Non-personalized Recommendations Research Articles

Accounting for Personalization in Personalization Algorithms: YouTube’s Treatment of Conspiracy Content

This article investigates under which video watch conditions YouTube’s recommender system tends to develop a preference for conspiracy-classified videos. Whereas existing research on so-called filter bubbles and rabbit holes tends to rely on non-personalized recommendations and on standard watch patterns, this study puts personalization and diversified user strategies at the center of its design. 20 authenticated bots have been instructed to watch YouTube content based on four distinct watch strategies. In a baseline strategy, bots watched non-conspiracy videos only. Treatment strategies involved watching conspiracy-classified content, selected based on either non-personalized, partly-personalized, or fully-personalized input. Bots watched a total of 15 videos, and after each video their top 20 homepage recommendations were collected and classified as either conspiracy-related or not. This allowed us to measure the impact of each video watched and of each watch strategy on the proportion of conspiracy-classified content recommended at each step. The same experiment has been reverted, exposing the treatment groups to non-conspiracy videos only, to assess the persistence of this pattern. Our results show that users primed with conspiracy-classified content tend to quickly receive a much larger proportion of conspiracy-classified recommendations. Inverting this pattern proves significantly more difficult than generating it. There are also indications that watch strategies relying on personalized content as input might produce stronger effects. This article contributes evidence to the argument that YouTube’s recommendation system is prone to generating strong, potentially pernicious recommendation patterns. Moreover, it contributes a replicable methodology that puts personalization at the center of the stage in the study of content personalization algorithms.

The effect of perceived usefulness of recommender systems and information sources on purchase intention

PurposeThe current study aims to empirically analyze the influence of different information sources, together with the persuasiveness of recommender systems (RSs) on the consumer’s purchase intention (PI). It also expands the research on RSs from the point of view of consumer behavior and psychology, considering perceived usefulness and relevance. In addition, it analyzes how different types of personalized recommendations, along with non-personalized ones, influence PI.Design/methodology/approachThe proposed model has been validated using partial least squares structural equation modeling (PLS-SEM), based on the data collected from 597 online shoppers.FindingsThis study proves that both information search and RSs influence PI, being complementary rather than mutually exclusive. Recommender systems’ findings indicate that the PI is primarily influenced by the perceived relevance of RSs, the information provided by manufacturers and reviews. Moreover, only the influence of the perceived usefulness of personalized recommendations strongly affects PI. Conversely, non-personalized recommendations do not affect PI.Practical implicationsDevelopers should focus on increasing the perceived usefulness and relevance of RSs. Thus, they could adopt the hybridization of RSs with the aggregation of both personal shopping behavior and social network contacts. It should integrate information signals from multiple sources to include sentiment extracted from reviews or links to the manufacturer’s page. Furthermore, the recommendation of discounted products must be only for products preferred by customers, because only these influence the PI.Originality/valueThis research provides a structural model that examines together, for the first time, the influence on the PI of the main RSs and sources of information.

Predicting Ecological Momentary Assessments in an App for Tinnitus by Learning From Each User's Stream With a Contextual Multi-Armed Bandit

Ecological Momentary Assessments (EMA) deliver insights on how patients perceive tinnitus at different times and how they are affected by it. Moving to the next level, an mHealth app can support users more directly by predicting a user's next EMA and recommending personalized services based on these predictions. In this study, we analyzed the data of 21 users who were exposed to an mHealth app with non-personalized recommendations, and we investigate ways of predicting the next vector of EMA answers. We studied the potential of entity-centric predictors that learn for each user separately and neighborhood-based predictors that learn for each user separately but take also similar users into account, and we compared them to a predictor that learns from all past EMA indiscriminately, without considering which user delivered which data, i.e., to a “global model.” Since users were exposed to two versions of the non-personalized recommendations app, we employed a Contextual Multi-Armed Bandit (CMAB), which chooses the best predictor for each user at each time point, taking each user's group into account. Our analysis showed that the combination of predictors into a CMAB achieves good performance throughout, since the global model was chosen at early time points and for users with few data, while the entity-centric, i.e., user-specific, predictors were used whenever the user had delivered enough data—the CMAB chose itself when the data were “enough.” This flexible setting delivered insights on how user behavior can be predicted for personalization, as well as insights on the specific mHealth data. Our main findings are that for EMA prediction the entity-centric predictors should be preferred over a user-insensitive global model and that the choice of EMA items should be further investigated because some items are answered more rarely than others. Albeit our CMAB-based prediction workflow is robust to differences in exposition and interaction intensity, experimentators that design studies with mHealth apps should be prepared to quantify and closely monitor differences in the intensity of user-app interaction, since users with many interactions may have a disproportionate influence on global models.

Personalized vs non-personalized recommendations: how recommender systems, recommendation sources and recommendation platforms affect trial of YouTube videos among digital natives in Saudi Arabia

Purpose This study aims to examine the difference between personalized and non-personalized recommendations in influencing YouTube users’ video choices. In addition, whether men and women have a significant difference in using recommendations was compared and the predictors of recommendation video use frequency were explored. Design/methodology/approach A survey of 524 Saudi Arabia college students was conducted using computer-assisted self-administered interviews to collect their video recommendation sources and how likely they follow the recommendation from different sources. Findings Video links posted on social media used by the digital natives were found as the most effective form of recommendation shows that social approval is important in influencing trials. Recommendations can succeed in both personalized and non-personalized ways. Personalized recommendations as in YouTube recommended videos are almost the same as friends and family’s non-personalized posting of video links on social media in convincing people to watch the videos. Contrary to expectations, Saudi men college students are more likely to use recommendations than women students. Research limitations/implications The use of a non-probability sample is a major limitation and self-reported frequency may result in over- or under-estimation of video use. Practical implications Marketers will realize that they may not need the personalized recommendation from the large site. They can use social media recommendations by the consumers’ friends and family. E-mail is the worst platform for a recommendation. Social implications Recommendation is a credible source and can overcome the avoidance of advertising. Its influence on consumers will be increasing in years to come with the algorithmic recommendation and social media use. Originality/value This is the first study to compare the influence of different online recommendation sources and compare personalized and non-personalized recommendations. As recommendation is growing more and more important with algorithm development online, the study results have high reference values to marketers in Islamic countries and beyond.

Tubes and bubbles topological confinement of YouTube recommendations.

The role of recommendation algorithms in online user confinement is at the heart of a fast-growing literature. Recent empirical studies generally suggest that filter bubbles may principally be observed in the case of explicit recommendation (based on user-declared preferences) rather than implicit recommendation (based on user activity). We focus on YouTube which has become a major online content provider but where confinement has until now been little-studied in a systematic manner. We aim to contribute to the above literature by showing whether recommendation on YouTube exhibits phenomena typical of filter bubbles, tending to lower the diversity of consumed content. Starting from a diverse number of seed videos, we first describe the properties of the sets of suggested videos in order to design a sound exploration protocol able to capture latent recommendation graphs recursively induced by these suggestions. These graphs form the background of potential user navigations along non-personalized recommendations. From there, be it in topological, topical or temporal terms, we show that the landscape of what we call mean-field YouTube recommendations is often prone to confinement dynamics. Moreover, the most confined recommendation graphs i.e., potential bubbles, seem to be organized around sets of videos that garner the highest audience and thus plausibly viewing time.

Where Do Local Experts Go? Evaluating User Geo-Topical Similarity for Top-N Place Recommendation

Recent years have witnessed the ever-growing popularity of location-based social network (LBSN) services. Top-N place recommendation, which aims at retrieving N most appealing places for a target user, has thus gained increasing importance. Yet existing solutions to this problem either provide non-personalized recommendations by selecting nearby popular places, or resort to collaborative filtering (CF) by treating each place as an independent item, overlooking the geographical and semantic correlations among places. In this paper, we propose GoTo, a collaborative recommender that provides top-N personalized place recommendation in LBSNs. Compared with existing methods, GoTo achieves its effectiveness by exploiting the wisdom of the so-called local experts, namely those who are geographically close and have similar preferences with regard to a certain user. At the core of GoTo lies a novel user similarity measure called geo-topical similarity, which combines geographical and semantic correlations among places for discovering local experts. In specific, the geo-topical similarity uses Gaussian mixtures to model users’ real-life geographical patterns, and extracts users’ topical preferences from the attached tags of historically visited places. Extensive experiments on real LBSN datasets show that compared with baseline methods, GoTo can improve the performance of top-N place recommendation by up to 50% in terms of accuracy.

ArchReco: a software tool to assist software design based on context aware recommendations of design patterns

This work describes the design, development and evaluation of a software Prototype, named ArchReco, an educational tool that employs two types of Context-aware Recommendations of Design Patterns, to support users (CS students or professionals) who want to improve their design skills when it comes to training for High Level Software models. The tool’s underlying algorithms take advantage of Semantic Web technologies, and the usage of Content based analysis for the computation of non-personalized recommendations for Design Patterns. The recommendations’ objective is to support users in functions such as finding the most suitable Design Pattern to use according to the working context, learn the meaning, objectives and usages of each Design Pattern. The current work presents the Semantic Modeling of the Software Design process through the definition of the context that defines the Software Design process and in particular the representation of the Design Patterns as Ontology model, the implemented Context Aware Recommendation Algorithms and the evaluation results extracted from a user based testing for the ArchReco prototype.