John O'Donovan

Trust in recommender systems

Recommender systems have proven to be an important response to the information overload problem, by providing users with more proactive and personalized information services. And collaborative filtering techniques have proven to be an vital component of many such recommender systems as they facilitate the generation of high-quality recom-mendations by leveraging the preferences of communities of similar users. In this paper we suggest that the traditional emphasis on user similarity may be overstated. We argue that additional factors have an important role to play in guiding recommendation. Specifically we propose that the trustworthiness of users must be an important consideration. We present two computational models of trust and show how they can be readily incorporated into standard collaborative filtering frameworks in a variety of ways. We also show how these trust models can lead to improved predictive accuracy during recommendation.

TasteWeights: a visual interactive hybrid recommender system

This paper presents an interactive hybrid recommendation system that generates item predictions from multiple social and semantic web resources, such as Wikipedia, Facebook, and Twitter. The system employs hybrid techniques from traditional recommender system literature, in addition to a novel interactive interface which serves to explain the recommendation process and elicit preferences from the end user. We present an evaluation that compares different interactive and non-interactive hybrid strategies for computing recommendations across diverse social and semantic web APIs. Results of the study indicate that explanation and interaction with a visual representation of the hybrid system increase user satisfaction and relevance of predicted content.

PeerChooser: visual interactive recommendation

Collaborative filtering (CF) has been successfully deployed over the years to compute predictions on items based on a users correlation with a set of peers. The black-box nature of most CF applications leave the user wondering how the system arrived at its recommendation. This note introduces PeerChooser, a collaborative recommender system with an interactive graphical explanation interface. Users are provided with a visual explanation of the CF process and opportunity to manipulate their neighborhood at varying levels of granularity to reflect aspects of their current requirements. In this manner we overcome the problem of redundant profile information in CF systems, in addition to providing an explanation interface. Our layout algorithm produces an exact, noiseless graph representation of the underlying correlations between users. PeerChoosers prediction component uses this graph directly to yield the same results as the benchmark. Users then improve on these predictions by tweaking the graph to their current requirements. We present a user-survey in which PeerChooser compares favorably against a benchmark CF algorithm.

Smallworlds: visualizing social recommendations

We present SmallWorlds, a visual interactive graph‐based interface that allows users to specify, refine and build item‐preference profiles in a variety of domains. The interface facilitates expressions of taste through simple graph interactions and these preferences are used to compute personalized, fully transparent item recommendations for a target user. Predictions are based on a collaborative analysis of preference data from a users direct peer group on a social network. We find that in addition to receiving transparent and accurate item recommendations, users also learn a wealth of information about the preferences of their peers through interaction with our visualization. Such information is not easily discoverable in traditional text based interfaces. A detailed analysis of our design choices for visual layout, interaction and prediction techniques is presented. Our evaluations discuss results from a user study in which SmallWorlds was deployed as an interactive recommender system on Facebook.

Is trust robust?: an analysis of trust-based recommendation

Systems that adapt to input from users are susceptible to attacks from those same users. Recommender systems are common targets for such attacks since there are financial, political and many other motivations for influencing the promotion or demotion of recommendable items [2].Recent research has shown that incorporating trust and reputation models into the recommendation process can have a positive impact on the accuracy and robustness of recommendations. In this paper we examine the effect of using five different trust models in the recommendation process on the robustness of collaborative filtering in an attack situation. In our analysis we also consider the quality and accuracy of recommendations. Our results caution that including trust models in recommendation can either reduce or increase prediction shift for an attacked item depending on the model-building process used, while highlighting approaches that appear to be more robust.

Understanding Information Credibility on Twitter

Increased popularity of microblogs in recent years brings about a need for better mechanisms to extract credible or otherwise useful information from noisy and large data. While there are a great number of studies that introduce methods to find credible data, there is no accepted credibility benchmark. As a result, it is hard to compare different studies and generalize from their findings. In this paper, we argue for a methodology for making such studies more useful to the research community. First, the underlying ground truth values of credibility must be reliable. The specific constructs used to define credibility must be carefully defined. Secondly, the underlying network context must be quantified and documented. To illustrate these two points, we conduct a unique credibility study of two different data sets on the same topic, but with different network characteristics. We also conduct two different user surveys, and construct two additional indicators of credibility based on retweet behavior. Through a detailed statistical study, we first show that survey based methods can be extremely noisy and results may vary greatly from survey to survey. However, by combining such methods with retweet behavior, we can incorporate two signals that are noisy but uncorrelated, resulting in ground truth measures that can be predicted with high accuracy and are stable across different data sets and survey methods. Newsworthiness of tweets can be a useful frame for specific applications, but it is not necessary for achieving reliable credibility ground truth measurements.

An analysis of topical proximity in the twitter social graph

Standard approaches of information retrieval are increasingly complemented by social search even when it comes to rational information needs. Twitter, as a popular source of real-time information, plays an important role in this respect, as both the follower-followee graph and the many relationships among users provide a rich set of information pieces about the social network. However, many hidden factors must be considered if social data are to successfully support the search for high-quality information. Here we focus on one of these factors, namely the relationship between content similarity and social distance in the social network. We compared two methods to compute content similarity among twitter users in a one-per-user document collection, one based on standard term frequency vectors, the other based on topic associations obtained by Latent Dirichlet Allocation (LDA). By comparing these metrics at different hop distances in the social graph we investigated the utility of prominent features such as Retweets and Hashtags as predictors of similarity, and demonstrated the advantages of topical proximity vs. textual similarity for friend recommendations.

Moodplay: Interactive Mood-based Music Discovery and Recommendation

A large body of research in recommender systems focuses on optimizing prediction and ranking. However, recent work has highlighted the importance of other aspects of the recommendations, including transparency, control and user experience in general. Building on these aspects, we introduce MoodPlay, a hybrid recommender system music which integrates content and mood-based filtering in an interactive interface. We show how MoodPlay allows the user to explore a music collection by latent affective dimensions, and we explain how to integrate user input at recommendation time with predictions based on a pre-existing user profile. Results of a user study (N=240) are discussed, with four conditions being evaluated with varying degrees of visualization, interaction and control. Results show that visualization and interaction in a latent space improve acceptance and understanding of both metadata and item recommendations. However, too much of either can result in cognitive overload and a negative impact on user experience.

WiGis: a framework for scalable web-based interactive graph visualizations

Traditional network visualization tools inherently suffer from scalability problems, particularly when such tools are interactive and web-based. In this paper we introduce WiGis –Web-based Interactive Graph Visualizations. WiGis exemplify a fully web-based framework for visualizing large-scale graphs natively in a users browser at interactive frame rates with no discernible associated startup costs. We demonstrate fast, interactive graph animations for up to hundreds of thousands of nodes in a browser through the use of asynchronous data and image transfer. Empirical evaluations show that our system outperforms traditional web-based graph visualization tools by at least an order of magnitude in terms of scalability, while maintaining fast, high-quality interaction.

Effects of Information Availability on Command-and-Control Decision Making Performance, Trust, and Situation Awareness

Objective: We investigated how increases in task-relevant information affect human decision-making performance, situation awareness (SA), and trust in a simulated command-and-control (C2) environment. Background: Increased information is often associated with an improvement of SA and decision-making performance in networked organizations. However, previous research suggests that increasing information without considering the task relevance and the presentation can impair performance. Method: We used a simulated C2 task across two experiments. Experiment 1 varied the information volume provided to individual participants and measured the speed and accuracy of decision making for task performance. Experiment 2 varied information volume and information reliability provided to two participants acting in different roles and assessed decision-making performance, SA, and trust between the paired participants. Results: In both experiments, increased task-relevant information volume did not improve task performance. In Experiment 2, increased task-relevant information volume reduced self-reported SA and trust, and incorrect source reliability information led to poorer task performance and SA. Conclusion: These results indicate that increasing the volume of information, even when it is accurate and task relevant, is not necessarily beneficial to decision-making performance. Moreover, it may even be detrimental to SA and trust among team members. Application: Given the high volume of available and shared information and the safety-critical and time-sensitive nature of many decisions, these results have implications for training and system design in C2 domains. To avoid decrements to SA, interpersonal trust, and decision-making performance, information presentation within C2 systems must reflect human cognitive processing limits and capabilities.