One hundred important questions facing forestry research: an exploratory case study of AI-generated agendas, biases, and future scenarios

This exploratory case study examines the potential and limitations of using a single generative Artificial Intelligence (AI) model to rapidly synthesize research questions in forestry. Using a specific prompting strategy, we compare DeepSeek-generated present-day questions with those projected for the twenty-second century and further examine geographic biases through a developing-country comparison. Our analysis reveals that DeepSeek captures main ecological concepts and generates comprehensive, thematically structured question lists. However, the outputs are inherently shaped by training-data biases, favoring technology-rich, data-intensive domains while underrepresenting developing-region priorities, local knowledge systems, and critical but less-documented issues, such as environmental pollution and forestry education. The future-oriented questions illustrate the model’s capacity for horizon scanning but span a spectrum from plausible trajectories to speculative scenarios. These findings highlight that while a large language model (LLM) can accelerate the generation of candidate research agendas, it cannot substitute for inclusive, context-sensitive expert validation. We conclude that robust research prioritization should combine LLM’s analytical speed with diverse expert perspectives to ensure equitable and comprehensive agenda setting.