Jeroen Keppens

Knowledge based crime scenario modelling

A crucial concern in the evaluation of evidence related to a major crime is the formulation of sufficient alternative plausible scenarios that can explain the available evidence. However, software aimed at assisting human crime investigators by automatically constructing crime scenarios from evidence is difficult to develop because of the almost infinite variation of plausible crime scenarios. This paper introduces a novel knowledge driven methodology for crime scenario construction and it presents a decision support system based on it. The approach works by storing the component events of the scenarios instead of entire scenarios and by providing an algorithm that can instantiate and compose these component events into useful scenarios. The scenario composition approach is highly adaptable to unanticipated cases because it allows component events to match the case under investigation in many different ways. Given a description of the available evidence, it generates a network of plausible scenarios that can then be analysed to devise effective evidence collection strategies. The applicability of the ideas presented here are demonstrated by means of a realistic example and prototype decision support software.

On compositional modelling

Many solutions to AI problems require the task to be represented in one of a multitude of rigorous mathematical formalisms. The construction of such mathematical models forms a difficult problem which is often left to the user of the problem-solver. This void between problem-solvers and their problems is studied by the eclectic field of automated modelling. Within this field, compositional modelling, a knowledge-based methodology for system-modelling, has established itself as a leading approach. In general, a compositional modeller organises knowledge in a structure of composable fragments that relate to particular system components or processes. Its embedded inference mechanism chooses the appropriate fragments with respect to a given problem, instantiates and assembles them into a consistent system model. Many different types of compositional modeller exist, however, with significant differences in their knowledge representation and approach to inference. This paper examines compositional modelling. It presents a general framework for building and analysing compositional modellers. Based on this framework, a number of influential compositional modellers are examined and compared. The paper also identifies the strengths and weaknesses of compositional modelling and discusses some typical applications.

A model based reasoning approach for generating plausible crime scenarios from evidence

Robust decision support systems (DSSs) for crime investigation are difficult to construct because of the almost infinite variation of plausible crime scenarios. Thus existing approaches avoid any explicit reasoning about crime scenarios. They focus on problems such as intelligence analysis and profiling. This paper introduces a novel model based reasoning technique that enables DSSs to automatically construct representations of crime scenarios. It achieves this by storing the component events of the scenarios instead of entire scenarios and by providing an algorithm that can instantiate and compose these component events into useful scenarios. This approach is more adaptable to unanticipated cases than one that represents scenarios explicitly because it allows component events to match the case under investigation in many different ways. The approach presented herein is applied to and illustrated with examples from an application of the differentiation between homicidal, suicidal, accidental and natural death.

Concept map assessment for teaching computer programming

A key challenge of effective teaching is assessing and monitoring the extent to which students have assimilated the material they were taught. Concept mapping is a methodology designed to model what students have learned. In effect, it seeks to produce graphical representations (called concept maps) of the concepts that are important to a given domain and how they are related, according to the students. In recent decades various methods have emerged to evaluate concept maps, each measuring different features of concept maps. New approaches are still being developed. Few guidelines are available regarding the method to choose for a given application. This paper is a literature review that consists of two parts. The first is a review of the many automated and manual techniques of concept map analysis. The second is a critical and reflective commentary on these techniques.

Probabilistic abductive computation of evidence collection strategies in crime investigation

This paper presents a methodology for integrating two approaches to building decision support systems (DSS) for crime investigation: symbolic crime scenario abduction [16] and Bayesian forensic evidence evaluation [5]. This is achieved by means of a novel compositional modelling technique that allows for automatically generating a space of models describing plausible crime scenarios from given evidence and formally represented domain knowledge. The main benefit of this integration is that the resulting DSS is capable to formulate effective evidence collection strategies useful for differentiating competing crime scenarios. A running example is used to demonstrate the theoretical developments.

Towards qualitative approaches to Bayesian evidential reasoning

A crucial aspect of evidential reasoning in crime investigation involves comparing the support that evidence provides for alternative hypotheses. Recent work in forensic statistics has shown how Bayesian Networks (BNs) can be employed for this purpose. However, the specification of BNs requires conditional probability tables describing the uncertain processes under evaluation. When these processes are poorly understood, it is necessary to rely on subjective probabilities provided by experts. Accurate probabilities of this type are normally hard to acquire from experts. Recent work in qualitative reasoning has developed methods to perform probabilistic reasoning using coarser representations. However, the latter types of approaches are too imprecise to compare the likelihood of alternative hypotheses. This paper examines this shortcoming of the qualitative approaches when applied to the aforementioned problem, and identifies and integrates techniques to refine them.

Argument diagram extraction from evidential Bayesian networks

Bayesian networks (BN) and argumentation diagrams (AD) are two predominant approaches to legal evidential reasoning, that are often treated as alternatives to one another. This paper argues that they are, instead, complimentary and proposes the beginnings of a method to employ them in such a manner. The Bayesian approach tends to be used as a means to analyse the findings of forensic scientists. As such, it constitutes a means to perform evidential reasoning. The design of Bayesian networks that accurately and comprehensively represent the relationships between investigative hypotheses and evidence remains difficult and sometimes contentious, however. Argumentation diagrams are representations of reasoning, and are used as a means to scrutinise reasoning (among other applications). In evidential reasoning, they tend to be used to represent and scrutinise the way humans reason about evidence. This paper examines how argumentation diagrams can be used to scrutinise Bayesian evidential reasoning by developing a method to extract argument diagrams from BN.

On extracting arguments from Bayesian network representations of evidential reasoning

Bayesian networks are a predominant approach to analyse the findings of forensic scientists. In part, this is due to the way the Bayesian approach fits the scientific method employed in forensic practice. The design of Bayesian networks that accurately and comprehensively represent the relationships between investigative hypotheses and evidence remains difficult and sometimes contentious, however. Recent research has shown that argumentation can inform the construction of Bayesian networks. But argumentation is a distinct approach to evidential reasoning with its on representation formalisms. This issue could be alleviated if it were easy to represent Bayesian networks as argumentation diagrams. This position paper presents an investigation into the similarities, differences and synergies between Bayesian networks and argumentation diagrams and shows a first version of an algorithm to extract argumentation diagrams from Bayesian networks.

Compositional Bayesian modelling for computation of evidence collection strategies

As forensic science and forensic statistics become increasingly sophisticated, and judges and juries demand more timely delivery of more convincing scientific evidence, crime investigation is becoming progressively more challenging. In particular, this development requires more effective and efficient evidence collection strategies, which are likely to produce the most conclusive information with limited available resources. Evidence collection is a difficult task, however, because it necessitates consideration of: a wide range of plausible crime scenarios, the evidence that may be produced under these hypothetical scenarios, and the investigative techniques that can recover and interpret the plausible pieces of evidence. A knowledge based system (KBS) can help crime investigators by retrieving and reasoning with such knowledge, provided that the KBS is sufficiently versatile to infer and analyse a wide range of plausible scenarios. This paper presents such a KBS. It employs a novel compositional modelling technique that is integrated into a Bayesian model based diagnostic system. These theoretical developments are illustrated by a realistic example of serious crime investigation.

Compositional model repositories via dynamic constraint satisfaction with order-of-magnitude preferences

The predominant knowledge-based approach to automated model construction, compositional modelling, employs a set of models of particular functional components. Its inference mechanism takes a scenario describing the constituent interacting components of a system and translates it into a useful mathematical model. This paper presents a novel compositional modelling approach aimed at building model repositories. It furthers the field in two respects. Firstly, it expands the application domain of compositional modelling to systems that can not be easily described in terms of interacting functional components, such as ecological systems. Secondly, it enables the incorporation of user preferences into the model selection process. These features are achieved by casting the compositional modelling problem as an activity-based dynamic preference constraint satisfaction problem, where the dynamic constraints describe the restrictions imposed over the composition of partial models and the preferences correspond to those of the user of the automated modeller. In addition, the preference levels are represented through the use of symbolic values that differ in orders of magnitude.