VALCRI - Visual Analytics for sense-making in Criminal Intelligence Analysis - an Integrating Project whose goal is to develop an integrated, multi-function system prototype at TRL-5, validated in a user environment. VALCRI is developing a suite of integrated functions that are intended to facilitate human reasoning and analytic discourse. By being tightly coupled with semi-automated human-mediated semantic knowledge extraction, VALCRI will respond to human analysts in both a proactive and reactive manner, and work with analysts as a human-technology team, responding and anticipating needs as a Joint Cognitive System.Synergy: In VALCRI, we have been careful in the allocation of functions between human users and technology. We have aimed for a design that seeks to leverage machine augmentation of human reasoning, pattern finding and matching, especially in complex and dynamic environments often with no or limited, missing, incomplete, uncertain or ambiguous data. Humans are (still) superior to machines in such contexts.Automation: We are very cautious to avoid creating automation stovepipes that can lead to the well-known problems associated with ‘automation surprise’. This occurs when automation produces unexpected outcomes or fail, causing a loss of awareness and leading to possible loss of control. In automated intelligence analysis systems, this can happen due to algorithmic opacity, or the lack of computational transparency -- i.e. the provision of information in a human-understandable manner that makes visible how the underlying computation produced its recommendations.Machine learning: Using machine learning (ML) techniques, VALCRI will support the search for semantically similar data across the different data sets applied in a variety of use cases for analytic techniques such as Comparative Case Analysis, Associative Search, Maps and timeline analysis, and Dispersions Diagrams.Such tools can then be used to generate and test the logic of assemblies of propositions that facilitate storytelling and the creation of explanations.
Research was organised according to multiple work streams:Research work stream included several PhD students and post-doctoral fellows and researchers in the following areas:
This informed the Systems Development work stream which included industrial companies and universities
A 4-year work package was also dedicated to the development of a anonymised and realistic database of over 1 million crime records from several datasets. The sense-making research used cognitive engineering approaches to understand the nature of thinking and reasoning processes that the human-machine team, using smart technologies such as Machine Learning, would be required to support.
VALCRI website"AI detective analyses police data to learn how to crack cases"; New SCientist; 10/05/17
"Plotting a path through crime data"; EU Research digital magazine