29th July 2025
Artificial Intelligence (AI) is often perceived as a futuristic technology. This short article by Lucy Hall (Data and Evidence Specialist at the HLA) is a companion piece to the forthcoming report from the Humanitarian Leadership Academy and Data Friendly Space: ‘AI in the humanitarian sector in 2025: mapping current practice and future potential.’
Lucy outlines how AI has evolved over time, and key developments that have influenced and shaped humanitarian action in the modern era.
Ancient Logic and Mechanical Imagination (circa 300BCE – 1600s)
Long before machines could think, humans were already trying to understand how thinking worked. The origins of AI can be traced back to the 4th century where philosophers such as Aristotle laid the foundations for logic and deductive reasoning; asking questions about the nature of thought, learning and ethics that we are still grappling with in today’s society.
Engineers in the 1st century were building programmable machines designed to mimic intelligent behaviour through utilising resources available to them at the time.
These early thought experiments and mechanical experiments laid the foundations for what would eventually become AI as we know it today through developing a system of rules and cause-effect reasoning which is a concept that underpins AI today.
The birth of the AI field (1950s-1980s)
Throughout the 20th century, advances in mathematics, computer science and symbolic logic brought together these philosophical questions into the realm of science and engineering.
In 1950, British founder of modern computing Alan Turing posed the questions “Can computers think?” This laid the foundations for the research and development of AI.
The term ‘Artificial Intelligence’ itself was first documented in 1956, marking the beginning of the field that would grow from theory to applications used around the world today.
Over the next two decades, researchers would develop rule based systems, logic engines and early machine translation tools. Throughout the decades tools such as ELIZA created by Joseph Weizenbaum at the Massachusetts Institute of Technology (MIT) which was a computer returning results mimicking human speech, and SHRDLU which inspired applications in further AI usage, including in humanitarian action for agriculture and public health.
Applied AI in humanitarian work emerges (1980s – 1990s): Early Warning Systems
The mid 1980s marked a turning point for the humanitarian sector as AI theories and foundations began to influence real world humanitarian operations.
The most significant development came in 1985 with the development and launch of FEWSNET (Famine Early Warning Systems Network).
Created by USAID, it was a response to the famines in the Horn of Africa; combining climate data, satellite imagery, market monitoring and expert analysis to predict food insecurity.
This was one of the first AI systems used for crisis prevention and does remain active today.
This laid further groundwork for the next iteration of AI. Research in the next decade diversified into machine learning, which allows computers to improve with experience by integrating data and strengthening algorithms and enhancing decision making.
Humanitarian systems began using rule-based models, early risk classification tools and the development of GIS (Geographic Information Systems) to map and respond to crises.
Classical AI in Humanitarian Systems (2000s – 2010s)
In the early 21st century, the rise of more powerful computing and more available and consistent data paved the way for increasingly sophisticated tools:
- IPC (Integrated Food Security Phase Classification) was formalised in the mid 2000s as a common language for analysing food insecurity
- DEEP (Data Entry and Exploration Platform) (2017 – 2025) was launched to help structure qualitative data from humanitarian assessments using AI informed classification and collaborative analysis
- GIS (Geographical Information Systems) became more advanced and more frequently used in mapping disaster responses, displacement tracking and agriculture analysis.
Other tools like the INFORM Severity and INFORM Risk tools use AI-like logic and are widely used in humanitarian planning, coordination and decision making practices.
Generative AI and the Humanitarian AI Surge (2020s)
Generative AI (GenAI) arrived with significant global interest, particularly with the launch of ChatGPT in 2022.
While both classical and GenAI rely on existing data to produce an output, the primary innovation of GenAI is how it presents the output.
Classic AI tools produces content in the format of structured outputs (such as risk scores, classifications or forecasts) that need to be interpreted to make sense of by humans. GenAI produces content in natural language. This enables more accessible content, producing both challenges and opportunities.
The road ahead: from tools to trust
The next era of humanitarian AI will be shaped not only by algorithms, but decisions about ethics, local leadership, data sovereignty and the human experience.
The evolution of AI and its application in humanitarian work reflects technological progress and application of this progress for global societal impact.
Early questions asked:“What is intelligence?”
More modern questions have asked “Can machines think” and “Can we predict crises before they happen?”.
GenAI is now asking: “Can machines help us plan, communicate and act faster?”
All of these questions have relied on human intelligence to provide the answers, so the question looking to the future is: “Can AI serve the world in ways which are trustworthy, inclusive, and safe?”
The next era of humanitarian AI will be shaped not only by algorithms, but decisions about ethics, local leadership, data sovereignty and the human experience.
If the past teaches us anything its this: the most impactful advancements in humanitarian AI isn’t just build with powerful technology, but with purpose and people at the centre of the journey.