Share
Predictive analytics is now a core workstream of the Centre. Below, we provide details on where we will focus our efforts so that the humanitarian system can reliably use models to trigger action. Our work is just getting started—join us by contributing to model development or becoming a peer review expert. Read more about elements of the programme in this factsheet.
The digital era is, by now, a reality in the humanitarian sector. In August, the number of datasets available on the Humanitarian Data Exchange (HDX) reached a staggering 10,000, and it keeps growing every day. Data from a variety of sources is becoming increasingly available to humanitarian practitioners. At the same time, algorithms are now part of daily life, following stunning advancements in data science and machine learning.
The influence of tech on humanitarians was apparent in the findings of the data literacy survey conducted by the Centre in January 2019. When asked what topics humanitarians are most interested in learning more about, respondents said big data, predictive analytics, and statistics. This response held true for people working at all levels in the humanitarian sector, from senior leaders to data managers.
Using data science to anticipate humanitarian crises represents a great opportunity for organizations to respond earlier, saving and protecting more lives than ever before. But despite some isolated applications, predictive analytics is still an exploratory space for the sector, especially when it comes to developing models that trigger financing.
Given the Centre’s role and expertise with data, it has been exploring how best to advance the application of predictive analytics in humanitarian response. The Centre’s 2018 and 2019 Data Fellows piloted models (for Somalia and South Sudan) and developed frameworks for model governance. A workshop was held in April 2019 with over a dozen partners to understand the current state of model development and use. Based on this research and increasing demand, the Centre will now include predictive analytics as a core aspect of its work.
In early September, I left my role at the Internal Displacement Monitoring Centre and decided to step outside of my comfort zone to take on the challenge of leading the Centre’s predictive work. As a scientific researcher, this is an exciting opportunity to bring my knowledge and methods in service to the humanitarian community.
The Centre’s predictive analytics focus
1. Modelling
It is far better to foresee even without certainty than not to foresee at all 1
We will develop new models, from design to application, and support existing partner models for use in humanitarian operations. We will collaborate with OCHA offices on models that support more forward-looking Humanitarian Response Plans. We will also be a key partner in the CERF’s anticipatory action pilots, which are intended to determine whether and how the fund, with its established systems and global remit, can best play a role in financing anticipatory action based on agreed thresholds and triggers.
We will leverage the unique position of the Centre to get a good understanding of the data available and of the key questions predictive models should answer, recognising the importance of involving decision makers from the beginning. A model cannot be actionable if the demand is not well understood and discussed within the context of its application.
2. Quality Assurance
All models are wrong, but some are useful 2
Models are a simplification of reality. Therefore their limitations, boundaries, and fields of applicability should be clearly understood and communicated. We will support the creation of standards for the development, validation, and use of predictive analytics. We want to increase the reliability of models and reduce the risk of ineffective actions. We believe that an objective, expert review process may be the best way to address existing scepticism while still seizing the opportunity represented by predictive analytics.
As a first step, the Centre has developed a Peer Review Framework for Predictive Analytics in Humanitarian Response. The initial Framework was developed by Dani Poole during her fellowship with the Centre in June and July 2019. The Framework consists of three steps: 1) readiness assessment; 2) model review; and 3) assessment and results. Models will be reviewed against three criteria: technical, ethical, and humanitarian relevance.
We are releasing the Framework for feedback this week and invite individuals working in academia, the private sector, and humanitarian organizations to take a closer look. In particular, let us know if the process seems feasible, and if the ethical matrix and technical and humanitarian relevance checklists are comprehensive. Also let us know if you want to become an expert reviewer or have your model reviewed as part of a pilot. Note: The Framework was updated in May 2021 and can be found here.
3. Community
Social innovation thrives on collaboration; on doing things with others, rather than just to them or for them 3
The Centre exists to create partnerships and our work on predictive analytics will only reinforce this. We recognise that we are not experts in areas such as climate science or food insecurity and we will therefore need to tap into specific domain expertise from around the world.
We will build capacity and community by convening events, developing case studies, and offering training on predictive modelling. We will work closely with the Centre’s Data Literacy team to ensure that humanitarians understand what predictive analytics can provide, including the limitations and uncertainties.
The Centre is also happy to host staff from partner organizations and academic institutions at its offices in The Hague Humanity Hub. These co-working arrangements are valuable for joint problem-solving and learning across sectors and crises. We will also promote a community of practice where modellers, researchers, and technical experts can share ideas.
“By working together, we can make anticipatory and early action become our default approach in addressing vulnerability and reducing the scale of a possible crisis.”
-Mark Lowcock, Under-Secretary-General for Humanitarian Affairs, OCHA
Next Steps
The Centre has been purposefully ambitious in drafting the strategy for this new workstream. We recognise that real progress will take time and may bring radical change to the sector. Although challenging, the development and integration of predictive analytics into OCHA’s processes, and those of the humanitarian system, has the potential to accelerate critical assistance to people in need before a situation escalates. It is a challenge we must try to meet together.
I hope you share our vision and passion and will engage with us on this topic. For more information, review the Peer Review Framework, visit the Predictive Analytics page on the Centre’s site, or contact our team at centrehumdata@un.org. We look forward to hearing from you!
1 Henri Poincaré, The Foundations of Science: Science and Hypothesis, The Value of Science, Science and Method, 1913 2 https://en.wikipedia.org/wiki/All_models_are_wrong 3 https://www.theguardian.com/lifeandstyle/2011/apr/03/happiness-how-to-find-it
