Digital Humanitarianism and Refugee Algorithms
Data modelling and simulation technologies are being to reimagine digital humanitarianism. Here, an interactive agent-based modelling simulation shows actors moving from across a specific border location in Myanmar. Origin to Destination on the right. Image courtesy of Luke Demarest, Emergent Migrations.
Digital technologies, in particular those which use algorithms and simulations to take on complex computation tasks, are increasingly ubiquitous in contemporary life. PPA has examined how humanitarian practice is being impacted by this new ‘digital turn’, and how the creative use of these new technologies might even guide policy-making and the practice of decision-making based on policy.
The use of algorithmic decision-making (ADM) in refugee resettlement is a relatively new area, and the management of migration, including forced migration, has thus far not been the subject of sustained scholarly treatment. Given the complexity of the resettlement landscape, a deeper understanding of what different types of ‘digital humanitarianism’ actually do – and crucially, what they cannot do or cannot tell us – is important to guide the broader discussion of how humanitarian practice engages with a wide array of increasingly complex digital technology. This technology brings with it both new and old challenges and risks to humanitarian interventions.
Image courtesy of Luke Demarest, Emergent Migrations.
On the one hand, digital technology is supposed to be emancipatory, freeing up vast resources through efficiency increases, enabling new forms of accountability and transparency, and sensitively taking into account complex requirements beyond non-digital ‘manual’ approaches (DH Network 2020). On the other, the digital is often seen as a disturbing enabler of increasing restrictions and control (Duffield 2016, Ajana 2019). A widespread body of research examines the risks and the pitfalls of this ‘tech turn’ in humanitarianism. However there remains a lack of widespread understanding and literature within humanitarianism related to the question of how these digital interventions and so-called ‘solutions’ actually work.
PPA has examined how the solutions work in practice, with regard to refugee resettlement, and has prototyped new ways of using worldbuilding to explore the possibilities for using algorithmic decision-making for fairer resettlement outcomes. Emerging out of this we have been established initial guidelines with which to approach this problem of how to tackle the challenges of explainable AI, of algorithms being seen as a black box and of inherent biases being built into systems. In light of these legitimate concerns and as a result of our research, PPA’s design principles show the way to approach the design of decision-making systems in this area:
Visualise the factors that combine to give different outcomes. It’s not difficult to visualise the models and logic that underpin these systems, and the indicators, factors and weightings that are combined to give different outcomes.
Transparent agendas: If you have a policy, make it visible. Make easy for people to see the effect of different factors - and toggle them off if needed, especially if there is a possibility of bias that needs to be investigated further. When designing and using these systems, it is best to admit that is always an agenda of some kind, and that there will always be trade-offs to be made.
People are not just data points: Explain their narrative. A decision-making system should be capable of revealing the narrative arc of the future lives of people impacted by decisions - they should not be restricted to dealing with numbers and a purely quantitative perspective, although these of course do have a significant role to play for an analyst. In addition, it is important to show how things are expected to change over time - and what contributes to these outcomes. (These design principles were presented at the CHI 2022 workshop, ‘AI Generated Characters: Deep Fakes for Good’).
Image courtesy of Luke Demarest, Emergent Migrations research project. Please see our Resources section for more information and other work in the field.