Humans interact with tens of objects daily, at home (e.g. cooking/cleaning) or outdoors (e.g. ticket machines/shopping bags), during working (e.g. assembly/machinery) or leisure hours (e.g. playing/sports), individually or collaboratively. When observing people interacting with objects, our vision assisted by the sense of hearing is the main tool to perceive these interactions. Let's take the example of boiling water from a kettle. We observe the actor press a button, wait and hear the water boil and the kettle's light go off before water is used for, say, preparing tea. The perception process is formed from understanding intentional interactions (called ideomotor actions) as well as reactive actions to dynamic stimuli in the environment (referred to as sensormotor actions). As observers, we understand and can ultimately replicate such interactions using our sensory input, along with our underlying complex cognitive processes of event perception. Evidence in behavioural sciences demonstrates that these human cognitive processes are highly modularised, and these modules collaborate to achieve our outstanding human-level perception. However, current approaches in artificial intelligence are lacking in their modularity and accordingly their capabilities. To achieve human-level perception of object interactions, including online perception when the interaction results in mistakes (e.g. water is spilled) or risks (e.g. boiling water is spilled), this fellowship focuses on informing computer vision and machine learning models, including deep learning architectures, from well-studied cognitive behavioural frameworks. Deep learning architectures have achieved superior performance, compared to their hand-crafted predecessors, on video-level classification, however their performance on fine-grained understanding within the video remains modest. Current models are easily fooled by similar motions or incomplete actions, as shown by recent research. This fellowship focuses on empowering these models through modularisation, a principle proven since the 50s in Fodor's Modularity of the Mind, and frequently studied by cognitive psychologists in controlled lab environments. Modularity of high-level perception, along with the power of deep learning architectures, will bring a new understanding to videos analysis previously unexplored. The targeted perception, of daily and rare object interactions, will lay the foundations for applications including assistive technologies using wearable computing, and robot imitation learning. We will work closely with three industrial partners to pave potential knowledge transfer paths to applications. Additionally, the fellowship will actively engage international researchers through workshops, benchmarks and public challenges on large datasets, to encourage other researchers to address problems related to fine-grained perception in video understanding.

As interaction on online Web-based platforms is becoming an essential part of people's everyday lives and data-driven AI algorithms are starting to exert a massive influence on society, we are experiencing significant tensions in user perspectives regarding how these algorithms are used on the Web. These tensions result in a breakdown of trust: users do not know when to trust the outcomes of algorithmic processes and, consequently, the platforms that use them. As trust is a key component of the Digital Economy where algorithmic decisions affect citizens' everyday lives, this is a significant issue that requires addressing. ReEnTrust explores new technological opportunities for platforms to regain user trust and aims to identify how this may be achieved in ways that are user-driven and responsible. Focusing on AI algorithms and large scale platforms used by the general public, our research questions include: What are user expectations and requirements regarding the rebuilding of trust in algorithmic systems, once that trust has been lost? Is it possible to create technological solutions that rebuild trust by embedding values in recommendation, prediction, and information filtering algorithms and allowing for a productive debate on algorithm design between all stakeholders? To what extent can user trust be regained through technological solutions and what further trust rebuilding mechanisms might be necessary and appropriate, including policy, regulation, and education? The project will develop an experimental online tool that allows users to evaluate and critique algorithms used by online platforms, and to engage in dialogue and collective reflection with all relevant stakeholders in order to jointly recover from algorithmic behaviour that has caused loss of trust. For this purpose, we will develop novel, advanced AI-driven mediation support techniques that allow all parties to explain their views, and suggest possible compromise solutions. Extensive engagement with users, stakeholders, and platform service providers in the process of developing this online tool will result in an improved understanding of what makes AI algorithms trustable. We will also develop policy recommendations and requirements for technological solutions plus assessment criteria for the inclusion of trust relationships in the development of algorithmically mediated systems and a methodology for deriving a "trust index" for online platforms that allows users to assess the trustability of platforms easily. The project is led by the University of Oxford in collaboration with the Universities of Edinburgh and Nottingham. Edinburgh develops novel computational techniques to evaluate and critique the values embedded in algorithms, and a prototypical AI-supported platform that enables users to exchange opinions regarding algorithm failures and to jointly agree on how to "fix" the algorithms in question to rebuild trust. The Oxford and Nottingham teams develop methodologies that support the user-centred and responsible development of these tools. This involves studying the processes of trust breakdown and rebuilding in online platforms, and developing a Responsible Research and Innovation approach to understanding trustability and trust rebuilding in practice. A carefully selected set of industrial and other non-academic partners ensures ReEnTrust work is grounded in real-world examples and experiences, and that it embeds balanced, fair representation of all stakeholder groups. ReEnTrust will advance the state of the art in terms of trust rebuilding technologies for algorithm-driven online platforms by developing the first AI-supported mediation and conflict resolution techniques and a comprehensive user-centred design and Responsible Research and Innovation framework that will promote a shared responsibility approach to the use of algorithms in society, thereby contributing to a flourishing Digital Economy.

As individuals, our daily routines rely on plastics in their many shapes and forms, whether as long lasting components of our homes and vehicles or as essential elements of important advances in medicine, water purification and infrastructure, or as packaging for cosmetics, food, drink, toiletries, cleaning products and healthcare products. These plastics are unrivalled materials: they are inexpensively synthesised, lightweight, recyclable and often deliver unmatchable performance. However, our love of plastics comes at a significant cost, as the environmental impact of these materials is massive, and growing. Genuinely sustainable plastics will need new forms of resource efficient materials, smart supply chains, and sustainable business practices, requiring holistic and integrated solutions. This proposal brings together diverse groups from across The University of Manchester to tackle this grand challenge of plastic waste. We seek solutions to the challenge of plastics pollution through an integrated approach that explicitly couples Manchester's strength in sociotechnological understanding and influence to our industry-guided solutions across chemistry, safety, materials, engineering and social sciences. The goal is to create a concerted, focussed consortium of diverse individuals who will lead stakeholder conversations, pitch multi-disciplinary projects that build from our strengths, and incubate these projects into translatable solutions. Through these collaborative efforts we will develop 6-12 projects building from our diverse expertise in urban recycling, sustainable business models, invisible plastic waste, valorising waste plastic streams, and new degradable polymers, and through them aim to: i) reduce the need for plastic by addressing demand, ii) improve the materials used to deliver better performance and clean degradation, iii) demonstrate new methods for recycling soft and mixed plastics/non-plastic films (currently very difficulty) and removal of micro plastics from source; and iv) create smart circular economies that allow users to take ownership of and reduce plastic waste. A multidisciplinary team of researchers at The University of Manchester will lead a portfolio of projects to tackle this grand challenge. Activities will be aligned with the first-of-its-kind Greater Manchester plan to drive down single-use plastics by 2020 and use the city-region as a living lab to innovate at speed and deploy solutions at scale.

The UK's carbon targets, as defined by the Climate Change Act of 2008, specify an emissions reduction of 80% by 2050, which the government has recently revised down to 'net zero' for the same year. In 2017, 17% of the UK's carbon emissions were associated with non-electric use in the residential sector (64.1 Mt CO2), the majority of which were associated with natural gas space heating, cooking and domestic hot water. The UK must therefore decarbonise residential heat to be able to meet its climate change targets, but, in combination with electric vehicles (EVs), this could lead to a 200-300% increase in the UK's annual electricity demand. In terms of deployment at scale, Air Source Heat Pumps (ASHP) operating either in isolation or as a hybrid gas system appear a key technology as they are not site specific and are applicable to both new build housing and retrofit. The UK's low voltage (LV) electricity network will not however, be able to operate with unconstrained electrical heating or EV charging loads. Both loads must be deferrable or scheduled in a manner to support the electricity network and maintain substations and feeders within limits. Household electric heating has the potential to operate as a significant deferrable load which LATENT is seeking to understand and harness. This can provide benefits across scales, namely to the UK (energy security and carbon targets), DNO (Distributed Network Operator as grid support), heat pump suppliers (by demonstrating added grid value), householders (in terms of bill reduction and avoidance of peaking dynamic tariffs) and electricity suppliers by applying aggregation techniques to minimise energy service costs. The key aim of LATENT therefore, is to be able to predict the impact of customers with electrical heating (predominantly ASHP) operating with 3rd party deferrable heating control on the LV network at the feeder / substation level. 3rd party control in this context would be through the energy service supplier, with whom, unlike the DNO, a household has an existing financial contract relationship. LATENT will inform industry of the potential of 3rd party control of deferrable heat through a rigorous field experiment, and, in doing so, accelerate the transition to decarbonised household heating. LATENT will determine the influence of householder personality trait (OCEAN traits: either positive / negative as Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism) alongside more traditional Census metrics such as educational attainment, house type etc to deliver a multi-variate regression model to describe deferrable heat reduction at the household level. A substation or feeder can then be analysed in terms of its household type mix (10% C+ detached, 30% E- flat etc) to produce a composite substation level, deferrable heat reduction estimate. This model will be realised through field trials with LATENT's industrial partner, Igloo Energy. Igloo have a customer base with smart heating systems and ASHP which support remote 3rd party control. LATENT will test (i) householder's stated acceptance to deferral of heating (in terms of temperature drop and duration) through focus groups and surveys, (ii) actual acceptance of heat deferral through heating season field trials, and (iii) operation of a commercial deferrable heat tariff with a sample of Igloo's customer base.

Global demand for broadband communications continues to increase substantially every year. A major factor contributing to this demand is the growing number of fixed and mobile broadband users, data-hungry applications like video as well as an ever-increasing number of network-connected everyday objects and machines. It is forecast that by 2020 the number of network-connected devices will reach 1000 times the world's population while data volumes transported over networks will progressively grow to Zettabytes and upwards. These trends pose entirely new challenges related to data volume, granularity, end-to-end connectivity and reach as well as increasing heterogeneity in network technologies (i.e. wireless and wired), networked-connected devices (i.e. sensors, mobile phones, computers, TVs, Data Centres) and services (i.e. Tbps data transfer for e-science, ultra-low latency financial transaction, real-time media streaming, kbps for sensor-based monitoring). Addressing these challenges necessitates radically new network models supporting convergence of traditionally separate network technology domains and offering high flexibility and adaptability in data granularity and throughput. TOUCAN aims to achieve ultimate network convergence enabled by a radically new technology agnostic architecture targeting a wide range of applications and end users. This architecture will facilitate optimal interconnection of any network technology domains, networked devices and data sets with high flexibility, resource and energy efficiency, and will aim to satisfy the full range of Quality of Service (QoS) and Quality of Experience (QoE) requirements. TOUCAN will realise its goals by including the network infrastructure and its control as part of the end-to-end service delivery chain. Important enablers will be that of separating the data and control planes, which will rely on Software Defined Networking (SDN) principles. TOUCAN will drastically evolve SDN to incorporate fundamentally new technology-specific interfacing and resource description followed by infrastructure resource abstraction, virtualisation and programmability. These features will enable any network technology and device to become "TOUCAN-ready" which means that the devices are programmable and interoperable. This is the foundation upon which the technology-agnostic feature of the TOUCAN architecture will be realized; thereby ultimate seamless end-to-end convergence will be achieved. TOUCAN will revolutionize the way we build and operate communication networks in a similar way that computer networks and more recently mobile terminals were transformed from platform-oriented to platform-agnostic solutions (e.g. through Linux and Android) and will drive towards commoditisation of network devices. Any new technology generation, regardless whether wired or wireless, will connect to the TOUCAN network in a plug-and-play fashion. Our research will open up a new network innovation eco-system, which will allow for the first time applications to compose, deploy and program their own virtual network infrastructures, as part of the service delivery mechanism to optimally support their specific and very diverse requirements. Such an environment will be able to adapt to challenging and unpredictable infrastructure and service evolution scenarios, meeting future application requirements. This highly challenging £12M project will bring together an Internationally renowned team of academics for a period of 5 years, allowing in depth technical exploration based on holistic and radical thinking in order to achieve the project goals. 58 person years of postdoctoral researcher time are requested for TOUCAN while the Universities have allocated a further 30 person years or more of PhD students. The TOUCAN consortium includes an impressive list of external partners who collectively are committing critical and tangible resources in excess of £3.6M.