Background:

Signpost, winner of the 2023 UN Sustainable Development Goal Action Award, is the leading digital information platform in the aid sector and governs an essential repository of humanitarian data. In 2024 we launched signpostai, an open consortium to build AI prototypes and solve complex humanitarian problems.

The first test case being researched by signpostai is the use of generative AI technology to help Signpost teams more efficiently communicate with crisis-affected communities at scale. To ensure that any deployment of AI follows core do-no-harm principles, extensive technical and programmatic AI de-risking is required.

Role overview/ summary

Protection Officer, signpostai (TTA) will work in conjunction with the Signpost technology and program teams to test applications for generative AI chat within a controlled testing environment. S/he role will create simulated client conversations to find flaws in the AI’s design before it is deployed to programs.

The TTA will attempt to exploit flaws in the generative AI bots developed under signpostai, document these flaws, and offer possible fixes so that the technology development team can identify the weakness of the tool and patch issues. S/he will also review, score, and provide feedback to outputs of the generative AI tools to inform possible programmatic applications and to ensure its safety.

Reporting: MIP: Senior Advisor for Responsive Information and Project Manager, AI

Responsibilities

• Create chat scenarios and mock client messages.
• Review and grade the quality of AI-generated responses (an approximate average of 100 a week) using a rubric based on established best practices for quality responses.
• Contribute substantive feedback on AI-generated responses that can be used to further train the AI.
• Keep step by step documentation of process, trends, and other notes relevant for dissemination.
• Analyze and document trends in bot responses and work with development team to suggest new prompts, training material, or rules for chatbot.
• Provide perspective on the readiness of chatbot for program-facing use.
• Document all details relating to test environment.

• High level of attention to detail.
• Strong computer proficiency and ability to quickly pick up new technology tools.
• English required. Strong preference for one of the following: Arabic, French, or Spanish.