Search for: All records

Smart speakers are becoming ubiquitous in daily life. The widespread and increasing use of smart speakers for news and information in society presents new questions related to the quality, source diversity and credibility, and reliability of algorithmic intermediaries for news consumption. While user adoption rates soar, audit instruments for assessing information quality in smart speakers are lagging. As an initial effort, we present a conceptual framework and data-driven approach for evaluating smart speakers for information quality. We demonstrate the application of our framework on the Amazon Alexa voice assistant and identify key information provenance and source credibility problems as well as systematic differences in the quality of responses about hard and soft news. Our study has broad implications for news media and society, content production, and information quality assessment. 

Civil wars are as frequent and debilitating now as ever. More often than not, their resolution consists of the negotiation of a peace accord that involves a number of provisions. Although previous work in political science indicates an underlying interdependence between provision implementation sequences, it is unclear how the structure and dynamics of this interdependence relate to the successful implementation of peace accords. To fill this gap, we systematically study peace processimplementationactivityfrom34peaceaccordscontaining 51 provisions negotiated between 1989 and 2015. We begin by constructing a bipartite network between peace accords and their provisions implementation and explore statistical properties of the structural underpinnings of peace processes. Then, we examine motifs (i.e., significantly frequent patterns) in provision implementation activity and uncover higher order correlations between provisions.Finally, we identify provision implementation sequences (i.e., meta-groups) that are most strongly associated with successful peace processes. Our empirical findings provide new insights for the implementation of peace accords by revealing temporal sequences of peace process implementation that help build confidence, enhance security, and ultimately prevent negative cascading effects in different stages of the peacebuilding process. 

Texture-based features computed on eye movement scan paths have recently been proposed for eye movement biometric applications. Feature vectors were extracted within this prior work by computing the mean and standard deviation of the resulting images obtained through application of a Gabor filter bank. This paper describes preliminary work exploring an alternative technique for extracting features from Gabor filtered scan path images. Namely, features vectors are obtained by downsampling the filtered images, thereby retaining structured spatial information within the feature vector. The proposed technique is validated at various downsampling scales for data collected from 94 subjects during free-viewing of a fantasy movie trailer. The approach is demonstrated to reduce EER versus the previously proposed statistical summary technique by 11.7% for the best evaluated downsampling parameter. 

Crowd financing is a burgeoning phenomenon that promises to improve access to capital by enabling borrowers with limited financial opportunities to receive small contributions from individual lenders towards unsecured loan requests. Faced with information asymmetry about borrowers' credibility, individual lenders bear the entire loss in case of loan default. Predicting loan payment is therefore crucial for lenders and for the sustainability of these platforms. To this end, we examine whether the ''wisdom'' of the lending crowd can provide reliable decision support with respect to projects' long-term success. Using data from Prosper.com, we investigate the association between the dynamics of lending behaviour and successful loan payment through interpretable classification models. We find evidence for collective intelligence signals in lending behaviour and observe variability in crowd wisdom across loan categories. We find that the wisdom of the lending crowd is most prominent in the auto loan category, but it is statistically significant for all other categories except student debt. Our study contributes new insights on how signals deduced from lending behaviour can improve the efficiency of crowd financing thereby contributing to economic growth and societal development. 

Photosensor oculography (PSOG) is a promising solution for reducing the computational requirements of eye tracking sensors in wireless virtual and augmented reality platforms. This paper proposes a novel machine learning-based solution for addressing the known performance degradation of PSOG devices in the presence of sensor shifts. Namely, we introduce a convolutional neural network model capable of providing shift-robust end-to-end gaze estimates from the PSOG array output. Moreover, we propose a transfer-learning strategy for reducing model training time. Using a simulated workflow with improved realism, we show that the proposed convolutional model offers improved accuracy over a previously considered multilayer perceptron approach. In addition, we demonstrate that the transfer of initialization weights from pre-trained models can substantially reduce training time for new users. In the end, we provide the discussion regarding the design trade-offs between accuracy, training time, and power consumption among the considered models.