The adoption of artificial intelligence (AI) systems in environments that involve high risk and high consequence decision-making is severely hampered by critical design issues. These issues include system transparency and brittleness, where transparency relates to (i) the explainability of results and (ii) the ability of a user to inspect and verify system goals and constraints; and brittleness, (iii) the ability of a system to adapt to new user demands. Transparency is a particular concern for criminal intelligence analysis, where there are significant ethical and trust issues that arise when algorithmic and system processes are not adequately understood by a user. This prevents adoption of potentially useful technologies in policing environments. In this article, we present a novel approach to designing a conversational agent (CA) AI system for intelligence analysis that tackles these issues. We discuss the results and implications of three different studies; a Cognitive Task Analysis to understand analyst thinking when retrieving information in an investigation, Emergent Themes Analysis to understand the explanation needs of different system components, and an interactive experiment with a prototype conversational agent. Our prototype conversational agent, named Pan, demonstrates transparency provision and mitigates brittleness by evolving new CA intentions. We encode interactions with the CA with human factors principles for situation recognition and use interactive visual analytics to support analyst reasoning. Our approach enables complex AI systems, such as Pan, to be used in sensitive environments, and our research has broader application than the use case discussed.

Visualising how social networks evolve is important in intelligence analysis in order to detect and monitor issues, such as emerging crime patterns or rapidly growing groups of offenders. It remains an open research question how this type of information should be presented for visual exploration. To get a sense of how users work with different types of visualisations, we evaluate a matrix and a node-link diagram in a controlled thinking aloud study. We describe the sense-making strategies that users adopted during explorative and realistic tasks. Thereby, we focus on the user behaviour in switching between the two visualisations and propose a set of nine strategies. Based on a qualitative and quantitative content analysis we show which visualisation supports which strategy better. We find that the two visualisations clearly support intelligence tasks and that for some tasks the combined use is more advantageous than the use of an individual visualisation.

Criminal Intelligence Analysis often requires a search different from the semantic and keyword based searching to reveal the associations among semantically and operationally connected objects within a crime knowledge base. In this paper we introduce associative search as a search along the networks of association between objects like people, places, other organizations, products, events, services, and so on. We also propose an associative search model based on the 5WH associated concepts of a crime, i.e. WHAT (what has happened), WHO (who was involved in the crime), WHEN (the temporal information of the crime), WHERE (the geo-spatial information of the crime) HOW (the modus-operandi used in committing a crime). We have employed Formal Concept Analysis theory to reveal the associations, highlighting Hot Spots, offender's profile and its associated offenders in a criminal activity.

This research aims to develop design guidelines for systems that support investigators and analysts in the exploration and assembly of evidence and inferences. We focus here on the problem of identifying candidate 'influencers' within a community of practice. To better understand this problem and its related cognitive and interaction needs, we conducted a user study using a system called INVISQUE (INteractive Visual Search and QUery Environment) loaded with content from the ACM Digital Library. INVISQUE supports search and manipulation of results over a freeform infinite 'canvas'. The study focuses on the representations user create and their reasoning process. It also draws on some pre-established theories and frameworks related to sense-making and cognitive work in general, which we apply as a 'theoretical lenses' to consider findings and articulate solutions. Analysing the user-study data in the light of these provides some understanding of how the high-level problem of identifying key players within a domain can translate into lower-level questions and interactions. This, in turn, has informed our understanding of representation and functionality needs at a level of description which abstracts away from the specifics of the problem at hand to the class of problems of interest. We consider the study outcomes from the perspective of implications for design.

This paper aims to address the problems low literacy (LL) users face when searching for information online. The first part of this paper summarizes the problems that LL user's face, and establishes a set of design principles for interfaces suitable for LL users. This is followed by a description of how these design principles are mapped to a novel interface for interactive data retrieval. The interface was realized into a working system and evaluated against a traditional web interface for both high literacy (HL) and LL users. The suitability of the designs was analyzed using performance data, subjective feedback and an observational analysis. The findings from the study suggest that LL users perform better and prefer the proposed designs over a traditional web interface.