Over the past decade, we have been engaged in an extensive research effort to build virtual humans and applications that use them.  Building a virtual human might be considered the quintessential AI problem, because it brings together many of the key features, such as autonomy, natural communication, sophisticated reasoning and behavior, that distinguish AI systems.  This paper describes major virtual human systems we have built and important lessons we have learned along the way.

We develop a novel advanced Particle Markov chain Monte Carlo algorithm that is capable of sampling from the posterior distribution of non-linear state space models for both the unobserved latent states and the unknown model parameters. We apply this novel methodology to five population growth models, including models with strong and weak Allee effects, and test if it can efficiently sample from the complex likelihood surface that is often associated with these models. Utilising real and also synthetically generated data sets we examine the extent to which observation noise and process error may frustrate efforts to choose between these models. Our novel algorithm involves an Adaptive Metropolis proposal combined with an SIR Particle MCMC algorithm (AdPMCMC). We show that the AdPMCMC algorithm samples complex, high-dimensional spaces efficiently, and is therefore superior to standard Gibbs or Metropolis Hastings algorithms that are known to converge very slowly when applied to the non-linear state space ecological models considered in this paper. Additionally, we show how the AdPMCMC algorithm can be used to recursively estimate the Bayesian Cram\'er-Rao Lower Bound of Tichavsk\'y (1998). We derive expressions for these Cram\'er-Rao Bounds and estimate them for the models considered. Our results demonstrate a number of important features of common population growth models, most notably their multi-modal posterior surfaces and dependence between the static and dynamic parameters. We conclude by sampling from the posterior distribution of each of the models, and use Bayes factors to highlight how observation noise significantly diminishes our ability to select among some of the models, particularly those that are designed to reproduce an Allee effect.

It is hypothesized that creativity arises from the self-mending capacity of an internal model of the world, or worldview. The uniquely honed worldview of a creative individual results in a distinctive style that is recognizable within and across domains. It is further hypothesized that creativity is domaingeneral in the sense that there exist multiple avenues by which the distinctiveness of one's worldview can be expressed. These hypotheses were tested using art students and creative writing students. Art students guessed significantly above chance both which painting was done by which of five famous artists, and which artwork was done by which of their peers. Similarly, creative writing students guessed significantly above chance both which passage was written by which of five famous writers, and which passage was written by which of their peers. These findings support the hypothesis that creative style is recognizable. Moreover, creative writing students guessed significantly above chance which of their peers produced particular works of art, supporting the hypothesis that creative style is recognizable not just within but across domains.

The Johnson-Lindenstrauss Lemma allows for the projection of $n$ points in $p-$dimensional Euclidean space onto a $k-$dimensional Euclidean space, with $k \ge \frac{24\ln \emph{n}}{3\epsilon^2-2\epsilon^3}$, so that the pairwise distances are preserved within a factor of $1\pm\epsilon$. Here, working directly with the distributions of the random distances rather than resorting to the moment generating function technique, an improvement on the lower bound for $k$ is obtained. The additional reduction in dimension when compared to bounds found in the literature, is at least $13\%$, and, in some cases, up to $30\%$ additional reduction is achieved. Using the moment generating function technique, we further provide a lower bound for $k$ using pairwise $L_2$ distances in the space of points to be projected and pairwise $L_1$ distances in the space of the projected points. Comparison with the results obtained in the literature shows that the bound presented here provides an additional $36-40\%$ reduction.

We mark up a corpus of LaTeX lecture notes semantically and expose them as Linked Data in XHTML+MathML+RDFa. Our application makes the resulting documents interactively browsable for students. Our ontology helps to answer queries from students and lecturers, and paves the path towards an integration of our corpus with external sites.

This thesis investigates the use of problem-specific knowledge to enhance a genetic algorithm approach to multiple-choice optimisation problems.It shows that such information can significantly enhance performance, but that the choice of information and the way it is included are important factors for success.Two multiple-choice problems are considered.The first is constructing a feasible nurse roster that considers as many requests as possible.In the second problem, shops are allocated to locations in a mall subject to constraints and maximising the overall income.Genetic algorithms are chosen for their well-known robustness and ability to solve large and complex discrete optimisation problems.However, a survey of the literature reveals room for further research into generic ways to include constraints into a genetic algorithm framework.Hence, the main theme of this work is to balance feasibility and cost of solutions.In particular, co-operative co-evolution with hierarchical sub-populations, problem structure exploiting repair schemes and indirect genetic algorithms with self-adjusting decoder functions are identified as promising approaches.The research starts by applying standard genetic algorithms to the problems and explaining the failure of such approaches due to epistasis.To overcome this, problem-specific information is added in a variety of ways, some of which are designed to increase the number of feasible solutions found whilst others are intended to improve the quality of such solutions.As well as a theoretical discussion as to the underlying reasons for using each operator,extensive computational experiments are carried out on a variety of data.These show that the indirect approach relies less on problem structure and hence is easier to implement and superior in solution quality.

The CausalityWorkbench project provides an environment to test causal discovery algorithms. Via a web portal, we provide a number of resources, including a repository of datasets, models, and software packages, and a virtual laboratory allowing users to benchmark causal discovery algorithms by performing virtual experiments to study artificial causal systems. We regularly organize competitions. In this paper, we explore the opportunities offered by development applications.

There is little doubt that the use of robots in introductory classes is an effective way to spark an initial interest in Computer Science and recruit students into our classes, and subsequently recruit some of them as Computer Science majors. But when the semester is over, the vast majority of our students are unlikely to see robots in the classroom again until they take advanced courses in AI or Robotics. It is time for those of us who are proponents of the use of robots in Introductory Computer Science to start thinking seriously about how we are using robots in our classes, and how to sustain the interest and enthusiasm of our students as they move on to more traditional courses. While the focus of this paper is on the use of robots in Introductory Computer Science courses, my goal is to initiate a more general discussion on the use of any sort of cool new technology (tangible or not) into both undergraduate and K-12 education. These technologies successfully attract students to study subjects that we ourselves are deeply engaged in. But we need to discuss as a community what happens when our individual classes conclude and the rest of their studies commence.

This paper discusses the development of a sequence of undergraduate courses forming the core curriculum in the Robotics Engineering (RBE) B.S. program at Worcester Polytechnic Institute (WPI). The laboratory robotics kit developed for the junior-level courses is presented in detail. The platform is designed to be modular and cost-effective and it is suitable for laboratory based robotics education. The system is ideal not only for undergraduate coursework but also may be adapted for graduate and undergraduate research as well as for exposing K-12 students to STEM.

Medical experts routinely need to identify the shapes of anatomical structures, and surgeons report that they depend substantially on touch to help them with this process. In this paper, we discuss possible reasons why touch may be especially important for anatomical shape recognition in surgery, and why in this domain haptic cues may be at least as informative about shape as visual cues. We go on to discuss modern surgical methods, in which these haptic cues are substantially diminished. We conclude that a potential future challenge is to find ways to reinstate these important cues and to help surgeons recognize shapes in the restricted sensory conditions of minimally invasive surgery.