[Paper Review] Reflections from the Workshop on AI-Assisted Decision Making for Conservation

Lily Xu, Esther Rolf+22 more|arXiv (Cornell University)|Jul 17, 2023

Species Distribution and Climate Change6 citations

TL;DR

This white paper synthesizes insights from a 2022 workshop on AI-assisted conservation decision-making, identifying key research gaps in AI applications for resource allocation, planning, and impact evaluation. It advocates for interdisciplinary collaboration to advance AI methods that address real-world conservation challenges, emphasizing causal inference, uncertainty quantification, and socially and financially sustainable deployment.

ABSTRACT

In this white paper, we synthesize key points made during presentations and discussions from the AI-Assisted Decision Making for Conservation workshop, hosted by the Center for Research on Computation and Society at Harvard University on October 20-21, 2022. We identify key open research questions in resource allocation, planning, and interventions for biodiversity conservation, highlighting conservation challenges that not only require AI solutions, but also require novel methodological advances. In addition to providing a summary of the workshop talks and discussions, we hope this document serves as a call-to-action to orient the expansion of algorithmic decision-making approaches to prioritize real-world conservation challenges, through collaborative efforts of ecologists, conservation decision-makers, and AI researchers.

Motivation & Objective

• Address the growing urgency of biodiversity loss and habitat degradation through scalable, data-driven conservation strategies.
• Identify critical research gaps in applying AI to conservation planning, resource allocation, and impact evaluation under uncertainty.
• Promote collaboration between AI researchers, ecologists, and conservation practitioners to develop methodologically novel, deployable solutions.
• Highlight the need for AI systems that account for social, financial, and ecological trade-offs in real-world conservation contexts.
• Advance methodological innovation in causal inference, uncertainty estimation, and human-AI collaboration for conservation decision-making.

Proposed method

• Synthesize discussions and presentations from the 2022 AI-Assisted Decision Making for Conservation workshop at Harvard’s Center for Research on Computation and Society.
• Categorize conservation challenges into three phases: understanding the world (via Earth observation and AI), acting in the world (via optimization and planning), and evaluating impact (via causal inference).
• Integrate AI techniques such as computer vision, deep learning, and probabilistic modeling for biodiversity monitoring and individual animal re-identification.
• Apply sequential decision-making frameworks and multi-agent planning to optimize ranger patrols, protected area design, and anti-poaching interventions.
• Use causal inference methods—including do-calculus and counterfactual reasoning—to estimate the impact of conservation actions when RCTs are infeasible.
• Emphasize human-in-the-loop systems that combine AI efficiency with human expertise, especially in pattern recognition and verification tasks.

Experimental results

Research questions

• RQ1 How can AI methods be extended beyond basic monitoring to support strategic resource allocation and action prioritization in conservation?
• RQ2 What novel methodological advances are needed to enable robust AI planning under uncertainty, model misspecification, and complex constraints?
• RQ3 How can causal inference techniques be adapted to estimate the real-world impact of conservation interventions when randomized trials are impractical?
• RQ4 What metrics and evaluation frameworks are needed to assess AI systems that balance accuracy, human effort, and deployment cost in conservation contexts?
• RQ5 How can AI systems be designed to communicate uncertainty effectively and build trust among conservation practitioners in out-of-distribution settings?

Key findings

• AI-assisted decision-making can significantly enhance conservation by improving monitoring through computer vision, enabling optimized action planning, and supporting causal impact evaluation.
• A hybrid human-AI system for elephant individual re-identification reduced reliance on manual verification by leveraging AI to rank candidate matches, cutting human effort while maintaining accuracy.
• Causal inference methods are essential for isolating the true effect of conservation actions—such as elephant population impacts on carbon stocks—when experimental controls are unfeasible.
• Uncertainty quantification and effective communication of uncertainty estimates are critical for trust and safe deployment, especially when models operate outside their training distribution.
• Successful AI deployment in conservation requires interdisciplinary collaboration and iterative, staged deployment with continuous assessment to manage social and financial trade-offs.
• Current AI applications in conservation are often limited to pattern recognition; future progress demands methodological innovation in optimization, sequential decision-making, and causal modeling under real-world constraints.