CAESAR: context-aware explanation based on supervised attention for service recommendations

Explainable recommendations have drawn more attention from both academia and industry recently, because they can help users better understand recommendations (i.e., why some particular items are recommended), therefore improving the persuasiveness of the recommender system and users’ satisfaction. However, little work has been done to provide explanations from the angle of a user’s contextual situations (e.g., companion, season, and destination if the recommendation is a hotel). To fill this research gap, we propose a new context-aware recommendation algorithm based on supervised attention mechanism (CAESAR), which particularly matches latent features to explicit contextual features as mined from user-generated reviews for producing context-aware explanations. Experimental results on two large datasets in hotel and restaurant service domains demonstrate that our model improves recommendation performance against the state-of-the-art methods and furthermore is able to return feature-level explanations that can adapt to the target user’s current contexts.

     

Designs for explaining intelligent agents

Explanation is an important capability for usable intelligent systems, including intelligent agents and cognitive models embedded within simulations and other decision support systems. Explanation facilities help users understand how and why an intelligent system possesses a given structure and set of behaviors. Prior research has resulted in a number of approaches to provide explanation capabilities and identified some significant challenges. We describe designs that can be reused to create intelligent agents capable of explaining themselves. The designs include ways to provide ontological, mechanistic, and operational explanations. These designs inscribe lessons learned from prior research and provide guidance for incorporating explanation facilities into intelligent systems. The designs are derived from both prior research on explanation tool design and from the empirical study reported here on the questions users ask when working with an intelligent system. We demonstrate the use of these designs through examples implemented using the Herbal high-level cognitive modeling language. These designs can help build better agents—they support creating more usable and more affordable intelligent agents by encapsulating prior knowledge about how to generate explanations in concise representations that can be instantiated or adapted by agent developers.     

Reflection-through-performance: personal implications of documenting health behaviors for the collective

Previous work has examined how technology can support health behavior monitoring in social contexts. These tools incentivize behavior documentation through the promise of virtual rewards, rich visualizations, and improved co-management of disease. Social influence is leveraged to motivate improved behaviors through friendly competition and the sharing of emotional and informational support. Prior work has described how by documenting and sharing behaviors in these tools, people engage in performances of the self. This performance happens as users selectively determine what information to share and hide, crafting a particular portrayal of their identity. Much of the prior work in this area has examined the implications of systems that encourage people to share their behaviors with friends, family, and geographically distributed strangers. In this paper, we report upon the performative nature of behavior sharing in a system created for a different social group: the local neighborhood. We designed Community Mosaic (CM), a system with a collectivistic focus: this tool asks users to document their behaviors using photographs and text, but not for their own benefit—for the benefit of others in their community. Through a 6-week deployment of CM, we evaluated the nature of behavior sharing in this system, including participants’ motivations for sharing, the way in which this sharing happened, and the reflexive impact of sharing. Our findings highlight the performative aspects of photograph staging and textual narration and how sharing this content led participants to become more aware and evaluative of their behaviors, and led them to try to eat more healthfully. We conclude with recommendations for behavior monitoring tools, specifically examining the implications of users’ perceived audience and automated behavioral tracking on opportunities for reflection-through-performance.     

Entertainment recommender systems for group of users

Recommender systems are used to recommend potentially interesting items to users in different domains. Nowadays, there is a wide range of domains in which there is a need to offer recommendations to group of users instead of individual users. As a consequence, there is also a need to address the preferences of individual members of a group of users so as to provide suggestions for groups as a whole. Group recommender systems present a whole set of new challenges within the field of recommender systems. In this article, we present two expert recommender systems that suggest entertainment to groups of users. These systems, jMusicGroupRecommender and jMoviesGroupRecommender, suggest music and movies and utilize different methods for the generation of group recommendations: merging recommendations made for individuals, aggregation of individuals’ ratings, and construction of group preference models. We also describe the results obtained when comparing different group recommendation techniques in both domains.     

Efficiently using contextual influence to recommend new items to ephemeral groups

Group recommender systems suggest items to groups of users that want to utilize those items together. These systems can support several activities that can be performed together with other people and are typically social, like watching TV or going to the restaurant. In this paper we study ephemeral groups, i.e., groups constituted by users who are together for the first time, and for which therefore there is no history of past group activities.Recent works have studied ephemeral group recommendations proposing techniques that learn complex models of users and items. These techniques, however, are not appropriate to recommend items that are new in the system, while we propose a method able to deal with new items too. Specifically, our technique determines the preference of a group for a given item by combining the individual preferences of the group members on the basis of their contextual influence, the contextual influence representing the ability of an individual, in a given situation, to guide the group’s decision. Moreover, while many works on recommendations do not consider the problem of efficiently producing recommendation lists at runtime, in this paper we speed up the recommendation process by applying techniques conceived for the top-$K$ query processing problem. Finally, we present extensive experiments, evaluating: (i) the accuracy of the recommendations, using a real TV dataset containing a log of viewings performed by real groups, and (ii) the efficiency of the online recommendation task, exploiting also a bigger partially synthetic dataset.     

Crowds,citizens and sensors: process and practice for mobilising learning

Participatory sensing is an emerging field in which citizens are empowered by technologies to monitor their own environments. Harvesting and analysing data gathered in response to personal or local enquiries can be seen as an antidote to information provided by official sources. Democratising sensing means that ordinary people can learn about and understand the world around them better and can be a part of the decision-making in improving environments for all. In this paper, we review and describe participatory sensing and discuss this in relation to making a series of prototype tools and applications for mobile users—Located Lexicon, Where’s Fenton? and Tall Buildings. In the first of these projects, Located Lexicon, we wanted to find out whether a lexicon of terms derived from user-generated content could enable the formation of Twitter like groups that allow users to engage in finding out more about their location. In the second project, Where’s Fenton? we made a publicly available app that involves users in counting the abundance and logging the location of deer in a park. This project focused specifically on anonymity of the user in collecting data for a specific enquiry. In the last project, Tall Buildings, we experimented with using dimensions of altitude, distance and speed to encourage users to physically explore a city from its rooftops. In all of these projects, we experiment with the pedestrian as a human sensor and the methods and roles they may engage in to make new discoveries. The underlying premise for our work is that it is not possible to calibrate people to be identical, so experimenting with crowd-sourced data opens up thinking about the way we observe and learn about the physical environment.     

Argument-based mixed recommenders and their application to movie suggestion

Recommender systems have become prevalent in recent years as they help users to access relevant items from the vast universe of possibilities available these days. Most existing research in this area is based purely on quantitative aspects such as indices of popularity or measures of similarity between items or users. This work introduces a novel perspective on movie recommendation that combines a basic quantitative method with a qualitative approach, resulting in a family of mixed character recommender systems. The proposed framework incorporates the use of arguments in favor or against recommendations to determine if a suggestion should be presented or not to a user. In order to accomplish this, Defeasible Logic Programming (DeLP) is adopted as the underlying formalism to model facts and rules about the recommendation domain and to compute the argumentation process. This approach has a number features that could be proven useful in recommendation settings. In particular, recommendations can account for several different aspects (e.g., the cast, the genre or the rating of a movie), considering them all together through a dialectical analysis. Moreover, the approach can stem for both content-based or collaborative filtering techniques, or mix them in any arbitrary way. Most importantly, explanations supporting each recommendation can be provided in a way that can be easily understood by the user, by means of the computed arguments. In this work the proposed approach is evaluated obtaining very positive results. This suggests a great opportunity to exploit the benefits of transparent explanations and justifications in recommendations, sometimes unrealized by quantitative methods.     

A unified framework for recommending items,groups and friends in social media environment via mutual resource fusion

Up to now, more and more online sites have started to allow their users to build the social relationships. Take the Last.fm for example (which is a popular music-sharing site), users can not only add each other as friends, but also join online interest groups where they shall meet people with common tastes. Therefore, in this environment, users might be interested in not only receiving item recommendations (such as music), but also getting friend suggestions so they might put them in the contact list, and group recommendations that they could consider joining. To support such demanding needs, in this paper, we propose a unified framework that provides three different types of recommendation in a single system: recommending items, recommending groups and recommending friends. For each type of recommendation, we in depth investigate the contribution of fusing other two auxiliary information resources (e.g., fusing friendship and membership for recommending items, and fusing user-item preferences and friendship for recommending groups) for boosting the algorithm performance. More notably, the algorithms were developed based on the matrix factorization framework in order to achieve the ideal efficiency as well as accuracy. We performed experiments with two large-scale real-world data sets that contain users’ implicit interaction with items. The results revealed the effective fusion mechanism for each type of recommendation in such implicit data condition. Moreover, it demonstrates the respective merits of regularization model and factorization model: the factorization is more suitable for fusing bipartite data (such as membership and user-item preferences), while the regularization model better suits one mode data (like friendship). We further enhanced the friendship’s regularization by integrating the similarity measure, which was experimentally proven with positive effect.     

A generalized taxonomy of explanations styles for traditional and social recommender systems

Recommender systems usually provide explanations of their recommendations to better help users to choose products, activities or even friends. Up until now, the type of an explanation style was considered in accordance to the recommender system that employed it. This relation was one-to-one, meaning that for each different recommender systems category, there was a different explanation style category. However, this kind of one-to-one correspondence can be considered as over-simplistic and non generalizable. In contrast, we consider three fundamental resources that can be used in an explanation: users, items and features and any combination of them. In this survey, we define (i) the Human style of explanation, which provides explanations based on similar users, (ii) the Item style of explanation, which is based on choices made by a user on similar items and (iii) the Feature style of explanation, which explains the recommendation based on item features rated by the user beforehand. By using any combination of the aforementioned styles we can also define the Hybrid style of explanation. We demonstrate how these styles are put into practice, by presenting recommender systems that employ them. Moreover, since there is inadequate research in the impact of social web in contemporary recommender systems and their explanation styles, we study new emerged social recommender systems i.e. Facebook Connect explanations (HuffPo, Netflix, etc.) and geo-social explanations that combine geographical with social data (Gowalla, Facebook Places, etc.). Finally, we summarize the results of three different user studies, to support that Hybrid is the most effective explanation style, since it incorporates all other styles.     

Folkommender: a group recommender system based on a graph-based ranking algorithm

With the rapid popularity of smart devices, users are easily and conveniently accessing rich multimedia content. Consequentially, the increasing need for recommender services, from both individual users and groups of users, has arisen. In this paper, we present a new graph-based approach to a recommender system, called Folkommender, that can make recommendations most notably to groups of users. From rating information, we first model a signed graph that contains both positive and negative links between users and items. On this graph we examine two distinct random walks to separately quantify the degree to which a group of users would like or dislike items. We then employ a differential ranking approach for tailoring recommendations to the group. Our empirical evaluations on two real-world datasets demonstrate that the proposed group recommendation method performs better than existing alternatives. We also demonstrate the feasibility of Folkommender for smartphones.     

kNN processing with co-space distance in SoLoMo systems

With the increasing popularity of smart phones, SoLoMo (Social-Location-Mobile) systems are expected to be fast-growing and become a popular mobile social networking platform. A main challenge in such systems is on the creation of stable links between users. For each online user, the current SoLoMo system continuously returns his/her kNN (k Nearest Neighbor) users based on their geo-locations. Such a recommendation approach is simple, but fails to create sustainable friendships. Instead, it would be more effective to tap onto the existing social relationships in conventional social networks, such as Facebook and Twitter, to provide a “better” friend recommendations.To measure the similarity between users, we propose a new metric, co-space distance, by considering both the user distances in the real world (physical distance) and the virtual world (social distance). The co-space distance measures the similarity of two users in the SoLoMo system. We compute the social distances between users based on their public information in the conventional social networks, which can be achieved by a few MapReduce jobs. To facilitate efficient computation of the social distance, we build a distributed index on top of the key-value store, and maintain the users’ geo-locations using an R-tree. For each query on finding potential friends around a location, we return kNN neighbors to each user based on their co-space distances. We propose a progressive top-k processing strategy and an adaptive-caching strategy to facilitate efficient query processing. Experiments with Gowalla dataset¹ show the effectiveness and efficiency of our recommendation approach.     

Exploring Patients’ Use Intention of Personal Health Record Systems: Implications for Design

Abstract

Personal health record (PHR) systems offer a technology for personal health information management (PHIM) activities. Despite efforts to increase the use of PHR systems as a mechanism to support better patient-centered care and improve information management across the continuum of care, PHR adoption remains low. The purpose of this study was to explore how to design a PHR system that can adequately support personal health information management activities. Using a mixed-methods approach (questionnaires and interviews), we identified the factors affecting a person’s intention to use PHRs and also described the personal health information management activities among people from a wide age range in the United States. Results indicated that the intention to use PHR systems was affected by system-related factors, such as perceived usefulness, health information understandability, personalization, and patient–clinician communication support, and user-related factors, such as social influence, self-efficacy, and willingness to share. Furthermore, five types of personal health information management activities were found, including storage, organization, maintenance, retrieval, and sharing. Informed by the study findings, we developed seven design recommendations to improve PHR systems. Future studies can focus on further validating these findings using other methods and be based on larger and more representative PHR users.     

A time-aware spatio-textual recommender system

Location-Based Social Networks (LBSNs) allow users to post ratings and reviews and to notify friends of these posts. Several models have been proposed for Point-of-Interest (POI) recommendation that use explicit (i.e. ratings, comments) or implicit (i.e. statistical scores, views, and user influence) information. However the models so far fail to capture sufficiently user preferences as they change spatially and temporally. We argue that time is a crucial factor because user check-in behavior might be periodic and time dependent, e.g. check-in near work in the mornings and check-in close to home in the evenings. In this paper, we present two novel unified models that provide review and POI recommendations and consider simultaneously the spatial, textual and temporal factors. In particular, the first model provides review recommendations by incorporating into the same unified framework the spatial influence of the users’ reviews and the textual influence of the reviews. The second model provides POI recommendations by combining the spatial influence of the users’ check-in history and the social influence of the users’ reviews into another unified framework. Furthermore, for both models we consider the temporal dimension and measure the impact of time on various time intervals. We evaluate the performance of our models against 10 other methods in terms of precision and recall. The results indicate that our models outperform the other methods.     

Mood modeling: accuracy depends on active logging and reflection

Current behavior change systems often demand extremely advanced sensemaking skills, requiring users to interpret personal datasets in order to understand and change behavior. We describe EmotiCal, a system to help people better manage their emotions, that finesses such complex sensemaking by directly recommending specific mood-boosting behaviors to users. This paper first describes how we develop the accurate mood models that underlie these mood-boosting recommendations. We go on to analyze what types of information contribute most to the predictive power of such models, and how we might design systems to reliably collect such predictive information. Our results show that we can derive very accurate mood models with relatively small samples of just 70 users. These models explain 61% of variance by combining: (a) user reflection about the effects of different activities on mood, (b) user explanations of how different activities affect mood, and (c) individual differences. We discuss the implications of these findings for the design of behavior change systems, as well as for theory and practice. Contrary to many recent approaches, our findings argue for the importance of active user reflection rather than passive sensing.     

Trust-inspiring explanation interfaces for recommender systems

A recommender system’s ability to establish trust with users and convince them of its recommendations, such as which camera or PC to purchase, is a crucial design factor especially for e-commerce environments. This observation led us to build a trust model for recommender agents with a focus on the agent’s trustworthiness as derived from the user’s perception of its competence and especially its ability to explain the recommended results. We present in this article new results of our work in developing design principles and algorithms for constructing explanation interfaces. We show the effectiveness of these principles via a significant-scale user study in which we compared an interface developed based on these principles with a traditional one. The new interface, called the organization interface where results are grouped according to their tradeoff properties, is shown to be significantly more effective in building user trust than the traditional approach. Users perceive it more capable and efficient in assisting them to make decisions, and they are more likely to return to the interface. We therefore recommend designers to build trust-inspiring interfaces due to their high likelihood to increase users’ intention to save cognitive effort and the intention to return to the recommender system.     

Human factors and information technology: responding to the challenge

Editor's note: Recommendations are given from a report published by the UK National Electronics Council on the importance of an awareness of human factors to information technology. Ergonomists alone cannot reshape attitudes to technological design, according to the report, and a concerted effort is needed from every group involved. For this reason, the NEC's recommendations are addressed to government, standards bodies, manufacturers, users and educators. Taken together, the recommendations from the basis of the initiative that the report's authors believe must be taken to meet the challenge posed by information technology. Although the report was published in the context of the UK Alvey ‘Advanced Information Technology’ programme, its content will be of interest in all countries working with information technology. The report represents a challenge from the human factors community to engineers, and as such is required reading for Displays subscribers. The journal will welcome responses and encourage debate on the subject.     

Preference dynamics with multimodal user-item interactions in social media recommendation

Recommender systems elicit the interests and preferences of individuals and make recommendations accordingly, a main challenge for expert and intelligent systems. An essential problem in recommender systems is to learn users’ preference dynamics, that is, the constant evolution of the explicit or the implicit information, which is diversified throughout time according to the user actions. Also, in real settings data sparsity degrades the recommendation accuracy. Hence, state-of-the-art methods exploit multimodal information of users-item interactions to reduce sparsity, but they ignore preference dynamics and do not capture users’ most recent preferences. In this article, we present a Temporal Collective Matrix Factorization (TCMF) model, making the following contributions: (i) we capture preference dynamics through a joint decomposition model that extracts the user temporal patterns, and (ii) co-factorize the temporal patterns with multimodal user-item interactions by minimizing a joint objective function to generate the recommendations. We evaluate the performance of TCMF in terms of accuracy and root mean square error, and show that the proposed model significantly outperforms state-of-the-art strategies.     

Black-Box Trace&Revoke Codes

We address the problem of designing an efficient broadcast encryption scheme which is also capable of tracing traitors. We introduce a code framework to formalize the problem. Then, we give a probabilistic construction of a code which supports both traceability and revocation. Given N users with at most r revoked users and at most t traitors, our code construction gives rise to a Trace&Revoke system with private keys of size O((r+t)logN) (which can also be reduced to constant size based on an additional computational assumption), ciphertexts of size O((r+t)logN), and O(1) decryption time. Our scheme can deal with certain classes of pirate decoders, which we believe are sufficiently powerful to capture practical pirate strategies. In particular, our code construction is based on a combinatorial object called (r,s)-disjunct matrix, which is designed to capture both the classic traceability notion of disjunct matrix and the new requirement of revocation capability. We then probabilistically construct (r,s)-disjunct matrices which help design efficient Black-Box Trace&Revoke systems. For dealing with “smart” pirates, we introduce a tracing technique called “shadow group testing” that uses (close to) legitimate broadcast signals for tracing. Along the way, we also proved several bounds on the number of queries needed for black-box tracing under different assumptions about the pirate’s strategies.     

An interaction model between human and system for intuitive graphical search interface

Our objective was to propose a new model which provides real interactions like human conversations. In this paper, we defined “interaction” to mean not only superficial interactions between human and systems but also internal elements inspiring one another. We proposed a new interaction model by defining four user elements namely user knowledge, information needs, thinking, and  feelings, and five system elements namely system knowledge, interaction algorithm, knowledge base, retrieval algorithm, and database. The key point is that users can understand inside the systems gradually and operate them flexibly in their own way to provide real interactions where users and systems inspire one another’s internal elements. We then defined system requirements to realize this model so that users can change and comprehend system knowledge and that users interact with the system constantly. We constructed an image retrieval system applying our proposed graphical search interface named Concentric Ring View and confirmed that all system requirements were satisfied. In a usability test with 12 university students, we confirmed that the proposed interaction model provided intuitive searches to users by inspiring internal elements between users and systems. Users could continuously change and comprehend system knowledge, synchronize user knowledge, shifting thinking and feeling, and changing information needs.     

On integrating confidentiality and functionality in a formal method

This paper proposes a formal method, based on Circus, for developing software systems that respect a joint specification of functionality and confidentiality attributes. We extend the semantics of Circus to capture the information that users can infer about a system’s behaviour, enabling confidentiality and functionality attributes of a system to be specified together. We represent inconsistencies between functionality and confidentiality properties as miracles, rendering insecure functionality infeasible. We present techniques for verifying that a system design’s functionality and confidentiality attributes are mutually consistent, and for ensuring that consistency is maintained by refinement steps.