Stewart Kettle and Patricia Lucas | 29 Feb 2024

There's been an earthquake with tens of thousands of people killed and millions displaced. You're managing the humanitarian response: how do you stay in touch with what's working and what's not?

Sluggish information flow is a problem for humanitarian NGOs. People working on the frontline often have great insights from their experiences on the ground, but getting that information to the ears of decision-makers is challenging. Surveys canvas the views and perceptions of large numbers of people quickly, but the information you get back can be pretty thin. Qualitative interviews provide richer detail, but include smaller numbers of people and are too slow for crisis managers with immediate decisions to make.

We think our AI interviewer can fill this gap, achieving scale and reach, with open, flexible, responsive questions. Combined with AI-assisted analysis of interview responses, you can unpick the details that frontline workers have to share at speed.

For UNICEF Türkiye, we interviewed 152 humanitarian workers in 48 hours with our Colectiv AI interviewer. In this case we asked about their views and experiences of distributing hygiene (or wash) kits to people displaced by the 2023 earthquake. In just 48 hours we were able to conduct 152 qualitative interviews in Turkish, with responses from across the earthquake affected regions. The results show that the kits were generally well received, but highlight some important barriers to access. Unlike a survey, they amplify voices from the frontline; in this case calling our attention to the experiences of the scale of need a year after this natural disaster:

"Bizi hala unutmadınız deprem olalı çok oldu ihtiyaçlarımız bitmedi bizi unutursunuz diye çok korkuyoruz...İyiki varsanız" (You still haven't forgotten us, it's been a long time since the earthquake, our needs are not over, we are very afraid that you will forget us...I'm glad you are here)

The interviews weren't perfect, we are a new startup and we are learning. Our interview was temporarily closed because we failed to anticipate how many and how quickly we would gather responses; we reached our initial predefined maximum within 24 hours. About 5% of people said the interview felt unnatural, and we want to keep improving on this. But 99% of people rated the interview very good, good or acceptable, and Colectiv was able to talk and analyse responses received in Turkish, Arabic and English without a hitch.

The potential for this tool is huge: we can collect and summarise nuanced data rapidly, remotely, and at scale to support decision-makers with real-time information in emergency situations.

The full results of our interviews are available here.