Filters for Book Series Diversity in a Conversational Book Recommender; An Actor-Network Theory Analysis of the Factors Impacting Voice Assistant Usage

Despite their widespread availability, recent reports have indicated that voice assistants have been underperforming in customer usage and revenue returns. The technical project describes internship work done on book recommender of a popular voice assistant to eliminate an issue with repetitive book series recommendations. This change aimed to increase the diversity of book series presented and ultimately increase the usage and engagement of the experience. The STS research paper seeks to understand the wider sociotechnical reasons behind voice assistant adoption. To do so, the paper examines the history of Siri and Alexa using Actor-Network Theory to understand the actors driving the direction of the technology and the impacts of their influences on its adoption. Through this work, we will be able to examine both the specific technical tasks involved with increasing voice assistant usage and the values embedded within the technology that have impeded adoption.
In the technical report, we focus specifically on modifications made to the conversational book recommender of a popular voice assistant. In the existing system, customers would sometimes be presented with multiple books from the same book series in a row. This created a repetitive or low-quality experience which negatively impact user satisfaction and engagement on the recommender. To address this issue, the technical work introduces additional filters to ensure the uniqueness of the book series presented and increase the diversity of book recommendations.
With the implementation of the technical project, the book recommender ensures that only one book per book series will be recommended within a single session. Additionally, the system will no longer present returning users with books from series that it has previously recommended. These changes were initially launched through A/B tests where the feature was first made available to a limited subset of users. While the results are unknown at the time of writing, they are projected to result in a small increase in the number of users interacting with and returning to the book recommender.
The STS research paper seeks to understand the factors negatively impacting voice assistant usage despite the widespread availability of the technology. The paper explores the history and development of Siri and Alexa using Actor-Network Theory to understand how the technologies came to exist in their current forms. This analysis focuses on the motivations behind the creation of the technology and the influences driving Alexa’s core design as a standalone smart speaker with a conversational user interface. Finally, the paper applies ideas from the Technology Acceptance Model to examine how the values influenced by different actors involved in the development of Alexa have impacted its perception of utility and ease of use.
The case studies of Siri and Alexa show that while Siri is the result of social and organizational factors involving government research and emerging smartphones, Alexa’s development was primarily driven by technologically deterministic ideas of creating the next big computing platform. This goal, in addition to influences by actors in anthropomorphizing the technology, has resulted in difficulties with using the devices and increased expectations of capabilities that are not present which may hinder adoption. The research indicates that future directions for improving voice assistants involve shifting focus to multi-modal devices with displays or through the creation of domain-specific voice assistants.
While incremental technical improvements can help improve the adoption of voice assistants, it is also important to understand how the ideas and attitudes towards the technology can impact its success. With voice assistants, these ideas extend beyond the individual technical components and emerge through the relationships and influences of actors involved with its creation. Addressing these social values is another crucial step in driving the adoption of voice assistants.

School of Engineering and Applied Science
Bachelor of Science in Computer Science
Technical Advisor: Briana Morrison
STS Advisor: Catherine Baritaud