Generative AI for Humanitarians

Introduction to Generative AI in Humanitarian Action

The "Generative AI for Humanitarians" report, published in September 2023 by the Digital Humanitarian Network (DH Network) and authored by Nasim Motalebi and Andrej Verity, explores the transformative potential of artificial intelligence in crisis response. This think-brief serves as a critical stepping stone for humanitarian practitioners and leaders aiming to mainstream Generative AI tools within their organizations. By examining technical capabilities and trending strategies, the paper provides a roadmap for the safe, ethical, and responsible adoption of AI to augment human capabilities rather than replace them in the field.

Technical Capabilities and Practical Use Cases

Generative AI represents a paradigm shift from traditional models that classify data to foundation models that create brand-new, unique artifacts like text, images, and video. For humanitarian organizations, these tools offer profound opportunities to enhance internal efficiencies—such as automating administrative work and donor reporting—as well as project-level applications like predicting climate events or analyzing forced migration patterns. The report details how technologies like Large Language Models (LLMs) and synthetic data generation can drastically reduce workloads, allowing staff to focus on substantive humanitarian issues.

Addressing Risks: Bias, Ethics, and Transparency

Despite the potential, the report underscores significant technical shortcomings including algorithmic bias, misinformation (hallucinations), and data privacy risks. Because AI models reflect the data they are trained on, they risk encoding and exacerbating historical social stereotypes or producing non-factual information. The authors emphasize that AI adoption is an accountability and data governance issue, requiring a "human-in-the-loop" approach to ensure outputs align with humanitarian values and legal standards.

The Road to a Humanitarian AI Ecosystem

To navigate this landscape, the report introduces "10 Rules of Thumb" and recommends building a robust Humanitarian AI Ecosystem. This strategic framework focuses on five key pillars: selecting appropriate data and technology, establishing governance and regulations, investing in AI upskilling, implementing continuous evaluation, and fostering multilateral coordination. By utilizing sandbox environments for safe testing and prioritizing transparency, organizations can harness the power of AI to deliver more effective assistance while remaining unwavering in their commitment to humanitarian principles