It has been estimated that over 600,000 Syrians have been killed since the start of the civil war, including tens of thousands of civilians killed in airstrike attacks. Predicting where and when strikes will occur and issuing time-critical warnings enabling civilians to seek safety is an ongoing challenge. It was this problem that motivated the development of Sentry Syria, an early warning system that alerts citizens to a possible airstrike. Sentry uses acoustic sensor data, reports from on-the-ground volunteers, and open media ‘scraping’ to detect warplanes in flight. It uses historical data and AI to validate information from these different data sources and then issues warnings to civilians 5-10 minutes in advance of a strike via social media, TV, radio and sirens. These extra minutes can be the difference between life and death.

Sentry Syria is just one example of an emerging approach in the humanitarian response we call collective crisis intelligence (CCI). CCI methods combine the collective intelligence (CI) of local community actors (e.g. volunteer plane spotters in the case of Sentry) with a wide range of additional data sources, artificial intelligence (AI) and predictive analytics to support crisis management and reduce the devastating impacts of humanitarian emergencies.

The increasing availability of data from a variety of sources, together with advancements in statistics and machine learning, is generating a growing interest in using models to gain insight and trigger anticipatory action. However, leaving the design and deployment of ‘AI for Good’ systems to technologists alone is now recognised as a risk, particularly for individuals and groups who are already extremely vulnerable.

By drawing on novel, localised data sources, including from responders and communities on the frontline, CCI solutions can build a richer local and social understanding of crises. Combining these with the processing power of AI technologies, such as predictive analytics, means humanitarians can have access to more timely and contextual data – which can be used for anticipatory action, effective response or sustainable recovery.

CCI, and the related participatory AI approach, offers the potential to mitigate the risks of applying AI and predictive analytics in a humanitarian setting – through codesign of solutions, the production of less biased data that reflects the experiences of the most affected populations, and community-based assessment of model efficacy.

Our landscape analysis of CCI solutions and rapid review of participatory approaches to the development of AI systems has been summarised in two new reports, ‘Collective crisis intelligence for frontline humanitarian response’ and ‘Participatory AI for humanitarian innovation: a briefing paper’.

The ‘collective crisis intelligence’ (CCI) project, delivered in partnership by Nesta’s Centre for Collective Intelligence Design, the IFRC’s Solferino Academy and the Cameroon and Nepal Red Cross, is one of the first attempts to develop and test new methods for involving crisis-affected populations and frontline responders in the development, evaluation and utilisation of new predictive AI models. It builds on our previous research on the relationship between AI and our collective human intelligence and responds directly to the need for more human-centred AI in the humanitarian sector. It aims to support humanitarian action that is both locally led and more inclusive by involving crisis-affected communities and frontline workers as key stakeholders – to ensure that crisis mitigation, response and recovery is rooted in the insights and experiences of those closest to the crisis.

Over the next year, we will research, develop, prototype and test CCI solutions working in partnership with two of the Red Cross Red Crescent’s National Societies in Nepal and Cameroon. The project comprises three key phrases:

1. Reviewing and analysing existing CCI solutions in the humanitarian sector,
2. Exploring locally tailored CCI concepts through design research,
3. Developing two CCI solutions for rapid prototyping and testing.

Our review of CCI solutions highlights how CCI is a nascent field, with many solutions at an early stage of development, and varying levels of integration into humanitarian workflows and systems. Adoption and scaling of solutions has been piecemeal, hindered by a range of technical, organisational and sectoral challenges such as the paucity of high quality datasets, the (growing) data and digital skills gaps, and a lack of leadership buy-in to new technical solutions entering the sector. By working together with the IFRC Solferino Academy and the Nepal and Cameroon National Societies we hope to learn more about potential pathways to solution integration in crisis response, with a focus on supporting National Societies to develop internal capacity to use these new methods in the years to come.

In our research we found limited information about the overall design process in which CCI solutions were developed. This raises questions on how the design and development of CCI typically happens, for example, how power and resources differences are negotiated in the initiatives and how this impacts the ability of frontline workers to engage local communities in the process. Perhaps the most significant barrier to the development of CCI solutions lies in the difficulty of working actively with affected local populations and sustaining participation as a crisis unfolds. This may be due to the need to prioritise speed of response, difficulties with connectivity or accessing affected locations, and the importance of considering the mental health of affected populations (and their ability to participate in the context of a crisis).

We believe CI methods can help build sustainable, equitable and meaningful data sources that truly reflect the insights and needs of crisis-affected communities. In addition, our exploration of participatory AI demonstrates how the development of these models can ensure that the resulting tools reflect the needs and values of the people and groups using them or impacted by their use. In this project we aim to test and evaluate participatory interventions during the algorithm development process, working with different groups from National Societies, frontline volunteers and where possible, local communities. We will give special attention to sharing reflections on the challenges, how we work to overcome them and the unique value of this particular partnership along the way through regular project updates.

For the IFRC and National Societies this is a valuable opportunity to not only advance the understanding and potential use of accelerated community engagement mechanisms, but also to play a leading role in co-creating the standards for (participatory) AI in the organisation and sector.

The Cameroon Red Cross hopes that further development of collective intelligence methods could advance data collection and processing systems, crisis anticipation and response capabilities and the ability to offer tailor made solutions to communities served by the organisation.

The Nepal Red Cross (NRCS) welcomes the practical role that the CCI pilot can play in helping the NRCS to advance their work on linking disaster response distributions with cash transfers. By participating in the pilot the NRCS hopes to enhance its capacity in the area of research through meaningful engagement and participation with crisis affected communities.

The Nepal and Cameroon National Society are in the midst of the research & design phase, and we’ve started mapping available data sources for training the model. We will be writing regular blog posts along the way to reflect on what we learn as the project unfolds and to share project updates - stay tuned!