International Rescue Committee Explores AI to Scale Humanitarian Aid, Raising Opportunities and Concerns

IRC's AI Initiative to Scale Humanitarian Aid

The International Rescue Committee (IRC) is leveraging artificial intelligence to address the growing demand for humanitarian assistance amidst increasing global displacement. With over 117 million people forcibly displaced in 2024 1

1

, the IRC is exploring AI-powered solutions to bridge the gap between needs and resources.

The Signpost Project and AI Integration

Since 2015, the IRC has invested in Signpost, a portfolio of mobile apps and social media channels providing multilingual assistance to people in dangerous situations 2

2

. Having reached 18 million people, the IRC now aims to significantly expand its reach using AI tools.

The organization is testing a network of AI chatbots to enhance the capacity of humanitarian officers and local organizations. This pilot project, currently operational in El Salvador, Kenya, Greece, and Italy, responds in 11 languages 3

3

. It utilizes large language models from tech giants like OpenAI, Anthropic, and Google, along with customer service software from Zendesk.

Potential Benefits and Scalability

Jeannie Annan, IRC's Chief Research and Innovation Officer, emphasizes the potential of AI to scale humanitarian responses:

"We're trying to really be clear about where the legitimate concerns are but lean into the optimism of the opportunities and not also allow the populations we serve to be left behind in solutions that have the potential to scale in a way that human to human or other technology can't." 4

4

The IRC aims to extend this technological infrastructure to other nonprofit humanitarian organizations at no cost, creating shared resources for less technically focused organizations.

Addressing Risks and Ethical Concerns

While AI offers promising solutions, it also presents significant risks:

1. Accuracy and timeliness: Outdated or incorrect information could be dangerous in rapidly changing situations 5

   5

   .
2. Data security: The system could potentially amass sensitive data about vulnerable populations, risking exploitation by malicious actors 5

   5

   .

To mitigate these risks, the IRC has implemented several measures:

• Local vetting of chatbot responses to ensure accuracy and contextual sensitivity.
• Escalation of complex queries to human responders.
• Agreements with tech providers to prevent AI models from training on sensitive data.
• Data anonymization, including removal of personal information and location data 6

  6

  .

Expert Perspectives on Responsible AI Implementation

Cathy Petrozzino from MITRE emphasizes the importance of responsible AI implementation, urging organizations to consider technology effectiveness, fairness, and data privacy [7]. She advocates for diverse input in governing and designing AI initiatives, including technical experts, context specialists, legal experts, and user representatives.

Helen McElhinney, executive director of CDAC Network, stresses the importance of consulting with displaced people and obtaining informed consent for AI-driven services [8]. She argues that while this may increase development time, it is crucial for addressing safety and ethical concerns.

Future Developments and Considerations

As part of the Signpost project, the IRC is also exploring other AI-powered tools, such as digital automated tutors and data-integrating maps for crisis preparation and response [9]. These developments highlight the potential for AI to revolutionize humanitarian aid, while also underscoring the need for careful implementation and ongoing evaluation to ensure ethical and effective deployment in serving vulnerable populations.