Discovering Complementary Roles: An Account of Human-AI Teamwork Applying NLP to Research Climate Fragility and Cultural Heritage

Natural language processing (NLP) techniques offer new capacities to extract insights from large volumes of unstructured textual data for public policy research. However, realizing this potential requires discovering effective human-AI collaboration models. Through an account of two attempts applying NLP to study climate fragility and cultural heritage preservation in Southeast Asia, this paper shares lessons on navigating human-AI team dynamics and identifying complementary roles. Initial experiments revealed challenges bridging human contextual knowledge with NLP capabilities. By interactively scoping analysis, iteratively evaluating outputs, and openly communicating, a more fluent workflow emerged. The human researcher fulfilled an orienting role grounding NLP in the research context while the AI assistant adopted a responsive part performing tasks required. The outputs were then checked by the human researcher manually, who then raises details of the data not covered by the NLP processing, further proposing directions of investigation. Then the cycle of activity repeats till the human researcher finds the outcome presenting satisfactory theoretical discussions to the literature. Their interdependent contributions produced novel analyses not feasible on author’s own individually. The discoveries led to actionable principles for human-AI coordination, including transparent task negotiation, result-driven refinement of prompts, layered responsibility, feedback exchange, and role fluidity. By delineating emerging complementary roles, this paper provides practical insights into blending AI strengths with human wisdom for impactful policy research using natural language techniques.