Examining the effects of personalized App recommender systems on purchase intention: A self and social-interaction perspective

Abstract

(ProQuest: ... denotes formulae omitted.)1.IntroductionThe emergence of user-based Web services and personalization technologies has allowed many companies to provide personalized content or services to users [Hill and Troshani 2010] and business intelligence [Foshay and Kuziemsky 2014; McBride 2014]. User-based Web services and personalization technologies refer to Web services and technologies that employ personal user information. Rapid improvements in Web, mobile facilities, and services have significantly increased the variety of choices available to customers, and have led to the development of a very large number of mobile software programs called (application programs for PCs and mobile devices, such as smartphones and tablets). Apps allow users to perform specific tasks on their desktops and mobile devices (e.g., iPads and Macintosh computers). For example, Apple OS X operates all programs as apps, and many Macintosh (Mac) users search for apps in Web-based app stores.Several apps have been developed. For example, Apple's App Store stocked more than one million apps in January 2015 [Gartner, 2013]. Most app store sales are made in recreational categories, such as entertainment, social networking, and music. Faced with the large amount of content available in many forms in an ultra-competitive environment, along with intense pricing pressure, would-be customers find it difficult to identify and select appropriate apps from among the many similar ones that are available [Herlocker et al. 2004]. Gartner [2013] reported that annual downloads from mobile app stores reached 102 billion in 2013, up from 64 billion in 2012, whereas total revenue reached $26 billion in 2013, up from $18 billion in 2012 (see Appendix 1). Therefore, app stores have had to devise methods for assisting customers in their search for appropriate apps that satisfy their needs.Previous literature on recommender systems has considered online content, such as news and movies. However, app markets have changed recently with the availability of more than one million apps, and it is difficult for customers to easily find suitable apps on their mobile devices. App recommender systems are therefore important because in the app market context, they can reduce customer search cost while yielding better search results. To help customers find suitable apps, personalized recommender systems (PRS) have been developed, thus allowing the delivery of enhanced, customized information or products in response to Web searches [Liang et al. 2007; Tam and Ho 2005; Wang and Benbasat 2007]. PRS determine and employ user preferences in order to generate recommendations that help such users select personally helpful and interesting items [Benlian et al. 2012; Liang et al. 2007; Tam and Ho 2005; Herlocker et al. 2004]. Their purpose is to retain customers by making it less appealing or attractive for them to switch, and to facilitate customer searches for products or information [Shani and Gunawardana 2011; Hess et al. 2009; Xiao and Benbasat 2007]. PRS are based on the premise that users already exposed to relevant Web content seek less information and spend less time making decisions [Choi et al. 2014; Choi et al. 2011; Tam and Ho 2005, 2006].To this end, previous studies on recommender systems have proposed various predictive metrics, such as accuracy, novelty, and variety, which differ from users' perceptual evaluations of app recommender systems [Shani and Gunawardana 2011; Palanivel and Sivakumar 2010; Adomavicius and Tuzhilin 2005; Herlocker et al. 2004]. Among those metrics, accuracy is important in order to ensure that recommended items are those items that users want to find based on their preference. Nevertheless, app stores cannot rely on the accuracy measure alone to evaluate recommender systems [Palanivel and Sivakumar 2010]. User behavior in terms of choice or purchase does not always correlate to high recommender accuracy [McNee et al. …