The MovieLens datasets are widely used in education, research, and industry. They downloaded hundreds of thousands times each year, reflecting their use popular press programming books, traditional online courses, software. These a product member activity the movie recommendation system, an active research platform that has hosted many experiments since its launch 1997. This article documents history datasets. We include discussion lessons learned from running long-standing, live perspective...

We design a field experiment to explore the use of social comparison increase contributions an online community. find that, after receiving behavioral information about median user's total number movie ratings, users below demonstrate 530 percent in monthly while those above decrease their ratings by 62 percent. When given outcome average net benefit score, above-average mainly engage activities that help others. Our findings suggest effective personalized can level public goods provision....

Question and answer (Q&A) sites such as Yahoo! Answers are places where users ask questions others them. In this paper, we investigate predictors of quality through a comparative, controlled field study responses provided across several online Q&A sites. Along with quantitative results concerning the effects factors question topic rhetorical strategy, present two high-level messages. First, you get what pay for in Answer was typically higher Google (a fee-based site) than free studied,...

A tagging community's vocabulary of tags forms the basis for social navigation and shared expression.We present a user-centric model evolution in communities based on community influence personal tendency. We evaluate our an emergent system by introducing features into MovieLens recommender system.We explore four tag selection algorithms displaying applied other members. analyze 'effect evolution, utility, adoption, user satisfaction.

Eli Pariser coined the term 'filter bubble' to describe potential for online personalization effectively isolate people from a diversity of viewpoints or content. Online recommender systems - built on algorithms that attempt predict which items users will most enjoy consuming are one family technologies potentially suffers this effect. Because have become so prevalent, it is important investigate their impact in these terms. This paper examines longitudinal impacts collaborative...

Tens of thousands questions are asked and answered every day on social question answer (Q&A) Web sites such as Yahoo Answers. While these generate an enormous volume searchable data, the problem determining which answers archival quality has grown. One major component this is prevalence conversational questions, identified both by Q&A academic literature that intended simply to start discussion. For example, a "do you believe in evolution?" might successfully engage users discussion, but...

Increasingly, algorithms are used to make important decisions across society. However, these usually poorly understood, which can reduce transparency and evoke negative emotions. In this research, we seek learn design principles for explanation interfaces that communicate how decision-making work, in order help organizations explain their stakeholders, or support users' "right explanation". We conducted an online experiment where 199 participants different understand algorithm making...

Algorithmic decision-making systems are increasingly used throughout the public and private sectors to make important decisions or assist humans in making these with real social consequences. While there has been substantial research recent years build fair algorithms, less seeking understand factors that affect people's perceptions of fairness systems, which we argue is also for their broader acceptance. In this research, conduct an online experiment better fairness, focusing on three sets...

Recent developments in user evaluation of recommender systems have brought forth powerful new tools for understanding what makes recommendations effective and useful. We apply these methods to understand how users evaluate recommendation lists the purpose selecting an algorithm finding movies. This paper reports on experiment which we asked compare produced by three common collaborative filtering algorithms dimensions novelty, diversity, accuracy, satisfaction, degree personalization, select...

Community Question Answering (CQA) services enable their users to exchange knowledge in the form of questions and answers. These communities thrive as a result small number highly active users, typically called experts , who provide large high-quality useful Expert identification techniques community managers take measures retain community. There is further value identifying during first few weeks participation it would allow nurture them. In this article we address two problems: (a) How...

Explanations are important for users to make decisions on whether take recommendations. However, algorithm generated explanations can be overly simplistic and unconvincing. We believe that humans overcome these limitations. Inspired by how people explain word-of-mouth recommendations, we designed a process, combining crowdsourcing computation, generates personalized natural language explanations. modeled key topical aspects of movies, asked crowdworkers write based quotes from online movie...

The essence of a recommender system is that it can recommend items personalized to the preferences an individual user. But typically users are given no explicit control over this personalization, and instead left guessing about how their actions affect resulting recommendations. We hypothesize any algorithm will better fit some users' expectations than others, leaving opportunities for improvement. To address challenge, we study puts in hands users. Specifically, build evaluate incorporates...

Recommender systems are not one-size-fits-all; different algorithms and data sources have strengths, making them a better or worse fit for users use cases. As one way of taking advantage the relative merits algorithms, we gave ability to change algorithm providing their movie recommendations studied how they make this power. We conducted our study with launch new version MovieLens recommender that supports multiple allows choose want provide recommendations. examine log from user...

As users browse a recommender system, they systematically consider or skip over much of the displayed content. It seems obvious that these eye gaze patterns contain rich signal concerning users' preferences. However, because tracking data is not available to most systems, signals are widely incorporated into personalization models. In this work, we show it possible predict by combining easily-collected user browsing with from small number in grid-based interface. Our technique able leverage...

Many online communities use tags - community selected words or phrases to help people find what they desire. The quality of varies widely, from that capture akey dimension an entity those are profane, useless, unintelligible. Tagging systems must often select a subset available display users due limited screen space. Because spread have seen, selecting good not only improves individual's view tags, it also encourages them create better in the future. We explore implicit (behavioral) and...

To achieve high quality initial personalization, recommender systems must provide an efficient and effective process for new users to express their preferences. We propose that this goal is best served not by the classical method where begin expressing preferences individual items - inefficient way convert a user's effort into improved personalization. Rather, we can groups of items. test idea designing evaluating interactive across are automatically generated clustering algorithms....

The technical barriers for conversing with recommender systems using natural language are vanishing. Already, there commercial that facilitate interactions an AI agent. For instance, it is possible to say "what should I watch" Apple TV remote get recommendations. In this research, we investigate how users initially interact a new deepen our understanding of the range inputs these technologies can expect. We deploy interface system, observe users' first and follow-up queries, measure...

Many small online communities would benefit from increased diversity or activity in their membership. Some run the risk of dying out due to lack participation. Others struggle achieve critical mass necessary for diverse and engaging conversation. But what tools are available these increase participation? Our goal this research was spark contributions movielens.org discussion forum, where only 2% members write posts. We developed personalized invitations, messages designed entice users visit...

Recommender systems algorithms are generally evaluated primarily on machine learning criteria such as recommendation accuracy or top-n precision. In this work, we evaluate six from a user-centric perspective, collecting both objective user activity data and subjective perceptions. field experiment involving 1508 users who participated for at least month, compare built using techniques, ranging supervised matrix factorization, contextual bandit to Q learning. We found that the design in...

Related item recommenders operate in the context of a particular item. For instance, music system's page about artist Radio-head might recommend other similar artists such as The Flaming Lips. Often central to these recommendations is computation similarity between pairs items. Prior work has explored many algorithms and features that allow for scores, but little evaluated approaches from user-centric perspective. In this work, we build evaluate six scoring span range activity- content-based...

Social question and answer (Q&A) Web sites field a remarkable variety of questions: while one user seeks highly technical information, another looks to start social exchange. Prior work in the has adopted informal taxonomies types as mechanism for interpreting behavior community outcomes. In this work, we contribute formal taxonomy deepen our understanding nature intent questions that are asked online. Our is grounded Aristotelian rhetorical theory, complemented by contributions leading...