The main theme that underlies this research project is automatedreasoning, an applied sub-discipline of mathematical logic. Logichas found applications in many areas of computer sciencesuch as the verification of digital circuits, reasoning aboutprograms and knowledge representation. One of the most fundamentalaspects in this context is to automatically decide whether aparticular formula is a logical consequence of a given set ofassumptions. The set of assumptions may describe complex relationsbetween diseases and their symptoms, and one possible reasoning taskwould be to confirm or reject a diagnosis based on observed symptomsand medical history.In this research project, we investigate applications ofmathematical logic in knowledge representation. One of the primechallenges in this area is to design logical formalisms that strikea balance between the two conflicting goals of expressiveness (theability to formally represent the application domain) andcomputational tractability. The family of modal logics, conceived ina broad way, combines both aspects and serves as the mathematicalfoundation of a large number of knowledge representation formalisms.The core ingredient of modal logic is the possibility to qualifylogical assertions to hold in a certain way. Depending on thecontext, we may for instance stipulate that assertion holds `alwaysin the future', `with a likelihood of at least 50%' or `normally'.Together with names for individual entities, this allows us toformulate assertions like `the likelihood of congestion on Queen'sRoad is greater than 30%', and complex knowledge bases arise bycombining different logical primitives. Automated reasoning thenallows us to mechanically verify e.g. the consistency of scientifichypotheses against an existing knowledge base. Our goal is to builda modular and practical knowledge representation system that allowsto represent and reason about knowledge represented in this way,based on a large and diverse class of logical primitives, includinge.g. the coalitional behaviour of agents, quantitative uncertainty,counterfactual reasoning and default assumptions. This goes waybeyond the current state of the art, where only logical primitiveswith a relational interpretation are supported by automated tools.Recent research has shown these new logical features can beaccounted for in a uniform way by passing to a more generalmathematical model, known as `coalgebraic semantics'. This richerframework does not only provide a uniform umbrella for a largenumber of reasoning principles, but also supports a richmathematical theory that has by now matured to the extent which putsthe development of automated tools within reach. The researchchallenge that this proposal addresses is the further development of thesetheoretical results as to bring them to bear on practical applications.As a concrete case study, we will use the Cool system to formalisequantitative models in Systems Biology.

Most modern cryptographic constructions are accompanied by a proof of security, in which the difficulty of violating the security of the construction (e.g., distinguishing ciphertexts for an encryption scheme) is reduced to the difficulty of solving a certain algebraic problem. Cryptographic proofs of security - also called reductions - thus lie at the heart of provable security, yet writing and verifying cryptographic reductions is currently a time-intensive and manual process, with most reductions highly individualised for a specific primitive or algebraic setting. By identifying proof techniques common to many settings, the landscape of both reductions and the hardness assumptions that constructions rely on for security can be vastly simplified. In a previous project, we demonstrated that certain proof techniques could also be applied outside of the settings for which they were originally intended, and moreover could be applied to show the equivalence of certain ad-hoc assumptions and more well-established assumptions. Thus, rather than avoid ad-hoc assumptions by providing new constructions or writing new reductions, we demonstrated that the security of a variety of existing constructions - which had relied previously on these ad-hoc assumptions for security - could now be considered secure under a milder assumption. In this work, we will formalise techniques that are common across different proofs in a fashion that makes them easier to reuse, verify, and apply to new settings. This will not only make reductions easier to both write and understand, but also expand the applicability of useful proof techniques.

We propose to study privacy management by investigating how individuals learn and benefit from their membership of social or functional groups, and how such learning can be automated and incorporated into modern mobile and ubiquitous technologies that increasingly pervade society. We will focus on the privacy concerns of individuals in the context of their use of pervasive technologies, such as Smartphones and personal sensors which share data in the Cloud. We aim to contribute to research in three areas: (1) software engineering of adaptive systems that guide their users to manage their privacy; (2) development of machine learning techniques to alleviate the cognitive and physical load of eliciting and personalising users' privacy requirements; and (3) empirical investigation of the privacy behaviour of, and in, groups, in the context of both collaboration and conflict. The ability to control and maintain privacy is central to the preservation of identity. In recent years, social psychologists have made a core distinction between personal identity (which refers to what makes us unique, as individuals, compared to other individuals) and social identity (which refers to our sense of ourselves as members of a social group and the meaning that group has for us). In the latter case, our sense of who we are can be derived from our membership of social groups. Identity is not fixed, but is rather the outcome of a dynamic process. We can move from a personal to a social identity (and back again) depending on the context. We can move between different social identities (for example, as a male, a father, a worker, a football fan, English, British, etc). Identity matters because it provides a prism through which we perceive the world, experience events, decide how to act, and understand our relationships to other people. It tells who is and who is not of us, who is for us and who is against us. Understanding the identity process is therefore key to assessing the impact that privacy and security policies have on people's behaviours. This is essential in order to be able to deliver systems that can express and analyse users' privacy requirements and, at runtime, self-adapt and guide users as they move from context to context. Broadly speaking, our proposed project asks the following two questions and attempts to answer them from both a social psychology and a computing perspective: Can privacy be a distributed quality (across 'the group')? If so, under what conditions might this be the case? Can the group protect the privacy of the individual? If so, how does the group manage the privacy-related behaviour of its members? The research challenges for the project are to devise non-intrusive yet rigorous ways in which to study privacy, both using pervasive technologies (such as life-logging cameras and biometric sensors) and in order to deliver more effective privacy management. At the heart of the project is a hypothesis that individuals are able to better manage their privacy by adopting or learning from the 'wisdom of groups' - we use this term as an acknowledgement of the crowd sourcing movement, also adapted by others in the catchphrase 'wisdom of friends'. Our novelty is in extending this idea to exploit the wisdom of particular subsets of people - groups whose positions and knowledge are more nuanced than a crowd. Our technical challenge is to investigate what we call the privacy dynamics of individuals as they relate to their membership of social, professional or other groups, to develop computational (machine learning) techniques that support such dynamics, and then to deliver privacy management capabilities interactively, autonomously, and adaptively as individuals' contexts change.

Data analysis is key to understanding timely phenomena from climate change to social media, from diseases to political conflicts, from the human brain to migration. In order to complement statistical analysis and modern machine learning approaches for data analysis, visualisation techniques and interactive interfaces support human-in-the-loop control over these systems as well as human sensemaking in cases where data is uncertain, requires greater overview for the generation of hypotheses, and effective communication to larger audiences. While more and more tools, such as Tableau, Gephi or Microsoft's PowerBI are democratising the use of data visualisation, using data visualisations to their full extend requires training novice analysts in tools, techniques, and interactive exploration, as well as communication and presentation. This project aims to free the analyst from their burden of exploring a data set from the beginning while having to chose among tools, learn their workflows, and create visualisations themselves. Rather, it aims to support novice analysts through a system that automatically displays information about a data set to an analyst while explaining visualisation techniques and findings. In such a "data tour", an analyst starts as a passive reader following a set of visualisations and textual explanations. Respective visualisations will be explained to the analyst. As the analyst becomes familiar with visualisations and their data, they are invited to explore the data by themselves through an interactive interface and communicate the system in which aspects they are most interested in. Creating effective data tours draws inspiration from previous work on using comics for data-driven storytelling (htttp://datacomics.net), visualisation cheatsheets (http://visualizationcheatsheets.github.io) and approaches to data visualisation literacy, data mining for networks, and human-computer interaction. To provide for specific data sets and contact with novice analysts for evaluating our tool, this project involves collaborators in history, archeology, sociology and network science and their complex geo-temporal networks including social networks, archeological trading networks, family networks, and Twitter networks. To create compelling data tours for these data sets we lack significant understanding of - current exploration strategies employed by analysts and their barriers to analysis, - ways of automatically extracting and annotating patterns-of-interest in networks, and - ways of creating meaningful explanatory sequences and high-level structures for data tours. This research involves a coordinated approach of field studies, visualisation and interface design, implementation, and user-centered evaluation. During a brief first phase, we will closely work with experts in Humanities research to create effective visualisations for their networks; in a second phase we mine and present insights from these data sets, and in the last phase, we investigate ways to structure and present findings in data tours. Our research will open new questions in how far storytelling and explaining visualisations can be supported by intelligent agents, i.e., computer programs, that partner with humans and engage in a dialogue. Our research may inspire new forms of intelligent interfaces that foresee an analyst's tasks and understand their specific interest in the data. Researchers in the digital humanities, social sciences, and network analysis will benefit from better support for visualising their geo-temporal networks and semi-automatic ways to analyse and lead to a better understanding of their data and new collaborative research agendas using visual analysis. Our project aims to provide impulses for commercial products and recommendation engines and will provide companies with knowledge and techniques to build customised data tours for their clients.

Context: software systems have become ever larger and more complex. This inevitably leads to software defects, whose debugging is estimated to cost the global economy 312 billion USD annually. Reducing the number of software defects is a challenging problem, and is particularly important considering the strong pressure towards rapid delivery. Such pressure impedes different parts of the software source code to all receive equally large amount of inspection and testing effort. With that in mind, machine learning approaches have been proposed for predicting defect-inducing changes in the source code as soon as these changes finish being implemented. Such approaches could enable software engineers to target special testing and inspection attention towards parts of the source code most likely to induce defects, reducing the risk of committing defective changes. Problem: the predictive performance of existing approaches is unstable, because the underlying defect generating process being modelled may vary over time (i.e., there may be concept drift). This means that practitioners cannot be confident about the prediction ability of existing approaches -- at any given point in time, predictive models may be performing very well or failing dramatically. Aim and vision: SPDISC aims at creating more stable models for predicting defect-inducing changes, through the development of a novel machine learning approach for automatically adapting to concept drift. When integrated with software versioning systems, the models will provide early, reliable and automated defect-inducing change alerts throughout the lifetime of software projects. Impact: SPDISC will enable a transformation in the way software developers review and commit their changes. By creating stable models to make software developers aware of defect-inducing changes as soon as these are implemented, it will allow targeted inspection and testing attention towards defect-inducing code throughout the lifetime of software projects. This will reduce the debugging cost and ultimately lead to better software quality. Proposed approach: an online learning algorithm will be developed to process incoming data as they become available, enabling fast reaction to concept drift. Concept drift will be detected using methods designed to cope with class imbalance, which typically occurs in prediction of defect-inducing software changes. Class imbalance refers to the issue of having a much smaller number of defect-inducing changes than the number of safe changes. The proposed approach will also make use of data from different projects (i.e., transfer learning between domains) to speed up adaptation to concept drift. Novelty: SPDISC is the first proposal to look into the stability of predictive performance over time in the context of defect-inducing software changes. Most previous work ignored the fact that predictions are required over time, being oblivious of the instability of predictive performance in this problem. To deal with instability, SPDISC will develop the first online transfer learning approach for predicting defect-inducing software changes. Ambitiousness: online transfer learning between domains with concept drift is not only a very new area of research in software engineering, but also in machine learning. Very few approaches exist for that, and none of them can deal with class-imbalanced problems. Therefore, SPDISC will not only advance software engineering by enabling a transformation in the way software developers review and commit their changes, but also advance the area of machine learning itself. Timeliness: given the current size and complexity of software systems, the increased number of life-critical applications, and the high competitiveness of the software industry, approaches for improving software quality and reducing the cost of producing and maintaining software are currently of utmost importance.