The efficacy of digital solutions to increase energy efficiency, including technical optimisations and behavioural influence, has long been a subject of debate within sustainable HCI (SHCI). While the viewpoints of policymakers and academics are frequently published (and often contradictory), less is known about the views of those on the ground. In this paper we ask: What are energy professionals’ views of digital energy-saving interventions and their users? What are the challenges they face implementing these interventions? Based on a university campus case study with twelve semi-structured interviews and a focus group with energy and facilities’ professionals, we illustrate how they strongly advocate digital efficiency as a pathway to sustainability; yet, this optimism is in apparent tension with key barriers they identify to realising ‘their seamless visions’, particularly the complexities of the human behaviour they are seeking to optimise. These findings underscore the seductiveness of techno-optimism and the need for more systemic change.

Most research depends to some extent on technologies and computational infrastructures including, and perhaps especially, HCI.Despite the noted environmental impacts associated with information communication technology (ICT) globally, to date little consideration has been given as to how to limit the impact of research and innovation processes themselves.Working to understand the technical and cultural drivers of this impact within the specific but resource-intensive domain of High Performance Computing (HPC), we conducted 25 interviews with academic researchers, providers, funders, and commissioners of HPC.We find intersecting sociocultural and technical dimensions that link to research institutions like conferences, funders, and universities that reinforce and embed, rather than challenge, expectations of growth and waste.At a time when large scale cloud systems, generative AI and ever larger models are multiplying, we argue to de-escalate demand for computing, aiming for more moderate, responsible and meaningful use of computational infrastructures-including within HCI itself.

Demand for delivery of take-away meals to customers has been growing worldwide, with deliveries often performed by non-specialised gig economy couriers working for online platform operators such as Deliveroo or Just Eat. This has led to the introduction of the ‘meal delivery problem’, characterised by a series of individual pickup and delivery tasks to be assigned to available couriers. While there is a vast set of algorithms proposed in the literature that aim to minimise total workload, very little attention has been given to equitably distributing work between couriers. We propose a new multi-objective problem that is aiming at distributing orders equitably between couriers as well as minimising total workload, where all information is known upfront. We propose an integer linear programming (ILP) model with a weighted objective function that is used to derive the Pareto front in small-scale problems by exploiting the ϵ−constraint approach. This formulation has been proven to solve in a reasonable time for problems with up to 60 orders, however, the optimal Pareto front can only be computed within a reasonable time for problems up to 30 orders. For problems with more orders, we propose a Variable Neighbourhood Search (VNS) algorithm, for which the fitness evaluation evolves in order to explore a wider set of the solution space. The VNS is compared against the ILP and also tested on more realistic size instances with up to 3123 orders, improving the performance over the business as usual and shows that equitable distribution of work can be achieved alongside reducing the total travelled distance.

The heterogeneity of Deep Learning models, libraries, and hardware poses an important challenge for improving model inference performance. Auto-tuners address this challenge via automatic tensor program optimization towards a target-device. However, auto-tuners incur a substantial time cost to complete given their design necessitates performing tensor program candidate measurements serially within an isolated target-device to minimize latency measurement inaccuracy. In this article we propose DOPpler, a parallel auto-tuning measurement infrastructure. DOPpler allows for considerable auto-tuning speedup over conventional approaches whilst maintaining high-quality tensor program optimization. DOPpler accelerates the auto-tuning process by proposing a parallel execution engine to efficiently execute candidate tensor programs in parallel across the CPU-host and GPU target-device, and overcomes measurement inaccuracy by introducing a high-precision on-device measurement technique when measuring tensor program kernel latency. DOPpler is designed to automatically calculate the optimal degree of parallelism to provision fast and accurate auto-tuning for different tensor programs, auto-tuners and target-devices. Experiment results show that DOPpler reduces total auto-tuning time by 50.5% on average whilst achieving optimization gains equivalent to conventional auto-tuning infrastructure.

Cloud datacenters capable of provisioning high performance Machine Learning-as-a-Service (MLaaS) at reduced resource cost is achieved via auto-tuning: automated tensor program optimization of Deep Learning models to minimize inference latency within a hardware device. However given the extensive heterogeneity of Deep Learning models, libraries, and hardware devices, performing auto-tuning within Cloud datacenters incurs a significant time, compute resource, and energy cost of which state-of-the-art auto-tuning is not designed to mitigate. In this paper we propose Trimmer, a high performance and cost-efficient Deep Learning auto-tuning framework for Cloud datacenters. Trimmer maximizes DL model performance and tensor program cost-efficiency by preempting tensor program implementations exhibiting poor optimization improvement; and applying an ML-based filtering method to replace expensive low performing tensor programs to provide greater likelihood of selecting low latency tensor programs. Through an empirical study exploring the cost of DL model optimization techniques, our analysis indicates that 26–43% of total energy is expended on measuring tensor program implementations that do not positively contribute towards auto-tuning. Experiment results show that Trimmer achieves high auto-tuning cost-efficiency across different DL models, and reduces auto-tuning energy use by 21.8–40.9% for Cloud clusters whilst achieving DL model latency equivalent to state-of-the-art techniques.

By CHI 2022, fifteen years will have passed since the emergence of Sustainable HCI (SHCI), which now constitutes an important subfield of HCI. In this paper, we draw on two SHCI corpora to ask: Has SHCI progressed? How has the field responded to prominent critiques? Have we identified and adopted constructive strategies for impacting environmental unsustainability? We further show the wide array of competencies SHCI researchers have been called to develop, and how this has been reflected in subsequent work. Our analysis identifies significant shifts in the SHCI landscape, toward research that is diverse and holistic, but also away from efforts to address the urgent climate crisis. We posit that SHCI has tended to take on far more than it could reasonably expect to deliver, and propose ‘Green Policy informatics’ as a pathway that enables SHCI to leverage a more traditional HCI skillset in addressing climate change.

Growing and even excessive use of digital technology has unquestionably fuelled demand for digital devices and online services leading to a wide range of societal and environmental impacts. In sustainability terms, ICT as a whole is estimated to produce up to nearly 4% of global greenhouse gas emissions. As presumed responsible innovators, the HCI community should now consider design strategies that will reduce use and demand for digital technology for the good of both its users and the planet—strategies perhaps even seen as retrogressive in an era where digital technology is constantly implicated in innovation and economic growth. Prior work has noted the potential to design “more moderate” interactions for sustainability, simultaneously addressing negative societal impacts on users’ wellbeing, relationships, productivity at work, and privacy. In this paper, we explore how we may design intentionally moderate digital interactions that retain our participants’ “more meaningful” experiences. We report on the outcomes of two design workshops to uncover experiences of meaningful device and service use, to inform practical designs for ‘moderate and meaningful’ interaction. From this, we offer design recommendations that aim to address the multiple negative impacts that digital technology can create, and discuss the possible barriers to these designs.

To accelerate the training of Deep Learning (DL) models, clusters of machines equipped with hardware accelerators such as GPUs are leveraged to reduce execution time. State-of-the-art resource managers are needed to increase GPU utilization and maximize throughput. While co-locating DL jobs on the same GPU has been shown to be effective, this can incur interference causing slowdown. In this article we propose Horus: an interference-aware and prediction-based resource manager for DL systems. Horus proactively predicts GPU utilization of heterogeneous DL jobs extrapolated from the DL model's computation graph features, removing the need for online profiling and isolated reserved GPUs. Through micro-benchmarks and job co-location combinations across heterogeneous GPU hardware, we identify GPU utilization as a general proxy metric to determine good placement decisions, in contrast to current approaches which reserve isolated GPUs to perform online profiling and directly measure GPU utilization for each unique submitted job. Our approach promotes high resource utilization and makespan reduction; via real-world experimentation and large-scale trace driven simulation, we demonstrate that Horus outperforms other DL resource managers by up to 61.5 percent for GPU resource utilization, 23.7-30.7 percent for makespan reduction and 68.3 percent in job wait time reduction.

From Fanger's seminal work on thermal comfort in the 1970s, standards governing temperatures in the workplace enshrine clothing level calculations based on full business suits, and building regulations developed using only male metabolic data, locking in a default male perspective. Even later work that highlights gender biases with regard to metabolism calculation, inclusive of both genders has focused on younger women, and the voices of older working women are missing from this discourse. We invited women over 45 to explore what they find important in workplace thermal comfort, and how devices and interfaces might meet their needs and also encourage thermal adaptivity. Our study highlights factors such as 'fresh air', and the importance of empathy to fellow inhabitants. We bring new voices to the thermal comfort discourse which supports reducing energy use in the workplace, improving thermal environments and ensuring the needs of a diverse, aging workforce are considered.

Widespread sensing devices enable a world in which physical spaces become personalised in the presence of mobile users. An important example of such personalisation is the use of pervasive displays to show content that matches the requirements of proximate viewers. Despite prior work on prototype systems that use mobile devices to personalise displays, no significant attempts to trial such systems have been carried out. In this article, we report on our experiences of designing, developing and operating the world’s first comprehensive display personalisation service for mobile users. Through a set of rigorous quantitative measures and 11 potential user/stakeholder interviews, we demonstrate the success of the platform in realising display personalisation, and offer a series of reflections to inform the design of future systems.

Internet use and online services underpin everyday life, and the resultant energy demand is almost entirely hidden, yet significant and growing: it is anticipated to reach 21% of global electricity demand by 2030 and to eclipse half the greenhouse gas emissions of transportation by 2040. Driving this growth, real-time video streaming ('watching') is estimated at around 50% of all peak data traffic. Using a mixed-methods analysis of the use of 66 devices (e.g. smart TVs, tablets) across 20 participants in 9 households, we reveal the online activity of domestic watching and provide a detailed exploration of video-on-demand activities. We identify new ways in which watching is transitioning in more rather than less data demanding directions; and explore the role HCI may play in reducing this growing data demand. We further highlight implications for key HCI and societal stakeholders (policy makers, service providers, network engineers) to tackle this important issue.

The evaluation of research artefacts is an important step to validate research contributions. Sub-disciplines of HCI often pursue primary goals other than usability, such as Sustainable HCI (SHCI), HCI for development, or health and wellbeing. For such disciplines, established evaluation methods are not always appropriate or sufficient, and new conventions for identifying, discussing, and justifying suitable evaluation methods need to be established. In this paper, we revisit the purpose and goals of evaluation in HCI and SHCI, and elicit five key elements that can provide guidance to identifying evaluation methods for SHCI research. Our essay is meant as a starting point for discussing current and improving future evaluation practice in SHCI; we also believe it holds value for other subdisciplines in HCI that encounter similar challenges while evaluating their research.

Road congestion, air pollution and sustainability are increasingly important in major cities. We look to understand how last-mile deliveries in the parcel sector are impacting our roads. Using formative field work and quantitative analysis of consignment manifests and location data, we identify how the effectiveness of life-style couriers is contributing to both environmental and non-environmental externalities. This paper presents an analysis of delivery performances and practices in last-mile logistics in central London, quantifying the impacts differing levels of experience have on overall round efficiency. We identify eleven key opportunities for technological support for last-mile parcel deliveries that could contribute to both driver effectiveness and sustainability. We finish by examining how HCI can lead to improved environmental and social justice by re-considering and realizing future collaborative visions in last-mile logistics.

Motivated by mobile devices' growing demand for connectivity, and concern in HCI with the energy intensity and sustainability of networked services, in this paper we reveal the impact of applications on smartphones and tablets in terms of network demand and time use. Using a detailed mixed methods study with eight participants, we first provide an account of how data demand has meaning and utility in our participants' social practices, and the timing and relative impacts of these. We then assess the scale of this demand by drawing comparison between our fine-grained observations and a more representative dataset of 398 devices from the Device Analyzer corpus. Our results highlight the significant categories of data demanding practice, and the identification of where changes in app time and duration of use might reduce or shift demand to reduce services' impacts.

This paper empirically explores the role that mobile devices have come to play in everyday practice, and how this links to demand for network connectivity and online services. After a preliminary device-logging period, thirteen participants were interviewed about how they use their iPhones or iPads. Our findings build a picture of how, through use of such devices, a variety of daily practices have come to depend upon a working data connection, which sometimes surges, but is at least always a trickle. This aims to inform the sustainable design of applications, services and infrastructures for smartphones and tablets. By focusing our analysis in this way, we highlight a little-explored challenge for sustainable HCI and discuss ideas for (re)designing around the principle of 'light-weight' data 'needs'.

In many parts of the world, the electricity supply industry makes the task of dealing with unpredictable spikes and dips in production and demand invisible to consumers, maintaining a seemingly unlimited supply. A future increase in reliance on time-variable renewable sources of electricity may lead to greater fluctuations in supply. We engaged remote islanders as equal partners in a research project that investigated through technology-mediated enquiry the topic of synchronising energy consumption with supply, and together built a prototype renewable energy forecast display. A number of participants described a change in their practices, saving high energy tasks for times when local renewable energy was expected to be available, despite having no financial incentive to do so. The main contributions of this paper are in: 1) the results of co-development sessions exploring systems supporting synchronising consumption with supply and 2) the findings arising from the deployment of the prototype.

Social Software Engineering (Social SE), that is SE aiming to promote positive social change, is a rapidly emerging area. Here, software and digital artefacts are seen as tools for social change, rather than end products or 'solutions'. Moreover, Social SE requires a sustained buy-in from a range of stakeholders and end-users working in partnership with multidisciplinary software development teams often at a distance. This context poses new challenges to software engineering: it requires both an agile approach for handling uncertainties in the software development process, and the application of participatory, creative design processes to bridge the knowledge asymmetries and the geographical distances in the partnership. This paper argues for the role of design thinking in Social SE and highlights its implications for software engineering in general. It does so by reporting on the contributions that design thinking -- and in particular physical design -- has brought to (1) the problem space definition, (2) user requirements capture and (3) system feature design of a renewable energy forecasting system developed in partnership with a remote Scottish Island community.

Tightly regulating indoor building temperatures using mechanical heating and cooling contributes significantly to worldwide greenhouse gas emissions. One promising approach for reducing the energy demand associated with indoor climate control is the adaptive model for thermal comfort. In this paper, we explore the challenges and opportunities for supporting the transition toward adaptive thermal comfort in conventionally heated buildings. We replaced the heating control system for eight university undergraduates living on campus for fifty days from January-March 2013. We report on the participants' experiences of living with and adapting to the change in conditions. We reflect on the lessons arising from our intervention for researchers and practitioners seeking to design for sustainability and thermal comfort.

To date, research in sustainable HCI has dealt with eco-feedback, usage and recycling of appliances within the home, and longevity of portable electronics such as mobile phones. However, there seems to be less awareness of the energy and greenhouse emissions impacts of domestic consumer electronics and information technology. Such awareness is needed to inform HCI sustainability researchers on how best to prioritise efforts around digital media and IT. Grounded in inventories, interview and plug energy data from 33 undergraduate student participants, our findings provide the context for assessing approaches to reducing the energy and carbon emissions of media and IT in the home. In the paper, we use the findings to discuss and inform more fruitful directions that sustainable HCI research might take, and we quantify how various strategies might have modified the energy and emissions impacts for our participants.

There is increasing interest in using digital signage to deliver highly personalised content. However, display personalization presents a number of architectural design challenges in particular, how best to provide personalisation without unduly compromising viewers' privacy. While previous research has focused on understanding specific elements of the overall vision, our work presents details of the first significant attempt at a system that integrates future pervasive display networks and mobile devices to support display personalisation. We describe a series of usage models and design goals for display personalisation and then present Tacita, a system that supports these models and goals. Our architecture includes mobile, display and cloud-based elements and provides comprehensive personalisation features while preventing the creation of user profiles within the display infrastructure, thus helping to preserve users' privacy. An initial evaluation of our prototype implementation of the architecture is also included and demonstrates the viability of the Tacita approach.

As the cost of display hardware falls so the number of public display networks being deployed is increasing rapidly. While these networks have traditionally taken the form of digital signage used for advertising and information there is increasing interest in the vision of "open display networks" [8]. A key component of any open display network is an effective channel for disseminating applications created by third-parties and recent research has proposed a display-oriented "application store" as one such channel [5]. In this paper we present a critical analysis of the requirements and design of display application stores -- providing insights designed to help the implementers of future application stores.

In four university student kitchens over twenty-one days, we captured participants' food preparation activity, quantified the greenhouse gas emissions and direct energy connected to the food and cooking, and talked to participants about their food practices. Grounded in this uniquely detailed micro-account, our findings inform sustainable design for cooking and eating at home and quantify the potential impacts. We outline the relation of the impacts to our participants' approaches to everyday food preparation, the organisation of their time, and the role of social meals. Our technique allows evaluation of opportunities for sustainable intervention design: at the appliance, in the digitally-mediated organisation of meals and inventory management, and more broadly in reflecting upon and reshaping diet.

In many parts of the world, mechanical heating and cooling is used to regulate indoor climates, with the aim of maintaining a uniform temperature. Achieving this is energy-intensive, since large indoor spaces must be constantly heated or cooled, and the difference to the outdoor temperature is large. This paper starts from the premise that comfort is not delivered to us by the indoor environment, but is instead something that is pursued as a normal part of daily life, through a variety of means. Based on a detailed study of four university students over several months, we explore how Ubicomp technologies can help create a more sustainable reality where people are more active in pursuing and maintaining their thermal comfort, and environments are less tightly controlled and less energy-intensive, and we outline areas for future research in this domain.

Trends such as the increasing adoption of smartphones, the development of the service-oriented internet, and diffusion of sensing technologies into cities have the potential to combine in order to form a ubiquitous computing infrastructure. At the same time, as the computer diffuses into the physical world, it loses its location-neutrality, exposing the urgent need for a debate of design choices in ubiquitous computing. In this paper, we discuss the process of urban development as a source of inspiration for such design choices. Looking from the ground up, of particular interest is the opportunity to localize and democratize an emerging ubiquitous computing infrastructure. The design choices we negotiate today will determine the society in which we will live in the future.

Many global internet services today can be thought of as being managed 'top-down' without much appreciation or requirement for local control. New services such as location-based data stores implemented in cities around the world suggest opportunities for novel forms of management of data with relevance to a local context. We argue that such localization of ubiquitous system management provides an opportunity in supporting local off-line interaction and 'community building' and that urban development, which requires interaction between members of different communities, presents an interesting case where such systems could be realised.

Transient use of displays by mobile users was prophesied two decades ago. Today, convergence of a range of technologies enable the realization of this vision. For researchers in this space, one key question is where to physically locate the application for which the display has been appropriated. The emergence of cloud and cloudlet computing has increased the range of possible locations. In this paper we focus on understanding the extent to which application location impacts user experience when appropriating displays. We describe a usage model in which public displays can be appropriated to support spontaneous use of interactive applications, present an example architecture based on cloudlets, and explore how application location impacts user experience.

In this paper we reflect on our long-term experiences of developing, deploying and supporting an experimental digital signage system. Existing public display systems almost always feature a single point of control that is responsible for scheduling content for presentation on the network and provide sophisticated mechanisms for controlling play-out timing and relative ordering. Our experiences suggest that such complex feature-sets are unnecessary in many cases and may be counter productive in signage systems. We describe an alternative, simpler paradigm for encouraging widespread use of signage systems based on shared 'content channels' between content providers and display owners. Our system has been in continuous use for approximately 3 years. We reflect and draw lessons from how our user community has adopted and used the resulting public display network. We believe that these reflections will be of benefit to future developers of ubiquitous display networks.

An increasing trend in mobile and pervasive computing is the augmentation of everyday public spaces with local computation - leading to so called smart environments. However, there are no well accepted techniques for supporting spontaneous interaction between mobile users and these smart environments, though a wide range of techniques have been explored ranging from gesture recognition to downloading applications to a user's phone. In this paper we explore an approach to supporting such interaction based on the use of Bluetooth Device (user-friendly) Names as a control channel between users' mobile phones and computational resources in their local environment. Such an approach has many advantages over existing techniques though it is not without limitations. Our work focuses specifically on the use of Device Names to control and customize applications on large public displays in a campus environment. This paper describes our basic approach, a number of applications that we have constructed using this technique and the results of our evaluation work which has included a range of user studies and field trials. The paper concludes with an assessment of the viability of using our approach for interaction scenarios involving mobile users and computationally rich environments.

Large screen displays are being increasingly deployed in public areas for advertising, entertainment, and information display. Recently we have witnessed increasing interest in supporting interaction with such displays using personal mobile devices. To enable the rapid development of public large screen interactive applications, we have designed and developed the MAGIC Broker. The MAGIC Broker provides a set of common abstractions and a RESTful Web services protocol to easily program interactive public large screen display applications with a focus on mobile device interactions. We have carried out a preliminary evaluation of the MAGIC Broker via the development of a number of prototypes and believe our toolkit is a valid first step in developing a generic support infrastructure to empower developers of interactive large screen display applications.

A major challenge in pervasive computing is to develop systems that can reliably recognize human activity patterns, such as bathing from sensor data. Typical sensor deployments generate sparse datasets with thousands of sensor readings and few instances of activities. The imbalance between the number of features (i.e. sensors firing) and the classification targets (i.e. activities) complicates the learning process. In this paper, we propose a novel framework for discovering relationships between sensor signals and observed human activities from sparse datasets. The framework builds on the use of Bayesian networks for modeling activities by representing statistical dependencies between sensors. This allows us to solve two key problems: firstly, how to automatically determine an effective structure for a Bayesian network that recognizes a particular activity without human intervention; and, secondly, we address the pragmatic problem of sparse training data, where the data available to train the activity recognizers is limited. In our approach, we `learn' the structure of the Bayesian networks automatically from the sensor data. We optimize this process in 3 ways: firstly, we perform multicollinearity analysis to focus on orthogonal sensor data with minimal redundancy. Secondly, we propose Efron's bootstrapping to generate large training sets that capture important features of an activity. Finally, we find the best Bayesian network that explains our data using a heuristic search that is unbiased to the ordering between consecutive variables. We evaluate our approach using a data set gathered from MIT's PlaceLab. The inferred networks correctly identify activities for 85% of the time

This document introduces a novel access control architecture for publicly accessible wireless overlay networks. The architecture is designed to address the problems of ubiquitous Internet service provisioning within the city of Lancaster. The proposed access control mechanism is based on the concepts of secure user authentication, packet marking, and network-level packet filtering. The novelty of the architecture lies in its use of microcellular layer three networks to acquire fine grained access control in a link independent manner. The paper describes the concepts behind the access control architecture and demonstrates to what extent it addresses the security, performance and extensibility concerns of public access packet switched wireless networks.

In this paper, we describe our experiences of developing and evaluating GUIDE, an intelligent electronic tourist guide. The GUIDE system has been built to overcome many of the limitations of the traditional information and navigation tools available to city visitors. For example, group-based tours are inherently inflexible with fixed starting times and fixed durations and (like most guidebooks) are constrained by the need to satisfy the interests of the majority rather than the specific interests of individuals. Following a period of requirements capture, involving experts in the field of tourism, we developed and installed a system for use by visitors to Lancaster. The system combines mobile computing technologies with a wireless infrastructure to present city visitors with information tailored to both their personal and environmental contexts. In this paper we present an evaluation of GUIDE, focusing on the quality of the visitor's experience when using the system.

The GUIDE system has been developed to provide city visitors with a hand-held context-aware tourist guide. The system has been successfully deployed in a major tourist destination and is currently at the stage where it is publicly available to visitors who wish to explore the city. Reaching this stage has been the culmination of a number of distinct research efforts. In more detail, the development of GUIDE has involved: capturing a real set of application requirements, investigating the properties of a cell-based wireless communications technology in a built-up environment and deploying a network based on this technology around the city, designing and populating an information model to represent attractions and key buildings within the city, prototyping the development of a distributed application running across portable GUIDE units and stationary cell-servers and finally, evaluating the entire system during an extensive field-trial study. This paper reports on our results in each of these areas. We believe that through our work on the GUIDE project we have produced a blueprint for the development of interactive context-aware systems that should be of real value to those in the community who wish to develop such systems in a practical environment.

This article considers the importance of context in mobile systems. It considers a range of context-related issues and focus on location as a key issue for mobile systems. A design framework is described consisting of taxonomies of location, mobility, population, and device awareness. The design framework inorms the construction of a semantic model of space for mobile systems. The semantic model is reflected in a computational model built on a distriuted platform that allows contextual information to be shared across a number of mobile devices. The framework support the design of interactive mobile systems while the platform supports their rapid development.

Due to increasing computational power, long-life batteries and wireless links with improving bandwidth, distributed multimedia applications have become more and more attractive for mobile users. However, the varying quality of service (QoS) on the wireless link is still a major problem. One generally accepted solution to this problem utilises "proxies" which act on behalf of the mobile user and modify the communication data stream in order to match the current QoS conditions. These proxies are generally dependent on protocol layers, applications or application data. Therefore, in order to enable proxy objects to be flexibly inserted for a given multimedia stream, an architecture is required for managing supplied and installed proxy objects.