The SWARM Platform

The SWARM Project

In 2016, the Intelligence Advanced Research Projects Activity (IARPA) in the US Office of the Director of National Intelligence (ODNI) sent out a Broad Area Announcement seeking applications for a project called CREATE, which was seeking to develop a system that would fundamentally advance the quality of analytic reasoning in intelligence. The system had to utilise crowdsourcing, structured analytic techniques and be deployed on a cloud platform. Out of many applications, the University of Melbourne’s SWARM Project was selected along with three other ‘performer’ teams. After the first phase of the project, the SWARM Project’s testing demonstrated that our system, in the right conditions, improved the quality of analysis by over one standard deviation. Thus, at the end of the Phase One of CREATE, the SWARM Project was the only performer team to be selected for Phase Two.


Tapping into the crowd

Through this project, the Hunt Lab succeeded in achieving this fundamental advance by developing a cloud-based platform designed around the structured analytic technique ‘Contending Analyses’. The technique uses a group-sourcing model, where emergent teams collaborate online to develop different ‘takes’ on a task, refine them and then vote to select the best. Having analyses contend with each other results in a stronger end product. Biases are avoided, hidden assumptions are identified, and alternatives are incorporated.

Rather than a crowd, SWARM works with ‘teams’ of anywhere from 5 members to 50. The idea is to provide the scaffolding to enable and promote the contending analyses process, while not obstructing an organic development of collaboration or ideas. Within the team, a core group of members emerges that undertakes the substantive analysis. However, the platform is designed to allow team members to contribute without each member being required to do the ‘hard yards’. Not everyone in a team can or should write a complete analysis, and SWARM provides these members with opportunities to contribute that play to their strengths. While a core group of members undertake substantive analysis or report-writing, others review work, leave feedback, post useful resources, synthesise information, contribute to positive group dynamics and edit the final report.



van Gelder, T. J., & de Rozario, R. (2018). Contending analyses: A new model of collaboration for intelligence analysis. Journal of the Australian Institute of Professional Intelligence Officers, 26(3), 19–31.


Learn about other Hunt Lab projects