Crowdsourcing Linchpin Variables for Future Robustness

Imagining the future is an ongoing task that provides necessary perspective for strategic decisions and policy. Our project takes a new approach to this work.

In this project, the Hunt Lab is using a method called ‘back-casting’ to discover ways in which unlikely scenarios could occur in the future through a series of plausible events. Teams compete on a series of exercises in which team members collaborate on our online platform, SWARM, to research and think through the necessary ‘linchpin’ variables that would have to exist to generate a given future scenario. This method is used to plan, prepare for and mitigate possible eventualities that would not otherwise be forecast.

Our approach originates in our research and experience in eliciting complex analytical reasoning across industry. In particular, we apply capabilities and knowledge obtained in two key activities:

In the first relevant activity, two of our team members led the research and development for extreme operational risk scenario methodology at a major bank with operational exposure in 32 countries (de Rozario, 2015). The work covered strategic level operational risks such as major cybercrimes, physical threats to international operations, class actions, and natural disasters. The methodology aimed at “black swan” type exposures with very low probabilities (1 in 1000) and very high impacts (hundreds of millions of dollars). Crucial to the R&D task was to join qualitative scenario approaches to quantitative modelling, where the methodology and analysis had to pass scrutiny of detailed external regulatory review. Salient to our proposal is that the particular scenario method, including analysis of critical factors, was able to identify key variables that enabled more operationally focused modelling of the risk space.

In the second relevant activity, our team designed and implemented a platform for crowdsourcing intelligence analysis (Van Gelder and de Rozario, 2018; van Gelder and de Rozario, 2017). On the platform, crowds form into teams of 15 – 30 participants and produce reports on intelligence-type problems, typically over a period of a week. The methodology manifested on the platform promotes cognitive diversity through a process of contending analyses and analytical latitude. A major experiment with 24 teams from a mix of organisational intelligence functions, as well as professional analyst teams and general public teams recruited online, showed a significant improvement over results produced with “normal methods”, which typically involve individually led analysis. Moreover, teams recruited online showed a surprising strength of analytical capability, leading to the salient insight that reasoning about complex and analytically contentious problems can be crowdsourced.

In the current project, we join elements of the previous research: scenario analysis and crowdsourcing. We identify a small number of future situations of interest and crowdsource plausible scenarios and analysis of key variables that would lead to such situations. These key variables form a dimensional space that frames future scenarios and enables the identification of indicators to monitor.



de Rozario, R., 2015. Scenario Analytics – Presentation at the Centre of Excellence for Biosecurity Risk Analysis.

van Gelder, T., de Rozario, R., 2018. Contending analyses: A new model of collaboration for intelligence analysis. J. Aust. Inst. Prof. Intell. Off. 26, 19.

van Gelder, T., de Rozario, R., 2017. Pursuing Fundamental Advances in Human Reasoning, in: Artificial General Intelligence. Springer International Publishing, pp. 259–262.


Other Hunt Lab Projects