ID: 71
Presenting Author: Ridhee Gupta
Status: pending
This study applies LLMs to analyze public engagement data in impact assessments, revealing usability of AI and biases to enhance transparency, inclusivity, and equitable decision-making.
Meaningful public engagement is central to the legitimacy of impact assessments (IA), yet engagement records are often disorganized, diverse in format, and analyzed manually, potentially contributing to incomplete or biased engagement. This study explores how artificial intelligence (AI) can enhance transparency and communication by making sense of the complex data generated through consultation. Using public submissions and summaries from regional wind development assessments in Atlantic Canada, a workflow was developed to structure and analyze publicly available engagement documents. Natural language processing (NLP) models were trained to identify stakeholder groups, categorize themes, and trace how certain concerns gain or lose prominence across the assessment. These results were compared with those generated using large language models (LLMs) on the same datasets to assess the value LLMs bring to engagement analysis. Model outputs were validated by subject-matter experts to evaluate interpretive accuracy and utility. This process highlights engagement dynamics by visualizing major themes and responses across stakeholder and rightsholder groups, revealing potential disparities in treatment. The paper discusses methodological lessons in data preparation, model bias, and interpretability, along with ethical considerations around automating deliberation. Ultimately, it argues that AI tools can serve, not replace, participatory judgment by revealing hidden structures of influence in IA communication and by strengthening the evidence base for more equitable decision-making.
Ridhee Gupta is a MSc student at University of Victoria. Her interests lie at the intersection between impact assessment and data science to facilitate support for better decision-making.
Coauthor 1: Gerald Singh
Coauthor 2: Brandon Haworth
Coauthor 3: Shana Lee Hirsch
Coauthor 4: Gillian Gregory
Coauthor 5: Ross Moser