Crowdsourcing Quality Control of Online Information: A Quality-Based Cascade Model Abstract. We extend previous cascade models of social influence to investigate how the exchange of quality information among users may moderate cascade behavior, and the extent to which it may influence the effectiveness of collec- tive user recommendations on quality control of information. We found that while cascades do sometimes occur, their effects depend critically on the accu- racies of individual quality assessments of information contents. Contrary to predictions of cascade models of information flow, quality-based cascades tend to reinforce the propagation of individual quality assessments rather than being the primary sources that drive the assessments. We found even small increase in individual accuracies will significantly improve the overall effectiveness of crowdsourcing quality control. Implication to domains such as online health information Web sites or product reviews are discussed. Keywords: Web quality, user reviews, information cascades, social dynamics.
The rapid growth of the WWW has led to increasing concern for the lack of quality control of online information. Many have suggested that quality assessments of online information could be “crowdsourced” to the public by harnessing the collective intel- ligence of online users by allowing them recommend contents and share the recom- mendations with others. Although this kind of crowdsourcing seems to work well in many domains [2, 3, 9], recent studies, however, show that aggregate behavior is often subject to effects of cascades, in which the social dynamics involved during the accumulation of recommendations may not be effective in filtering out low-quality information . Indeed, previous models show that when people imitate choices of others, “bad cascades” may sometimes make low-quality information more popular than high-quality ones . Previous cascade models often assume that users imitate choices of others without direct communication of quality information. We extend previous models by assuming that users can make multiple choices among web sites, and they can communicate their quality assessments to other users. Our goal is to understand how individual quality assessments may moderate effects of information cascades. We investigate how cas- cades may be influenced by factors such as accuracies of individual quality assess- ments, confidence levels of accepting other users’ recommendations, and impact of aggregate user recommendations on choices. The goal is to simulate effects at the indi- vidual level to understand how the social dynamics may lead to different aggregate patterns of behavior. Specifically, we investigate how individual quality assessments propagate in a participatory Web environment, and whether information cascades may moderate the effectiveness of aggregate user recommendations, such as whether they may lead to “lock in” to low-quality information; and if so, to what extent will they impact the overall effectiveness of crowdsourcing quality control of information.
2.1 Quality Assessment of Online Information
While research has shown that people are in general good at choosing high quality Web sites, it is also found that people often browse passively by relying on what is readily available. For example, health information seekers are found to rely on search engines to choose online sources, and may utilize surface features (such as layout and structures of web pages) to infer quality . Accuracies of quality assessments are also found to vary among individuals, and can be influenced by various factors such as background knowledge, Internet experience, cognitive resources available, etc . Liao & Fu  also found that older adults, who tended to have declined cognitive abilities, were worse in assessment quality of online health information. Research in the field of e-commerce shows...
Please join StudyMode to read the full document