Intelligence analysis involves unpredictable processes and decision making about complex domains where analysts rely upon expertise. Artificial Intelligence (AI) systems could support analysts as they perform analysis tasks, to enhance their expertise. However, systems must also be cognisant about how expertise is gained and designed so that this is not impinged. In this paper, we describe the results of Cognitive Task Analysis interviews with 6 experienced intelligence analysts. We capture themes, in terms of their decision making paths during an analysis task, and highlight how each theme is both influenced by expertise and an influence upon expertise. We also identify important interdependencies between themes. We propose that our findings can be used to help design Human-Centered AI (HCAI) systems for supporting intelligence analysts.

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.

Dimensionality Reduction (DR) is a core building block in visualizing multidimensional data. For DR techniques to be useful in exploratory data analysis, they need to be adapted to human needs and domain-specific problems, ideally, interactively, and on-the-fly. Many visual analytics systems have already demonstrated the benefits of tightly integrating DR with interactive visualizations. Nevertheless, a general, structured understanding of this integration is missing. To address this, we systematically studied the visual analytics and visualization literature to investigate how analysts interact with automatic DR techniques. The results reveal seven common interaction scenarios that are amenable to interactive control such as specifying algorithmic constraints, selecting relevant features, or choosing among several DR algorithms. We investigate specific implementations of visual analysis systems integrating DR, and analyze ways that other machine learning methods have been combined with DR. Summarizing the results in a "human in the loop" process model provides a general lens for the evaluation of visual interactive DR systems. We apply the proposed model to study and classify several systems previously described in the literature, and to derive future research opportunities.

Euler diagrams are often used to visualize intersecting data sets in applications such as criminology; genetics, medicine, and computer file systems. One interesting aspect of these diagrams is that some data sets cannot be drawn without breaking one or more "wellformedness properties," which are considered to reduce the user comprehension of the diagrams. However, it is possible to draw the same data with different diagrams, each of which breaks different wellformedness properties. Hence, some properties are "swappable," so motivating the study of which of the alternatives would be best to use. This paper reports on the two empirical studies to determine how wellformedness properties affect comprehension. One study was with abstract data, the other was with concrete data that visualized students' enrollment on university modules. We have results from both studies that imply that diagrams with concurrency or disconnected zones perform less well than other some other properties. Further, we have no results that imply that diagrams with brushing points adversely affect performance. Our data also indicate that nonsimple curves are preferred less than diagrams with other properties. These results will inform both human diagram designers and the developers of automated drawing systems on the best way to visualize data using Euler diagrams.

Euler diagrams have a wide variety of uses, from information visualization to logical reasoning. In all of their application areas, the ability to automatically layout Euler diagrams brings considerable benefits. In this paper, we present a novel approach to Euler diagram generation. We develop certain graphs associated with Euler diagrams in order to allow curves to be added by finding cycles in these graphs. This permits us to build Euler diagrams inductively, adding one curve at a time. Our technique is adaptable, allowing the easy specification, and enforcement, of sets of well-formedness conditions; we present a series of results that identify properties of cycles that correspond to the well-formedness conditions. This improves upon other contributions toward the automated generation of Euler diagrams which implicitly assume some fixed set of well-formedness conditions must hold. In addition, unlike most of these other generation methods, our technique allows any abstract description to be drawn as an Euler diagram. To establish the utility of the approach, a prototype implementation has been developed.

Euler diagrams are effective tools for visualizing set intersections. They have a large number of application areas ranging from statistical data analysis to software engineering. However, the automated generation of Euler diagrams has never been easy: given an abstract description of a required Euler diagram, it is computationally expensive to generate the diagram. Moreover, the generated diagrams represent sets by polygons, sometimes with quite irregular shapes that make the diagrams less comprehensible. In this paper, we address these two issues by developing the theory of piercings, where we define single piercing curves and double piercing curves. We prove that if a diagram can be built inductively by successively adding piercing curves under certain constraints, then it can be drawn with circles, which are more esthetically pleasing than arbitrary polygons. The theory of piercings is developed at the abstract level. In addition, we present a Java implementation that, given an inductively pierced abstract description, generates an Euler diagram consisting only of circles within polynomial time.

Euler diagrams are frequently used for visualizing information about collections of objects and form an important component of various visual languages. Properties possessed by Euler diagrams correlate with their usability, such as whether the diagram has only simple curves or possesses concurrency. Sometimes, every diagram that represents some given information possesses some undesirable properties, and reducing the number of violations of undesirable properties is beneficial. In this paper we show how to count the number of violations from the reduced Euler graph. We then define various transformations on the Euler graph which can reduce the number of violations of a given property, but sometimes at the expense of increasing the number of violations of another property. These transformations can be used to improve the quality of the drawn diagram, which is important for effective information visualization.

Euler diagrams are collections of labelled closed curves. They are often used to represent information about the relationship between sets and, as such, they have numerous applications including: visualizing biological data, diagrammatic logics, and visual database querying. Various methods to automatically generate Euler diagrams have been proposed. Typically, the generation process starts with an abstract description of an Euler diagram, which is then converted to a planar dual graph. Finally, the process attempts to embed the Euler diagram from the dual graph. This paper describes a method for embedding wellformed Euler diagrams from dual graphs. There are several mechanisms to generate dual graphs but, prior to the novel work described here, no general method for embedding a wellformed Euler diagram from a dual graph had been demonstrated. The method in this paper achieves an embedding of any wellformed Euler diagram. The method first triangulates the dual graph. Then, using the faces of the triangulated graph, an edge labelling technique identifies the vertices of polygons which form the closed curves of the Euler diagram. The method is demonstrated by a Java implementation. In addition, this paper discusses a number of layout improvements that can be explored for this embedding method.

Graphical representation may provide effective means of making sense of the complexity and sheer volume of data produced by DNA microarray experiments that monitor the expression patterns of thousands of genes simultaneously. The ability to use ldquoabstractrdquo graphical representation to draw attention to areas of interest, and more in-depth visualizations to answer focused questions, would enable biologists to move from a large amount of data to particular records they are interested in, and therefore, gain deeper insights in understanding the microarray experiment results. This paper starts by providing some background knowledge of microarray experiments, and then, explains how graphical representation can be applied in general to this problem domain, followed by exploring the role of visualization in gene expression data analysis. Having set the problem scene, the paper then examines various multivariate data visualization techniques that have been applied to microarray data analysis. These techniques are critically reviewed so that the strengths and weaknesses of each technique can be tabulated. Finally, several key problem areas as well as possible solutions to them are discussed as being a source for future work.

Clustering is inherently a difficult problem, both with respect to the construction of adequate objective functions as well as to the optimization of the objective functions. In this paper, we suggest an objective function called the Weighted Sum Validity Function (WSVF), which is a weighted sum of the several normalized cluster validity functions. Further, we propose a Hybrid Niching Genetic Algorithm (HNGA), which can be used for the optimization of the WSVF to automatically evolve the proper number of clusters as well as appropriate partitioning of the data set. Within the HNGA, a niching method is developed to preserve both the diversity of the population with respect to the number of clusters encoded in the individuals and the diversity of the subpopulation with the same number of clusters during the search. In addition, we hybridize the niching method with the k-means algorithm. In the experiments, we show the effectiveness of both the HNGA and the WSVF. In comparison with other related genetic clustering algorithms, the HNGA can consistently and efficiently converge to the best known optimum corresponding to the given data in concurrence with the convergence result. The WSVF is found generally able to improve the confidence of clustering solutions and achieve more accurate and robust results.