigate overfitting, ensuring robust model performance across varied policy documents. The results show that ESN, when enhanced with an attention mechanism, has an F1 score of 0.79. This shows that the proposed approach can identify potential policy gaps.