The sharing economy model is a contested concept: on one hand, its proponents have praised it to be enabler of fair marketplaces, with all participants receiving equal opportunities; on the other hand, its detractors have criticised it for actually exacerbating preexisting societal inequalities. In this paper, we propose a scalable quantitative method to measure participants' diversity and inclusion in such marketplaces, with the aim to offer evidence to ground this debate. We apply the method to the case of the Airbnb hospitality service for the city of London, UK. Our findings reveal that diversity is high for gender, but not so for age and ethnicity. As for inclusion, we find strong signals of homophily both in terms of gender, age and ethnicity, thus suggesting that under-represented groups have significantly fewer opportunities to gain from this market model. Interestingly, the sentiment associated to same-group (homophilic) interactions is just as positive as that associated to heterophilic ones, even after controlling for Airbnb property's type, price and location. This suggests that increased diversity and inclusion are desirable not only for moral but also for economic and market reasons.

Border Theory suggests individuals create borders to manage the transitions between work and family (or, more generally, life) domains. The degree of separation or integration of domains across borders has an impact on the balance between work and life. Previous studies have shown individuals who perceive balance between work and life domains tend to be more satisfied with their lives, reporting higher physical and mental health. At times of crisis, such as during a pandemic, borders can be disrupted, affecting work-life balance and leading to a short- or long-term negative impact on well-being. Border theory provides a systematic lens through which to study these changes. However, changes cannot be studied using interviews or diaries as these are not at the scale required when societal disruptions occur. In this paper, we explore the feasibility of using a computational linguistic approach to operationalize border theory at scale, using readily available social media data. In particular, we make two main contributions. First, we design metrics to measure key characteristics of borders. This involves the application of a transformer-based topic modeling technique, BERTopic, to detect topics from social media data. Second, we apply this operationalization to a case study of around a million tweets posted by nearly two hundred teachers and journalists in the UK from the beginning of 2019 to the end of 2022. In so doing, we longitudinally study and compare the changes in borders between work and life before, during, and after COVID-19 lockdown periods.

AI regulations are expected to prohibit machine learning models from using sensitive attributes during training. However, the latest Natural Language Processing (NLP) classifiers, which rely on deep learning, operate as black-box systems, complicating the detection and remediation of such misuse. Traditional bias mitigation methods in NLP aim for comparable performance across different groups based on attributes like gender or race but fail to address the underlying issue of reliance on protected attributes. To partly fix that, we introduce NLPGuard, a framework for mitigating the reliance on protected attributes in NLP classifiers. NLPGuard takes an unlabeled dataset, an existing NLP classifier, and its training data as input, producing a modified training dataset that significantly reduces dependence on protected attributes without compromising accuracy. NLPGuard is applied to three classification tasks: identifying toxic language, sentiment analysis, and occupation classification. Our evaluation shows that current NLP classifiers heavily depend on protected attributes, with up to 23% of the most predictive words associated with these attributes. However, NLPGuard effectively reduces this reliance by up to 79%, while slightly improving accuracy.

Studies on depression in the workplace have mostly investigated its impact on individual employees. Little is known about its association with the company as a whole, or the state where the company is based. This is due to the lack of scalable methodologies operationalizing depression in the specific context of the workplace, and of data documenting potential distress. In this work, we adapted a work-related depression scale called Occupational Depression Inventory (ODI), gathered more than 350K employee reviews of 104 major companies across the whole US for the (2008-2020) years, and developed a deep-learning framework (called AutoODI) scoring these reviews on a composite ODI score. Presence of ODI mentions manifested itself not only at micro-level (companies scoring high in ODI suffered from low stock growth) but also at macro-level (states hosting these companies were associated with high depression rates, talent shortage, and economic deprivation). This new way of applying AutoODI onto company reviews offers both theoretical implications for the literature in computational social science, occupational health and economic geography, and practical implications for companies and policy makers.

Voice assistants have become hugely popular in the home as domestic and entertainment devices. Recently, there has been a move towards developing them for work settings. For example, Alexa for Business and IBM Watson for Business were designed to improve productivity, by assisting with various tasks, such as scheduling meetings and taking minutes. However, this kind of assistance is largely limited to planning and managing user's work. How might they be developed to do more by way of empowering people at work? Our research is concerned with achieving this by developing an agent with the role of a facilitator that assists users during an ongoing task. Specifically, we were interested in whether the modality in which the agent interacts with users makes a difference: How does a voice versus screen-based agent interaction affect user behavior? We hypothesized that voice would be more immediate and emotive, resulting in more fluid conversations and interactions. Here, we describe a user study that compared the benefits of using voice versus screen-based interactions when interacting with a system incorporating an agent, involving pairs of participants doing an exploratory data analysis task that required them to make sense of a series of data visualizations. The findings from the study show marked differences between the two conditions, with voice resulting in more turn-taking in discussions, questions asked, more interactions with the system and a tendency towards more immediate, faster-paced discussions following agent prompts. We discuss the possible reasons for why talking and being prompted by a voice assistant may be preferable and more effective at mediating human-human conversations and we translate some of the key insights of this research into design implications.

In this paper, we study the app store as a phenomenon from the developers' perspective to investigate the extent to which app stores affect software engineering tasks. Through developer interviews and questionnaires, we uncover findings that highlight and quantify the effects of three high-level app store themes: bridging the gap between developers and users, increasing market transparency and affecting mobile release management. Our findings have implications for testing, requirements engineering and mining software repositories research fields. These findings can help guide future research in supporting mobile app developers through a deeper understanding of the app store-developer interaction.

Airbnb is one of the most successful examples of sharing economy marketplaces. With rapid and global market penetration, understanding its attractiveness and evolving growth opportunities is key to plan business decision making. There is an ongoing debate, for example, about whether Airbnb is a hospitality service that fosters social exchanges between hosts and guests, as the sharing economy manifesto originally stated, or whether it is (or is evolving into being) a purely business transaction platform, the way hotels have traditionally operated. To answer these questions, we propose a novel market analysis approach that exploits customers’ reviews. Key to the approach is a method that combines thematic analysis and machine learning to inductively develop a custom dictionary for guests’ reviews. Based on this dictionary, we then use quantitative linguistic analysis on a corpus of 3.2 million reviews collected in 6 different cities, and illustrate how to answer a variety of market research questions, at fine levels of temporal, thematic, user and spatial granularity, such as (i) how the business vs social dichotomy is evolving over the years, (ii) what exact words within such top-level categories are evolving, (iii) whether such trends vary across different user segments and (iv) in different neighbourhoods.

Newly emerging urban IoT infrastructures are enabling novel ways of sensing how urban spaces are being used. However, the data produced by these systems are largely context-agnostic, making it difficult to discern what patterns and anomalies in the data mean. We propose a hybrid data approach that combines the quantitative data collected from an urban IoT sensing infrastructure with qualitative data contributed by people answering specific kinds of questions in situ. We developed a public installation, Roam-io, to entice and encourage the public to walk-up and answer questions to suggest what the data might represent and enrich it with subjective observations. The findings from an in the wild study on the island of Madeira showed that many passers-by stopped and interacted with Roam-io and attempted to make sense of the data and contribute in situ observations.

Code-free software similarity detection techniques have been used to support different software engineering tasks, including clustering mobile applications (apps). The way of measuring similarity may affect both the efficiency and quality of clustering solutions. However, there has been no previous comparative study of feature extraction methods used to guide mobile app clustering. In this paper, we investigate different techniques to compute the similarity of apps based on their textual descriptions and evaluate their effectiveness using hierarchical agglomerative clustering. To this end we carry out an empirical study comparing five different techniques, based on topic modelling and keyword feature extraction, to cluster 12,664 apps randomly sampled from the Google Play App Store. The comparison is based on three main criteria: silhouette width measure, human judgement and execution time. The results of our study show that using topic modelling, in addition to collocation-based and dependency-based feature extractors perform similarly in detecting app-feature similarity. However, dependency-based feature extraction performs better than any other in finding application domain similarity ( ρ = 0.7, p − v a l u e < 0.01). Current categorisation in the app store studied does not exhibit a good classification quality in terms of the claimed feature space. However, a better quality can be achieved using a good feature extraction technique and a traditional clustering method.

Existing platforms for sharing locative digital content rely on the use of mobile phones for accessing the content. This can be a major deterrent to wider public access and also hinders immediacy and 'in the moment' discoverability. Building on previous work in situated public installations, we developed Pinsight, a novel platform for enabling end-users, such as local communities, to create and share digital content in-situ with public audiences through physical interactive devices. Pinsight is based on a set of design principles that focus on supporting both the expressiveness of content creators and the appeal to public audiences. This paper describes the design of the platform and how it supports sharing knowledge in ways different to conventional media. Through preliminary evaluations and two in-the-wild studies, we explore how such a situated technology can be used by different user groups (content designers, history communities, local residents) for sharing content with public audiences (visitors, pedestrians, residents) in different contexts.

In many developing countries there remains a limited view on the socioeconomic status of the population, owing to the high cost associated with detailed and comprehensive surveying. This situation has encouraged a number of researchers to attempt to exploit alternative sources of data in order to derive estimates, including mobile phone data, which offers a rich depiction of the social dynamics of a population. Meanwhile, from the level of the individual to the city, access to information has been posited as an important factor in determining prosperity and economic development. In this paper we explore this relationship by simulating the flow of information through a mobile phone call graph in two sub-Saharan countries. We find a strong relationship between a location's average wealth and its access to information as determined by the simulations in one country, and a weaker correlation in the second country. This finding adds to recent evidence that mining patterns from mobile phone data represents a viable means to estimate poverty in places where traditionally derived estimates are lacking. We further investigate the impact of various factors on the empirical results in order to explain the variation between the two countries.

Crowdsourcing platforms have become important information providers after disaster events. While they can build on some prior experiences, it is not yet well understood how contributor capacity for such activities is constituted. To what extent are initiatives building a dormant task force that springs to action when it is needed? Alternatively, do they mainly rely on the recruitment of new contributors during disaster events, possibly at the expense of contribution quality? We seek to develop a better understanding of these relationships, using the example of the Humanitarian OpenStreetMap Team. In a large-scale quantitative study, we assess the outcomes of 26 campaigns with almost 20,000 participants. We find that event-centric campaigns can be significant recruiting and reactivation events, however that this is not guaranteed. Our analytical methods provide a means of interpreting key differences in outcomes. We close with recommendations relating to the promotion and coordination of event-centric campaigns in HOT and related platforms.

Crowd-mapping is a form of collaborative work that empowers users to share geographic knowledge. Despite geographic information being intrinsically evolving, little research has so far gone into analysing maintenance practices in these domains. In this paper, we quantitatively capture maintenance dynamics in geographic crowd-sourced datasets, in terms of: the extent to which different maintenance actions are taking place, the type of spatial information that is being maintained, who engages in these practices and where. We apply this method to 117 countries in OpenStreetMap, one of the most successful examples of geographic crowd-sourced datasets. Furthermore, we explore what triggers maintenance, by means of an online survey to which 96 OpenStreetMap contributors took part. Our findings reveal that, although maintenance practices vary substantially from country to country in terms of how widespread they are, strong commonalities exist in terms of what metadata is being maintained, by whom, and what triggers them.

Prior research suggests that public negative feedback on social knowledge sharing platforms can be powerfully demotivating to newcomers, particularly when it involves peer feedback mechanisms such as ratings and commenting systems. What is the impact on newcomer retention when feedback is private, and from a single peer reviewer? We study these effects using the example of the Humanitarian OpenStreetMap Team, a Wikipedia-style social mapping platform where the review process is closer to a teacher-learner model rather than a public peer review. We observe peer feedback for early contributions by 1,300 newcomers, and assess the impact of different classes of feedback, including performance feedback, corrective feedback, and verbal rewards. We find that verbal rewards and immediate feedback can have a powerful effect on newcomer retention. In order to better support such positive engagement effects, we recommend that system designers conceptually distinguish between mechanisms for quality control and for learner feedback.

How can civic technology be designed to encourage more public engagement? What new methods of data collection and sharing can be used to engender a different relationship between citizens and the state? One approach has been to design physical systems that draw people in and which they can trust, leading them to give their views, opinions or other data. So far, they have been largely used to elicit feedback or votes for one or two questions about a given topic. Here, we describe a physical system, called Sens-Us, which was designed to ask a range of questions about personal and sensitive information, within the context of rethinking the UK Census. An in-the-wild study of its deployment in a city cultural center showed how a diversity of people approached, answered and compared the data that had been collected about themselves with others. We discuss the findings in relation to the pros and cons of using this kind of innovative technology when wanting to promote civic engagement or other forms of public engagement.

Internet of things (IoT) devices and sensor kits have the potential to democratize the access, use, and appropriation of data. Despite the increased availability of low cost sensors, most of the produced data is "black box" in nature: users often do not know how to access or interpret data. We propose a "human-data design" approach in which end-users are given tools to create, share, and use data through tangible and physical visualizations. This paper introduces Physikit, a system designed to allow users to explore and engage with environmental data through physical ambient visualizations. We report on the design and implementation of Physikit, and present a two-week field study which showed that participants got an increased sense of the meaning of data, embellished and appropriated the basic visualizations to make them blend into their homes, and used the visualizations as a probe for community engagement and social behavior.

Organisers of large-scale crowdsourcing initiatives need to consider how to produce outcomes with their projects, but also how to build volunteer capacity. The initial project experience of contributors plays an important role in this, particularly when the contribution process requires some degree of expertise. We propose three analytical dimensions to assess first-time contributor engagement based on readily available public data: cohort analysis, task analysis, and observation of contributor performance. We apply these to a large-scale study of remote mapping activities coordinated by the Humanitarian OpenStreetMap Team, a global volunteer effort with thousands of contributors. Our study shows that different coordination practices can have a marked impact on contributor retention, and that complex task designs can be a deterrent for certain contributor groups. We close by providing recommendations about how to build and sustain volunteer capacity in these and comparable crowdsourcing systems.

Context: Categorising software systems according to their functionality yields many benefits to both users and developers. Goal: In order to uncover the latent clustering of mobile apps in app stores, we propose a novel technique that measures app similarity based on claimed behaviour. Method: Features are extracted using information retrieval augmented with ontological analysis and used as attributes to characterise apps. These attributes are then used to cluster the apps using agglomerative hierarchical clustering. We empirically evaluate our approach on 17,877 apps mined from the BlackBerry and Google app stores in 2014. Results: The results show that our approach dramatically improves the existing categorisation quality for both Blackberry (from 0.02 to 0.41 on average) and Google (from 0.03 to 0.21 on average) stores. We also find a strong Spearman rank correlation (ρ= 0.96 for Google and ρ= 0.99 for BlackBerry) between the number of apps and the ideal granularity within each category, indicating that ideal granularity increases with category size, as expected. Conclusions: Current categorisation in the app stores studied do not exhibit a good classification quality in terms of the claimed feature space. However, a better quality can be achieved using a good feature extraction technique and a traditional clustering method.

What is the relationship between urban form and citizens' well-being? In this paper, we propose a quantitative approach to help answer this question, inspired by theories developed within the fields of architecture and population health. The method extracts a rich set of metrics of urban form and well-being from openly accessible datasets. Using linear regression analysis, we identify a model which can explain 30% of the variance of well-being when applied to Greater London, UK. Outcomes of this research can inform the discussion on how to design cities which foster the well-being of their residents.

Gathering opinions from young children is challenging and different methods have been explored. In this paper we investigated how tangible devices can be used to gather feedback from children in the context of a theater performance. We introduce SmallTalk, a tangible survey system designed for use within a theater space to capture what children, aged 4 to 9, thought of a live performance they had just seen. We describe how the system was designed to build on previous feedback methods that had been tried; while at the same time meeting the constraints of the challenging theater context. We present results from seven deployments of SmallTalk and based on these we briefly discuss its value as a method for evaluating the theater performance. We then look at how the results validated the system design and present several design implications that more generally relate to tangible feedback systems for children.

Sharing economy platforms have become extremely popular in the last few years, and they have changed the way in which we commute, travel, and borrow among many other activities. Despite their popularity among consumers, such companies are poorly regulated. For example, Airbnb, one of the most successful examples of sharing economy platform, is often criticized by regulators and policy makers. While, in theory, municipalities should regulate the emergence of Airbnb through evidence-based policy making, in practice, they engage in a false dichotomy: some municipalities allow the business without imposing any regulation, while others ban it altogether. That is because there is no evidence upon which to draft policies. Here we propose to gather evidence from the Web. After crawling Airbnb data for the entire city of London, we find out where and when Airbnb listings are offered and, by matching such listing information with census and hotel data, we determine the socio-economic conditions of the areas that actually benefit from the hospitality platform. The reality is more nuanced than one would expect, and it has changed over the years. Airbnb demand and offering have changed over time, and traditional regulations have not been able to respond to those changes. That is why, finally, we rely on our data analysis to envision regulations that are responsive to real-time demands, contributing to the emerging idea of ``algorithmic regulation''.

Within the remit of `Data for Development' there have been a number of promising recent works that investigate the use of mobile phone Call Detail Records (CDRs) to estimate the spatial distribution of poverty or socio-economic status. The methods being developed have the potential to offer immense value to organisations and agencies who currently struggle to identify the poorest parts of a country, due to the lack of reliable and up to date survey data in certain parts of the world. However, the results of this research have thus far only been presented in isolation rather than in comparison to any alternative approach or benchmark. Consequently, the true practical value of these methods remains unknown. Here, we seek to allay this shortcoming, by proposing two baseline poverty estimators grounded on concrete usage scenarios: one that exploits correlation with population density only, to be used when no poverty data exists at all; and one that also exploits spatial autocorrelation, to be used when poverty data has been collected for a few regions within a country. We then compare the predictive performance of these baseline models with models that also include features derived from CDRs, so to establish their real added value. We present extensive analysis of the performance of all these models on data acquired for two developing countries -- Senegal and Ivory Coast. Our results reveal that CDR-based models do provide more accurate estimates in most cases; however, the improvement is modest and more significant when estimating (extreme) poverty intensity rates rather than mean wealth.

Many companies would like to redesign their workspaces to make them more pleasant and even fun places to work in. An assumption is it will result in social and economic benefits. However, it can be difficult to achieve because of cost, level of disruption and regulations. We present an alternative approach that provides an injection of playfulness into "drab" office buildings. A lightweight technology intervention was designed - Mood Squeezer - that asks people to reflect on their mood by squeezing a colored ball from a box set. The squeezes are mirrored back as an aggregate colorful visualization on a public floor display. An in-the-wild study showed how this intervention was successful at getting people to squeeze their mood, leading to a diversity of conversations throughout the building. We discuss how this lightweight approach to office augmentation can provide new opportunities for opening up a "closed" workplace.

Crowd-sourcing has become a popular form of computer mediated collaborative work and OpenStreetMap represents one of the most successful crowd-sourcing systems, where the goal of building and maintaining an accurate global map of the world is being accomplished by means of contributions made by over 1.2M citizens. However, within this apparently large crowd, a tiny group of highly active users is responsible for the mapping of almost all the content. One may thus wonder to what extent the information being mapped is biased towards the interests and agenda of this group of users. In this paper, we present a method to quantitatively measure content bias in crowd-sourced geographic information. We then apply the method to quantify content bias across a three-year period of OpenStreetMap mapping in 40 countries. We find almost no content bias in terms of what is being mapped, but significant geographic bias; furthermore, we find that bias in terms of meticulousness varies with culture.

Measuring socioeconomic deprivation of cities in an accurate and timely fashion has become a priority for governments around the world, as the massive urbanization process we are witnessing is causing high levels of inequalities which require intervention. Traditionally, deprivation indexes have been derived from census data, which is however very expensive to obtain, and thus acquired only every few years. Alternative computational methods have been proposed in recent years to automatically extract proxies of deprivation at a fine spatio-temporal level of granularity; however, they usually require access to datasets (e.g., call details records) that are not publicly available to governments and agencies. To remedy this, we propose a new method to automatically mine deprivation at a fine level of spatio-temporal granularity that only requires access to freely available user-generated content. More precisely, the method needs access to datasets describing what urban elements are present in the physical environment; examples of such datasets are Foursquare and OpenStreetMap. Using these datasets, we quantitatively describe neighborhoods by means of a metric, called Offering Advantage, that reflects which urban elements are distinctive features of each neighborhood. We then use that metric to (i) build accurate classifiers of urban deprivation and (ii) interpret the outcomes through thematic analysis. We apply the method to three UK urban areas of different scale and elaborate on the results in terms of precision and recall.

Organizers regularly want to understand the experiences of event goers and typically use survey methods, with researchers and clipboards. However, gathering opinions in such ways is difficult to do without disrupting the event goers' experience. In place of clipboard surveys, we developed a quite different form of tangible questionnaire, called VoxBox, which uses physical interactions to transform feedback giving into a playful and engaging experience that fits much more with the event itself. Here we question if such a device can successfully draw a diverse representation of event attendees to voice relevant opinions during the event. We describe an observational study of VoxBox based on two real-world deployments, and present findings on (1) the experiences VoxBox provides to facilitators and users; and (2) its capabilities as a means for opinion gathering. We conclude by discussing lessons learned, design implications, and the wide potential for tangible questionnaires in other application areas.

Gathering public opinions, such as surveys, at events typically requires approaching people in situ, but this can disrupt the positive experience they are having and can result in very low response rates. As an alternative approach, we present the design and implementation of VoxBox, a tangible system for gathering opinions on a range of topics in situ at an event through playful and engaging interaction. We discuss the design principles we employed in the creation of VoxBox and show how they encouraged wider participation, by grouping similar questions, encouraging completion, gathering answers to open and closed questions, and connecting answers and results. We evaluate these principles through observations from an initial deployment and discuss how successfully these were implemented in the design of VoxBox.

Governments and other organisations often rely on data collected by household surveys and censuses to identify areas in most need of regeneration and development projects. However, due to the high cost associated with the data collection process, many developing countries conduct such surveys very infrequently and include only a rather small sample of the population, thus failing to accurately capture the current socio-economic status of the country's population. In this paper, we address this problem by means of a methodology that relies on an alternative source of data from which to derive up to date poverty indicators, at a very fine level of spatio-temporal granularity. Taking two developing countries as examples, we show how to analyse the aggregated call detail records of mobile phone subscribers and extract features that are strongly correlated with poverty indexes currently derived from census data.

Crowd-mapping is a form of collaborative work that empowers citizens to collect and share geographic knowledge. OpenStreetMap (OSM) is a successful example of such paradigm, where the goal of building and maintaining an accurate global map of the changing world is being accomplished by means of local contributions made by over 1.2M citizens. While OSM has been subject to many country-specific studies, the relationship between national culture and economic affluence and users' participation has been so far unexplored. In this work, we systematically study the link between them: we characterise OSM users in terms of who they are, how they contribute, during what period of time, and across what geographic areas. We find strong correlations between these characteristics and national culture factors (e.g., power distance, individualism, pace of life, self expression), and well as Gross Domestic Product per capita. Based on these findings, we discuss design issues that developers of crowd-mapping services should consider to account for cross-cultural differences.

Public transport information systems have been shown to positively affect passengers' usage of their city's transport infrastructure, by providing information such as the location and schedules of trains and buses. These systems, however, lack qualitative information about passengers' ongoing experi

Urban crowd-sourcing has become a popular paradigm to harvest spatial information about our evolving cities directly from citizens. OpenStreetMap is a successful example of such paradigm, with an accuracy of its user-generated content comparable to that of curated databases (e.g., Ordnance Survey). Coverage is however low and most importantly non-uniformly distributed across the city. Being able to model the spontaneous growth of digital information in these domains is required, so to be able to plan interventions aimed at gathering content about areas that would otherwise be neglected. Inspired by models of physical urban growth developed by urban planners, we build a model of digital growth of crowd-sourced spatial information that is both easy to interpret and dynamic, so to be able to determine what factors impact growth and how these change over time. We build and test the model against five years of OpenStreetMap data for the city of London, UK. We then run the model against two other cities, chosen for their different physical and digital growth's characteristics, so to stress-test the model. We conclude with a discussion of the implications of this work on both developers and users of urban crowd-sourcing applications.

No abstract available.

Urbanization's rapid progress has modernized many people's lives but also engendered big issues, such as traffic congestion, energy consumption, and pollution. Urban computing aims to tackle these issues by using the data that has been generated in cities (e.g., traffic flow, human mobility, and geographical data). Urban computing connects urban sensing, data management, data analytics, and service providing into a recurrent process for an unobtrusive and continuous improvement of people's lives, city operation systems, and the environment. Urban computing is an interdisciplinary field where computer sciences meet conventional city-related fields, like transportation, civil engineering, environment, economy, ecology, and sociology in the context of urban spaces. This article first introduces the concept of urban computing, discussing its general framework and key challenges from the perspective of computer sciences. Second, we classify the applications of urban computing into seven categories, consisting of urban planning, transportation, the environment, energy, social, economy, and public safety and security, presenting representative scenarios in each category. Third, we summarize the typical technologies that are needed in urban computing into four folds, which are about urban sensing, urban data management, knowledge fusion across heterogeneous data, and urban data visualization. Finally, we give an outlook on the future of urban computing, suggesting a few research topics that are somehow missing in the community.

While the prediction accuracy of a large-scale recommender system can generally be improved by learning from more and more training data over time, it is unclear how well a fixed predictive model can handle the changing business dynamics in a real-world scenario. The adjustment of a predictive model is controlled by the hyperparameter settings of a selected algorithm. Although the problem of hyperparameter optimization has been studied for decades in various disciplines, the adaptiveness of the initially selected model is not as well understood. This paper presents an approach to continuously re-select hyperparameter settings of the algorithm in a large-scale retail recommender system. In particular, an automatic hyperparameter optimization technique is applied on collaborative filtering algorithms in order to improve prediction accuracy. Experiments have been conducted on a large-scale real retail dataset to challenge traditional assumption that a one-off initial hyperparameter optimization is sufficient. The proposed approach has been compared with a baseline approach and a widely used approach with two scalable collaborative filtering algorithms. The evaluations of our experiments are based on a 2-year real purchase transaction dataset of a large retail chain business, both its online e-commerce site and its offline retail stores. It is demonstrated that continuous hyperparameter optimization can effectively improve the prediction accuracy of a recommender system. This paper presents a new direction in improving the prediction performance of a large-scale recommender system.

Ubiquitous crowd-sourcing has become a popular mechanism to harvest knowledge from the masses. OpenStreetMap (OSM) is a successful example of ubiquitous crowd-sourcing, where citizens volunteer geographic information in order to build and maintain an accurate map of the changing world. Research has shown that OSM information is accurate, by comparing it with centrally maintained spatial information such as Ordnance Survey. However, we find that coverage is low and non uniformly distributed, thus challenging the suitability of ubiquitous crowd-sourcing as a mechanism to map the whole world. In this paper, we investigate what contextual factors correlate with coverage of OSM information in urban settings. We find that, although there is a direct correlation between population density and information coverage, other socio-economic factors also play an important role. We discuss the implications of these findings with respect to the design of urban crowd-sourcing applications.

A common metaphor to describe the movement of people within a city is that of blood flowing through the veins of a living organism. We often speak of the 'pulse of the city' when referring to flow patterns we observe. Here we extend this metaphor by hypothesising that by monitoring the flow of people through a city we can assess the city's health, as a nurse takes a patient's heart-rate and blood pressure during a routine health check. Using an automated fare collection dataset of journeys made on the London rail system, we build a classification model that identifies areas of high deprivation as measured by the Indices of Multiple Deprivation, and achieve a precision, sensitivity and specificity of 0.805, 0.733 and 0.810, respectively. We conclude with a discussion of the potential benefits this work provides to city planning, policymaking, and citizen engagement initiatives.

Policy makers are calling for new socio-economic measures that reflect subjective well-being, to complement traditional measures of material welfare as the Gross Domestic Product (GDP). Self-reporting has been found to be reasonably accurate in measuring one's well-being and conveniently tallies with sentiment expressed on social media (e.g., those satisfied with life use more positive than negative words in their Facebook status updates). Social media content can thus be used to track well-being of individuals. A question left unexplored is whether such content can be used to track well-being of entire physical communities as well. To this end, we consider Twitter users based in a variety of London census communities, and study the relationship between sentiment expressed in tweets and community socio-economic well-being. We find that the two are highly correlated: the higher the normalized sentiment score of a community's tweets, the higher the community's socio-economic well-being. This suggests that monitoring tweets is an effective way of tracking community well-being too.

Social (or folksonomic) tagging has become a very popular way to describe content within Web 2.0 websites. However, as tags are informally defined, continually changing, and ungoverned, it has often been criticised for lowering, rather than increasing, the efficiency of searching. To address this issue, a variety of approaches have been proposed that recommend users what tags to use, both when labeling and when looking for resources. These techniques work well in dense folksonomies, but they fail to do so when tag usage exhibits a power law distribution, as it often happens in real-life folksonomies. To tackle this issue, we propose an approach that induces the creation of a dense folksonomy, in a fully automatic and transparent way: when users label resources, an innovative tag similarity metric is deployed, so to enrich the chosen tag set with related tags already present in the folksonomy. The proposed metric, which represents the core of our approach, is based on the mutual reinforcement principle. Our experimental evaluation proves that the accuracy and coverage of searches guaranteed by our metric are higher than those achieved by applying classical metrics.

The widespread adoption of automated fare collection (AFC) systems by public transport authorities around the world means that, increasingly, people carry and use passive sensors (embedded inside of public transit tickets) to record their daily movements. Unlike mobile phones, the records held by AFC systems provide a rich and detailed source of data about peoples' transport habits: times of travel, modalities, destinations, trip durations, and fares paid. In this work, we explore the extent that this data offers the possibility to both build and measure future of travel-based ubiquitous computing applications. We focus on two potential end-users: first, how travellers may be aided by feedback mechanisms in order to re-align misperceptions of their travel behaviour and leverage this data to change their habits. In particular, we analyse differences between 85 travellers' surveyed perceptions of their public transport habits and their actual usage of the system. Second, how transport authorities can use this data to measure and implement incentive mechanisms that produce the expected impact. We use anonymised AFC data to measure the extent that financial incentives implemented by London's transport authority (such as peak-hour fares and student discounts) correlate with measurable changes in millions of travellers' behaviours.

As the public transport infrastructure of large cities expands, transport operators are diversifying the range and prices of tickets that can be purchased for travel. However, selecting the best fare for each individual traveller's needs is a complex process that is left almost completely unaided. By examining the relation between urban mobility and fare purchasing habits in large datasets from London, England's public transport network, we estimate that travellers in the city cumulatively spend, per year, up to approximately GBP 200 million more than they need to, as a result of purchasing the incorrect fares. We propose to address these incorrect purchases by leveraging the huge volumes of data that travellers create as they move about the city, by providing, to each of them, personalised ticket recommendations based on their estimated future travel patterns. In this work, we explore the viability of building a fare-recommendation system for public transport networks by (a) formalising the problem as two separate prediction problems and (b) evaluating a number of algorithms that aim to match travellers to the best fare. We find that applying data mining techniques to public transport data has the potential to provide travellers with substantial savings.

The rise of location-enabled mobile phones and location based services offers a great opportunity to apply personalization and recommender system technology to people's everyday lives. A variety of digital traces can now be used to infer how people move about their city and extract their context and habits. Personalization and recommender systems, potentially merged with the data that people store online (e.g., social networks, web ratings), can then not only be used to recommend new places and events that they may find interesting to attend, but, more broadly, personalize and enhance any service that people find themselves using.

Delay Tolerant Networking (DTN) protocols have been investigated as effective ways to distribute content in scenarios where the producers and consumers of content belong to the same geographical area. Common focus of all these approaches has been on maximising network delivery while treating messages as if they were all worth the same to recipients (i.e., messages are forwarded in a first-encountered/first-forwarded fashion). However, because of battery limitations on mobile devices, it is often the case that not all messages can be delivered. In this paper, we propose a new approach for priority-scheduling in participatory DTNs, whereby both message's value to the end-user as well as the likelihood of future encounters are combined to make the forwarding decision.

Traveller information, route planning, and service updates have become essential components of public transport systems: they help people navigate built environments by providing access to information regarding delays and service disruptions. However, one aspect that these systems lack is a way of tailoring the information they offer in order to provide personalised trip time estimates and relevant notifications to each traveller. Mining each user's travel history, collected by automated ticketing systems, has the potential to address this gap. In this work, we analyse one such dataset of travel history on the London underground. We then propose and evaluate methods to (a) predict personalised trip times for the system users and (b) rank stations based on future mobility patterns, in order to identify the subset of stations that are of greatest interest to the user and thus provide useful travel updates.

Thanks to advances in mobile technology, modern mobile devices have become essential companions, assisting their users in attaining their daily tasks. It will not be long before these devices will become recommending companions, advising users about what data (e.g., restaurants) and what services (e.g., podcast channels) they may enjoy in the local area at the present time. Because of the very nature of the items (both data and services) being suggested (i.e., location dependent and mobile with respect to the consuming user), recommendations cannot be computed on central servers and then pushed to the users. Rather, a novel decentralised mobile recommender service will have to be developed and deployed; instead of relying on global knowledge about users' profiles, such service will have to exploit the wisdom of local communities to compute recommendations. Moreover, because of resource limitations of mobile devices, the algorithms it will employ will have to be computationally light. In this paper, we propose diffeRS, a totally decentralised mobile recommender service specifically designed for pervasive environments. diffeRS crafts a virtual view of the local community's preferences, by exchanging users' profiles via radio technology (e.g., Bluetooth) during periods of colocation. Profiles are stored locally and recommendations are computed using a lightweight algorithm. As our experimental evaluations demonstrate, diffeRS achieves an accuracy and coverage that are comparable to those of centralized recommender systems in use today.

Collaborative Filtering (CF) algorithms, used to build web-based recommender systems, are often evaluated in terms of how accurately they predict user ratings. However, current evaluation techniques disregard the fact that users continue to rate items over time: the temporal characteristics of the system's top-N recommendations are not investigated. In particular, there is no means of measuring the extent that the same items are being recommended to users over and over again. In this work, we show that temporal diversity is an important facet of recommender systems, by showing how CF data changes over time and performing a user survey. We then evaluate three CF algorithms from the point of view of the diversity in the sequence of recommendation lists they produce over time. We examine how a number of characteristics of user rating patterns (including profile size and time between rating) affect diversity. We then propose and evaluate set methods that maximise temporal recommendation diversity without extensively penalising accuracy.

Digital content production and distribution has radically changed our business models. An unprecedented volume of supply is now on offer, whetted by the demand of millions of users from all over the world. Since users cannot be expected to browse through millions of different items to find what they might like, filtering has become a popular technique to connect supply and demand: trusted users are first identified, and their opinions are then used to create recommendations. In this domain, users' trustworthiness has been measured according to one of the following two criteria: taste similarity (i.e., “I trust those who agree with me”), or social ties (i.e., “I trust my friends, and the people that my friends trust”). The former criterion aims at identifying concordant users, but is subject to abuse by malicious behaviors. The latter aims at detecting well-intentioned users, but fails to capture the natural subjectivity of tastes. In this article, we propose a new definition of trusted recommenders , addressing those users that are both well-intentioned and concordant. Based on this characterisation, we propose a novel approach to information filtering that we call dependable filtering . We describe alternative algorithms realizing this approach, and demonstrate, by means of extensive performance evaluation on a variety of real large-scale datasets, the high degree of both accuracy and robustness they entail.

We propose FriendSensing, a framework that automatically suggests friends to mobile social-networking users. Using short-range technologies (e.g., Bluetooth) on her mobile phone, a social-networking user "senses" and keeps track of other phones in her proximity. FriendSensing processes proximity records using a variety of algorithms that are based on social network theories of geographical proximity and of link prediction. It then returns a personalized and automatically generated list of people the user may know. We evaluate the extent to which FriendSensing helps users find people they know against real mobility and social network data.

Collaborative Filtering aims to predict user tastes, by minimising the mean error produced when predicting hidden user ratings. The aim of a deployed recommender system is to iteratively predict users' preferences over a dynamic, growing dataset, and system administrators are confronted with the problem of having to continuously tune the parameters calibrating their CF algorithm. In this work, we formalise CF as a time-dependent, iterative prediction problem. We then perform a temporal analysis of the Netflix dataset, and evaluate the temporal performance of two CF algorithms. We show that, due to the dynamic nature of the data, certain prediction methods that improve prediction accuracy on the Netflix probe set do not show similar improvements over a set of iterative train-test experiments with growing data. We then address the problem of parameter selection and update, and propose a method to automatically assign and update per-user neighbourhood sizes that (on the temporal scale) outperforms setting global parameters.

This paper proposes Habit, an efficient multi-layered approach to content dissemination in Delay Tolerant Networks (DTN) that leverages information about nodes' colocation (physical layer) and their social network (application layer). More precisely, the regularity of users' colocation is learned based on historical colocation observations; also, the users' social network (or `network of interest') is dynamically propagated during periods of colocation; finally, these distinct pieces of information are locally combined and used to compute the paths that content should follow, in a way that maximises both precision (i.e., nodes receive only content they are interested in) and recall (i.e., all relevant content is received by interested nodes).

In wireless ad hoc networks, nodes are both routers and terminals, and they have to cooperate to communicate. Cooperation at the network layer means routing (finding a path for a packet), and forwarding (relaying packets for others). However, because wireless nodes are usually constrained by limited power and computational resources, a selfish node may be unwilling to spend its resources in forwarding packets that are not of its direct interest, even though it expects other nodes to forward its packets to the destination. In this paper, we propose a game-theoretic model to facilitate the study of the non-cooperative behaviors in wireless ad hoc networks and analyze incentive schemes to motivate cooperation among wireless ad hoc network nodes to achieve a mutually beneficial networking result.

Service providers as we know them nowadays are the always-on static web service providers, that aim at Five9 availability (99.999%). Formal, or de-facto, standards, such as WSDL and BPEL, have become technology enablers for the easy discovery, use and coordination of such services. However, we envisage tomorrow's services to become increasingly pervasive, being deployed within buildings, transport systems, markets, as well as people portable devices. Such services will be, by their own nature, simple and fine grained; as a consequence, service composition will become crucial to deliver rich functionalities that satisfy end-users requests. Composing services in mob.ile environments opens up significant challenges. In particular, the Five9 availability assumption no longer holds: the higher the dynamic nature of the environment, the higher the chances that services will move out-of-reach before the composition completes, causing the service as a whole to fail. We argue that, in order to enable the successful completion of compound services, the reliability of the composition must be measured and reasoned about. In order to do so, we propose to dynamically deploy a prediction model to estimate the duration of colocation between componentservices. These estimates are fed in input to a service composition semantics reasoner, which then autonomically selects those providers, within the current environment, that maximise the chances of successful compound service completion. We demonstrate the positive impact that the reliability reasoning has onto the ratio of successfully completed compound services in a typical human movement scenario.

To share services, portable devices may need to locate reputable in-range providers and, to do so, they may exchange ratings with each other. However, providers may well tweak ratings to their own advantage. That is why we have designed a new decentralized mechanism (dubbed MobiRate) with which portable devices store ratings in (local) tamperevident tables and check the integrity of those tables through a gossiping protocol. We evaluate the extent to which MobiRate reduces the impact of tampered ratings and consequently locates reputable service providers. We do so using real mobility and social network data. We also assess computational and communication costs of MobiRate on mobile phones.

People living in urban areas spend a considerable amount of time on public transport, for example, commuting to/from work. During these periods, opportunities for inter-personal networking present themselves, as many members of the public now carry electronic devices equipped with Bluetooth or other wireless technology. Using these devices, individuals can share content (e.g., music, news and video clips) with fellow travellers that are on the same train or bus. Transferring media content takes time; in order to maximise the chances of successful downloads, users should identify neighbours that possess desirable content and who will travel with them for long-enough periods. In this paper, we propose a user-centric prediction scheme that collects historical colocation information to determine the best content sources. The scheme works on the assumption that people have a high degree of regularity in their movements. We first validate this assumption on a real dataset, that consists of traces of people moving in a large city's mass transit system. We then demonstrate experimentally on these traces that our prediction scheme significantly improves communication efficiency, when compared to a memory(history)-less source selection scheme.

Recommender systems, based on collaborative filtering, draw their strength from techniques that manipulate a set of user-rating profiles in order to compute predicted ratings of unrated items. There are a wide range of techniques that can be applied to this problem; however, the k-nearest neighbour (kNN) algorithm has become the dominant method used in this context. Much research to date has focused on improving the performance of this algorithm, without considering the properties that emerge from manipulating the user data in this way. In order to understand the effect of kNN on a user-rating dataset, the algorithm can be viewed as a process that generates a graph, where nodes are users and edges connect similar users: the algorithm generates an implicit social network amongst the system subscribers. Temporal updates of the recommender system will impose changes on the graph. In this work we analyse user-user kNN graphs from a temporal perspective, retrieving characteristics such as dataset growth, the evolution of similarity between pairs of users, the volatility of user neighbourhoods over time, and emergent properties of the entire graph as the algorithm parameters change. These insights explain why certain kNN parameters and similarity measures outperform others, and show that there is a surprising degree of structural similarity between these graphs and explicit user social networks.

Social (or folksonomic) tagging has become a very popular way to describe, categorise, search, discover and navigate content within Web 2.0 websites. Unlike taxonomies, which overimpose a hierarchical categorisation of content, folksonomies empower end users by enabling them to freely create and choose the categories (in this case, tags) that best describe some content. However, as tags are informally defined, continually changing, and ungoverned, social tagging has often been criticised for lowering, rather than increasing, the efficiency of searching, due to the number of synonyms, homonyms, polysemy, as well as the heterogeneity of users and the noise they introduce. In this paper, we propose Social Ranking, a method that exploits recommender system techniques to increase the efficiency of searches within Web 2.0. We measure users' similarity based on their past tag activity. We infer tags' relationships based on their association to content. We then propose a mechanism to answer a user's query that ranks (recommends) content based on the inferred semantic distance of the query to the tags associated to such content, weighted by the similarity of the querying user to the users who created those tags. A thorough evaluation conducted on the CiteULike dataset demonstrates that Social Ranking neatly improves coverage, while not compromising on accuracy.

Using mobile devices, such as smart phones, people may create and distribute different types of digital content (e.g., photos, videos). One of the problems is that digital content, being easy to create and replicate, may likely swamp users rather than informing them. To avoid that, users may organize content producers that they know and trust in a web of trust. Users may then reason about this web of trust to form opinions about content producers with whom they have never interacted before. These opinions will then determine whether content is accepted. The process of forming opinions is called trust propagation. We design a mechanism for mobile devices that effectively propagates trust and that is lightweight and distributed (as opposed to previous work that focuses on centralized propagation). This mechanism uses a graph-based learning technique. We evaluate the effectiveness (predictive accuracy) of this mechanism against a large real-world data set. We also evaluate the computational cost of a J2ME implementation on a mobile phone.

Large scale market penetration of electronic devices equipped with Bluetooth technology now gives the ability to share content (such as music or video clips) between members of the public in a decentralised manner. Achieved using opportunistic connections, formed when they are colocated, in environments where Internet connectivity is expensive or unreliable, such as urban buses, train rides and coffee shops. Most people have a high degree of regularity in their movements (such as a daily commute), including repeated contacts with others possessing similar seasonal movement patterns. We argue that this behaviour can be exploited in connection selection, and outline a system for the identification of long-term companions and sources that have previously provided quality content, in order to maximise the successful receipt of content flies. We utilise actual traces and existing mobility models to validate our approach, and show how consideration of the colocation history and the quality of previous data transfers leads to more successful sharing of content in realistic scenarios.

Handheld devices have become sufficiently powerful that it is easy to create, disseminate, and access digital content (e.g., photos, videos) using them. The volume of such content is growing rapidly and, from the perspective of each user, selecting relevant content is key. To this end, each user may run a trust model - a software agent that keeps track of who disseminates content that its user finds relevant. This agent does so by assigning an initial trust value to each producer for a specific category (context); then, whenever it receives new content, the agent rates the content and accordingly updates its trust value for the producer in the content category. However, a problem with such an approach is that, as the number of content categories increases, so does the number of trust values to be initially set. This paper focuses on how to effectively set initial trust values. The most sophisticated of the current solutions employ predefined context ontologies, using which initial trust in a given context is set based on that already held in similar contexts. However, universally accepted (and time invariant) ontologies are rarely found in practice. For this reason, we propose a mechanism called TRULLO (trust bootstrapping by latently lifting context) that assigns initial trust values based only on local information (on the ratings of its user's past experiences) and that, as such, does not rely on third-party recommendations. We evaluate the effectiveness of TRULLO by simulating its use in an informal antique market setting. We also evaluate the computational cost of a J2ME implementation of TRULLO on a mobile phone.

Collaborative filtering has become an established method to measure users' similarity and to make predictions about their interests. However, prediction accuracy comes at the cost of user's privacy: in order to derive accurate similarity measures, users are required to share their rating history with each other. In this work we propose a new measure of similarity, which achieves comparable prediction accuracy to the Pearson correlation coefficient, and that can successfully be estimated without breaking users' privacy. This novel method works by estimating the number of concordant, discordant and tied pairs of ratings between two users with respect to a shared random set of ratings. In doing so, neither the items rated nor the ratings themselves are disclosed, thus achieving strictly-private collaborative filtering. The technique has been evaluated using the recently released Netflix prize dataset.

In recent years, various trust management models based on the human notion of trust have been proposed to support trust-aware decision making in pervasive systems. However, the degree of subjectivity embedded in human trust often clashes with the requirements imposed by the target scenario: on one hand, pervasive computing calls for autonomic and light-weight systems that impose minimum burden on the user of the device (and on the device itself); on the other hand, computational models of human trust seem to demand a large amount of user input and physical resources. The result is often a computational trust model that does not 'compute': either the degree of subjectivity it offers is limited, or its complexity compromises its usability. In this paper, we present an accurate and efficient trust prediction model that is based on a basic Kalman filter. We discuss simulation results to demonstrate that the predictor is capable of capturing the natural disposition to trust of the user of the device, while being autonomic and light-weight.

Rapid advances in wireless networking technologies have enabled mobile devices to be connected anywhere and anytime. While roaming, applications on these devices dynamically discover hosts and services with whom interactions can be started. However, the fear of exposure to risky transactions with potentially unknown entities may seriously hinder collaboration. To minimise this risk, an engineering approach to the development of trust-based collaborations is necessary. This paper introduces hTrust, a human trust management model and framework that facilitates the construction of trust-aware mobile systems and applications. In particular, hTrust supports: reasoning about trust (trust formation), dissemination of trust information in the network (trust dissemination), and derivation of new trust relationships from previously formed ones (trust evolution). The framework views each mobile host as a self-contained unit, carrying along a portfolio of credentials that are used to prove its trustworthiness to other hosts in an ad-hoc mobile environment. Customising functions are defined to capture the natural disposition to trust of the user of the device inside our trust management framework.

Mobile devices, such as mobile phones and personal digital assistants, have gained wide-spread popularity. These devices will increasingly be networked, thus enabling the construction of distributed applications that have to adapt to changes in context, such as variations in network bandwidth, battery power, connectivity, reachability of services and hosts, etc. In this paper, we describe CARISMA, a mobile computing middleware which exploits the principle of reflection to enhance the construction of adaptive and context-aware mobile applications. The middleware provides software engineers with primitives to describe how context changes should be handled using policies. These policies may conflict. We classify the different types of conflicts that may arise in mobile computing and argue that conflicts cannot be resolved statically at the time applications are designed, but, rather, need to be resolved at execution time. We demonstrate a method by which policy conflicts can be handled; this method uses a microeconomic approach that relies on a particular type of sealed-bid auction. We describe how this method is implemented in the CARISMA middleware architecture and sketch a distributed context-aware application for mobile devices to illustrate how the method works in practice. We show, by way of a systematic performance evaluation, that conflict resolution does not imply undue overheads, before comparing our research to related work and concluding the paper.

No abstract available.

No abstract available.

Mobile devices, such as mobile phones and personal digital assistants, have gained wide-spread popularity. These devices will increasingly be networked, thus enabling the construction of distributed mobile applications. These have to adapt to changes in context, such as variations in network bandwidth, exhaustion of battery power or reachability of services on other devices. We show how the construction of adaptive and context-aware mobile applications can be supported using a reflective middleware. The middleware provides software engineers with primitives to describe how context changes are handled using policies. These policies may conflict. In this paper, we classify the different types of conflicts that may arise in mobile computing. We argue that conflicts cannot be resolved statically at the time applications are designed, but, rather, need to be resolved at execution time. We demonstrate a method by which these policy conflicts can be treated. This method uses a micro-economic approach that relies on a particular type of sealed-bid auction.

xlinkit is a lightweight application service that provides rule-based link generation and checks the consistency of distributed Web content. It leverages standard Internet technologies, notably XML, XPath, and XLink. xlinkit can be used as part of a consistency management scheme or in applications that require smart link generation, including portal construction and management of large document repositories. In this article we show how consistency constraints can be expressed and checked. We describe a novel semantics for first-order logic that produces links instead of truth values and give an account of our content management strategy. We present the architecture of our service and the results of two substantial case studies that use xlinkit for checking course syllabus information and for validating UML models supplied by industrial partners.

Summary form only given. Middleware solutions for wired distributed systems cannot be used in a mobile setting, as mobile applications impose new requirements that run counter to the principle of transparency on which current middleware systems have been built. We propose the use of reflection capabilities and meta-data to pave the way for a new generation of middleware platforms designed to support mobility.

An increasing number of distributed applications will be written for mobile hosts, such as laptop computers, third generation mobile phones, personal digital assistants, watches and the like, with focus on peer-to-peer collaboration. Application engineers have to deal with a new set of problems caused by mobility, such as low bandwidth, context changes or loss of connectivity. During disconnection, independently from each others, users will typically update local replicas of shared data, possibly generated by peers. The resulting inconsistent replicas need to be reconciled upon re-connection. To support building mobile applications that use both replication and reconciliation over ad-hoc networks, we have designed XMIDDLE, a peer-to-peer middleware that targets mobile computing settings. In this paper we describe XMIDDLE and show how reflection capabilities are used to allow application engineers to influence replication and reconciliation techniques. XMIDDLE enables the transparent sharing of XML documents across heterogeneous mobile peers, allowing online and off-line access to data.