DIVA: Data Intensive Visual Analytics - Provenance and Uncertainty in Human Terrain Analysis

Data Intensive Visual Analytics can help address the data deluge by helping decision makers to rapidly reach informed and effective decisions in a range of situations.

This exploratory project will apply DIVA to defence and security applications in close collaboration with DSTL. It will investigate visual methods for effectively utilising the kinds of dynamic and uncertain data that are emerging from multiple and frequently conflicting sources. Methods will be developed to store, communicate and use metadata about (potentially conflicting, uncertain and messy) data origins, quality and analytical process. They will be transferable and apply at operational and strategic levels.

We draw together a team of UK academics with complimentary expertise. Contributors from Middlesex University and City University London have growing international reputations for developing innovative and applied visual analytics solutions and the theoretical work that supports this activity. Contributors from Loughborough University offer experience of information management and analysis in real time, harsh environments and the military context. The team will work closely to establish and evaluate the potential for DIVA in the area of Human Terrain Analysis.

The programme of work is designed to ensure close engagement between academics and DSTL colleagues. Short bursts of concerted activity focusing around a series of participatory design workshops will result in rapid development and evaluation. These intense periods of coordinated co-located activity will stimulate subsequent reflection and respond to feedback involving DSTL in an iterative process. A continuous bridging presence over a 12 month elapsed period (one researcher working at two sites) will support and consolidate this work. These efforts will address critical research issues faced by the emerging academic VA community:

* How can we best support analysts with information about data uncertainty and provenance?
These factors underlie analytic approaches in data intensive systems yet many issues remain unresolved.

* How can we capture, annotate and explain the analytic process?
Doing so will enable us to reproduce the analytic process and support communication and collaborative analysis.

* How do VA approaches apply in critical applications areas?
Close collaboration with DSTL will ensure that academic developments are grounded in and informed by an applications domain that is vital to national security.

The planned activity will produce schemas, methods and prototypes that address these questions, support analytical work and demonstrate DIVA potential in the military context.

The results are likely to have application impact across MOD and in wider disciplines to which VA is being increasingly applied, including significant data intensive areas in science, industry and government.

Findings will be communicated widely through national and international academic conferences, social media, press releases and at DSTL networking events. Software and functionality developed will be made available through a Creative Commons licence. Along with the knowledge derived through the planned research, this will be used by the UK Visual Analytics Community.

The project offers significant value, using existing skills, equipment and technology, and has low start-up costs. No recruitment is necessary with all participants employed in dynamic and successful research groups at the three participating institutions: City University London (lead), Middlesex University and Loughborough University. The programme of activity involves 24 months of research time over 12 months elapsed time and fits in well with the schedules and workloads of world class researchers operating in the international arena. All are committed to the work plan, which will contribute to institutional objectives in all cases and is supported by the US National Visual Analytics Centre.

This proposal is in direct response to a strategic assessment of the impact of emerging technology on defence and security by the MOD. All activity is designed to have direct impact in terms of demonstrating and evaluating this emerging capability in terms of established needs. Short-term impact is a key objective and will be evaluated during the project.

The proposed work will establish requirements and possibilities for using Visual Analytics in data intensive systems to support efficient and informed decision-making at operational and strategic levels in the context of large, diverse, contradictory and uncertain data sets.

The project is designed using techniques developed for knowledge sharing and participatory design to enable leading academics in the discipline to work rapidly and efficiently with DSTL colleagues in addressing these aims. This gives this exploratory project a high probability of success.

DSTL will benefit by gaining knowledge about visualization and VA possibilities that may be applied to various decision-making scenarios. This will be achieved by focussing on the Human Terrain - an area for which we have some evidence from serving officers that visualization may be helpful - but solutions are likely to be transferable to other scenarios in which data intensive systems are used. Our designs will account for these needs.

The approaches that we develop for establishing, encoding and using data provenance and information about analytics process will be of direct benefit to information owners and analysts in the military (SIO, IMgr, ISO, etc.) and beyond.

The research undertaken to develop VA approaches addresses key open academic research questions and are applicable to the analysis of broad swathes of data that are emerging in science, government, commerce and industry. The team's development of visualization and VA approaches in a range of domains demonstrates potential. These include: transportation management and planning, insurance risk, global climate modelling, local government service provision, environmental protection, voting behaviour, energy consumption, house price analysis and long and short term migration. In each case we have worked to develop analytic methods that provide access to data through interactive graphics that improve understanding, frequently with partners in government, commerce or industry. The methods that we propose developing through this project would have been beneficial in each of these cases.

We consider the work that we will be undertaking in a focussed DSTL case study as influencing the majority of these domains and being likely to have wide and significant impact in VA use cases within them.

Short term impacts are likely to involve awareness in DSTL and amongst project partners with a formalization of uncertainty, analytical process and data provenance. Medium term impacts will transfer this knowledge to the academic community. Longer term impacts will involve the uptake of these methods in other domains and disciplines. Our 'pathways to impact' and dissemination strategy support these objectives.