In the field of Institutional Analysis (IA), institutions describe the emergent rules that govern human behaviour and interaction. There are two key types: formal rules such as laws and policies, and informal rules such as norms and shared strategies. Evaluating the effectiveness of existing policies and developing new policies often requires a thorough understanding of informal rules. Traditional methods for identifying informal rules are predominantly qualitative, labour-intensive, and prone to bias, typically limited by small sample sizes. This thesis addresses the gap in standardised, quantitative approaches by developing a novel method to extract informal rules and social structures from survey data: Survey Informal Rule Extraction (SIRE).

The SIRE method involves preprocessing survey data, training decision trees to classify behaviours, extracting Institutional Grammar (IG) components, and generating structured Attribute, Deontic, Aim, Condition, Or else (ADICO) statements. The design process involves an iterative approach with three key steps: 1) exploring and comparing existing survey data from an institutional grammar lens, 2) evaluating different types of questions and identifying relationships in the data, and 3) testing various methods to analyse response data and convert questions into ADICO statements. These steps led to a systematic approach to understanding how informal rules shape human behaviour, supporting evidence-based policy design.

The effectiveness of SIRE is validated through three applications to datasets from the European Social Survey (ESS) and SCALAR. The first validation compares SIRE results with literature findings on household climate change adaptation, showing logical and comparable outcomes. The second validation demonstrates the method's utility in policy analysis, examining factors influencing the adoption of structural flood protection measures. The third validation showcased the method's potential as a strategic analytics tool for political parties, identifying conditions that increase the likelihood of voting for specific parties. The results demonstrate that the SIRE method has the potential to provide a clear representation of institutional rules, offering valuable insights for institutional and policy analysts.

While this thesis presents a novel method for extracting informal rules from survey data, limitations include potential generalizability issues due to the focused validation on specific datasets (SCALAR and ESS), AI-related inconsistencies in text processing, and methodological constraints such as the current inability to directly address open-ended questions or identify deontic components. Future work should focus on further validation across diverse contexts, integration with other analytical frameworks, and method automation. Improvements in data encoding, condition scope, and rule extraction could enhance the SIRE method's utility in policy development and behavioural analysis, addressing current limitations and expanding its applicability.

As climate change impacts rapidly escalate, ensuring the just adaptation of vulnerable communities has become a critical challenge. However, procedural justice—a key pillar of climate justice that focuses on the fairness of policies from a systemic standpoint—remains underexplored. Policy practitioners currently lack the tools to assess whether a policy can meaningfully engage affected groups. This research presents a novel approach to evaluating the procedural justice of climate change adaptation policies ex-ante, before their implementation.

While the impact of institutional structures on the adaptive capacity of systems is well-established, the direct relationship between these structures and the procedural justice of adaptive interventions has not been systematically studied. This research addresses this gap by applying the Institutional Analysis and Development (IAD) framework and incorporating Institutional Network Analysis to assess the procedural justice of climate adaptation interventions.

The research also demonstrates how Large Language Models support the automation of policy analysis tasks, and may be used to significantly improve the scale and timeliness of policy analysis research.

In this research we identify 4 values critical to procedurally just climate change adaptation policy institutions – namely Voice, Transparency, Accountability and Correctability. We operationalise these values in terms of institutional networks, and develop an empirical method for their identification. The work features a tool to automate extraction of policy statements from documents and features an annotation accuracy of 0.67.

Using this methodology, an empirical case of climate change policy ­— the expansion of Water and Wastewater Systems in Chennai — is investigated. The analysis shows that the institutional environment is severely lacking in procedural justice elements and that policy statements relevant to procedurally justice lack specificity.