By Aldo Benini. July 2018. Available here as a pdf
Summary
This is an invitation for humanitarian data analysts and others – assessment, policy and advocacy specialists, response planners and grant writers – to enhance the reach and quality of scenarios by means of so-called Bayesian belief networks. Belief networks are a powerful technique for structuring scenarios in a qualitative as well as quantitative approach. Modern software, with elegant graphical user interfaces, makes for rapid learning, convenient drafting, effortless calculation and compelling presentation in workshops, reports and Web pages.
In recent years, scenario development in humanitarian analysis has grown. Until now, however, the community has hardly ever tried out belief networks, in contrast to the natural disaster and ecological communities. This note offers a small demonstration. We build a simple belief network using information currently (mid-July 2018) available on a recent violent crisis in Nigeria. We produce and discuss several possible scenarios for the next three months, computing probabilities of two humanitarian outcomes.
Figure 1: Belief network with probability bar charts (segment)
We conclude with reflections on the contributions of belief networks to humanitarian scenario building and elsewhere. While much speaks for this technique, the growth of competence, the uses in workshops and the interpretation of graphs and statistics need to be fostered cautiously, with consideration for the real-world complexity and for the doubts that stakeholders may harbor about quantitative approaches. This note is in its first draft. It needs to be revised, possibly by several authors, in order to connect to progress in humanitarian scenario methodologies, expert judgment and workshop didactics
RD Comment: See also the comment and links provided below by Simon Henderson on his experience (with IOD/PARC) of trialing the use of Bayesian belief networks
Dear Rick
I hope you are well. Just looked at the article by Aldo Benini on your website – interesting application of the tool. I tried and failed to leave a comment. In that, I referred to work I’ve been involved in using BBNs to evaluate complex programmes where subjective ‘beliefs’ are the primary data driving decision-making. I’ve blogged about that at https://simonhendersonresearch.com/blog which you may be interested in.
That also refers to a paper on the IOD PARC website (written while I was with them) on the use of the tool to augment contribution analysis more generally – which may be of interest to you also. http://www.iodparc.com/resource/bayesian_networks_in_evaluation.html
all the best
Simon
Simon Henderson Research
Mob: 07889 393703
Land: 01795 530445
Skype: simon.henderson.shrl
http://www.simonhendersonresearch.com