A Usability Model for Web-Based Personalization Based on Privacy and Security

Personalization approaches have become popular as content providers have attempted to provide features that differentiate their sites from the sites of competitors while enhancing usability. For instance, the Google and MSN news pages offer the ability to modify the content and size of news categories. Additional personalization and customization features are also offered on sites such as Google News, MSN, and Yahoo!. Personalization requires the acquisition of data that can be used to modify the content or presentation of the web page. Users have demonstrated that they often claim to be privacy conscious, but still provide information easily to sites on the Internet for widely varying levels of benefit in return (Berendt, Gunther, & Spiekermann, 2005). In order to better align the privacy preferences of users to actual practice, a multi-phased survey was conducted that measured user preferences for how data should be obtained, used, and shared by sites. The survey included real-world personalization scenarios in order to identify the conditions under which users provide information to a site to support a personalized experience. The results revealed that personalization is not necessarily a differentiating feature in news site choice. With respect to privacy, users do not get substantial information from the privacy policies of sites they visit and are largely unaware of common data collection practices by news sites to support personalization. Users are generally less concerned about interface data privacy than the privacy of user-based data. Finally, privacy breaches are more likely to affect user satisfaction than unintentional security breaches. Additional analyses were performed using ordinal regression, principal components analysis, and non-parametric techniques for identifying differences and correlations. The results also facilitated the development of a usability model and associated recommendations for how sites that implement personalization should handle data used to support personalization. This includes communication mechanisms such as privacy policies and the impact of privacy and security breaches. As part of this research, a conceptual model is presented that demonstrates the relationships between factors such as usability, privacy and security, and complexity of a personalized architecture.