Netflix recommends: algorithms, film choice, and the history of the taste

Watching movies, television series, and reality shows on video streaming platforms like Netflix, either individually or in the company of friends and family, has become an integral part of people’s daily routines. Millions of users actively interact with these platforms in unique and unexpected ways, rendering the recommendation algorithms of Netflix-like companies never truly ‘perfect’. In Netflix Recommends: Algorithms, Film Choice, and the History of Taste, Mattias Frey shifts from the assumption of novelty and the prism of technological determinism, to investigate the algorithms behind recommender systems through a user-centred approach, examining the cultural recommendations and media-consumption choices from a media archaeological perspective. The author prompts us to remember the views of Huhtamo and Parikka (2011), who seek to identify continuities alongside uniqueness, to look for the old in the new, and to recognise the novelty that already exists in the old. By drawing lessons from these insights, Frey contextualises Video On Demand (VOD) recommender systems within the historical landscape of information regimes to scrutinise their working mechanisms, Frey challenges the myth of recommendation algorithms and reminds us that there has always been a need for gatekeepers, filters, attention focalises, curators, and recommenders. In doing so, he also discards the conventional scholarly approach that treats consumers as passive subjects.

The book can be divided into three main parts. In the book’s first part, Chapters 1 and 2, Frey thoroughly examines VOD recommender systems, exploring how they interfere with the ecologies of cultural mediation, taste, and choice. These chapters provide a conceptual understanding of how these systems operate, laying the foundation for the next part of the book. In Chapter 1, ‘Why We Need Film and Series Suggestions’, Frey demonstrates how VOD recommendation systems address a fundamental challenge film and series face: how to capture audience attention amidst a vast array of cultural products. By placing VOD systems within the broader context of historical information regimes that seek to capture consumers’ attention and guide their decision-making processes, the subsequent sections of this chapter highlight the necessity of these recommender systems in the current digital era.

In chapter 2, ‘How Algorithmic Recommender Systems Work’, Frey dives into the working mechanism of the algorithmic systems, outlining six important characteristics of how such recommendation systems function: (1) Recommendations are generated by subjecting viewers’ data to a sequence of algorithms, (2) providing personalised and more accurate suggestions which beomes the main value-added selling point of VOD services. (3) The system remediates traditional forms of cultural suggestions like word-of-mouth on a massive scale and via data-based collective intelligence; (4) but the computational mechanisms behind the systems are not transparent to users. However, (5) the system establishes credibility by emphasising scientific objectivity, technological innovation, and thoughtful design. Finally, (6) these personalised recommendations could introduce a new cultural taste that goes beyong traditional gatekeepers. Furthermore, this chapter emphasises the underlying social needs driving VOD recommender systems by juxtaposing them with preceding recommendation methods such as critical reviews, word of mouth, posters, and trailers. According to Frey, the enduring presence of these various forms of recommendation underscores a fundamental reality: recommendations play a distinct, essential, and valuable role in the consumption of films and series, as they help consumers to reduce the uncertainty associated with the quality of cultural products before consumption, while decreasing the amount of research they would otherwise need to make better judgments about the product’s worth (p, 27).

Chapters 3 and 4 form the second part of this book, presenting a comprehensive case-study analysis of Netflix. Chapter 3 discusses Netflix's origins and historical development since its founding in 1997. Additionally, it explores how this system evolved in response to the introduction of the new streaming model. Frey asserts that the designs and styles of the Netflix algorithmic recommendation system draw heavily from past forms and norms such as video-store classifiers, video-clerk suggestions, broadcast criticism, top-ten lists, and peer word-of-mouth tips and that system is not as mysterious as the research and discourses claim. Moreover, as this chapter reveals, to further captivate viewers and establish trust, Netflix’s engineers have optimised the recommendation engine by blending personalisation with other non-personalised factors, such as critical acclaim, short- and long-term popularity, novelty, and diversity. Consequently, Netflix’s personalised system remains, in fact, ‘resolutely unpersonalised’ (p. 66).

In Chapter 4, through an in-depth analysis of relevant academic, business, and technical discourses, company press releases, and publicity, Frey thoroughly explores the mythology surrounding Netflix’s recommendation system, focusing on the algorithm driven by the company’s extensive collection of big data. This chapter demonstrates how Netflix establishes its credibility and raises concerns among media scholars through a series of Public Relation tactics highlighting its system's scientific precision and objectivity. Similar to how film critics must perform authority, knowledge, and a certain level of distance or familiarity with the industry to establish their trustworthiness, Netflix achieves this by delivering a performance to the audience that embodies scientific objectivity, innovation, and differentiation. To break the illusion about the algorithm and take a closer look at the actual impact of Netflix’s system on the culture of films and series, at the end of this chapter, Frey advocates for a user-centric perspective in examining the Netflix recommendation system, as the unique experiences of diverse real users serve as evidence of the imperfections of the system.

Chapter 5, the final part of the book, continues the analysis from the previous chapter and presents the empirical findings and arguments on how algorithm-driven recommendations influence users’ taste acquisition and development on streaming platforms with a folk theories perspective. With data from an empirical study based on two nationwide (United States and United Kingdom) surveys and dozens of in-depth interviews, this chapter explains what role recommender systems play in the process of accessing and selecting films and series; how users perceive and understand the algorithmic mechanism; and how users evaluate the efficiency of the recommender system in relation to alternative sources of recommendations. In this chapter, Frey identifies several intriguing categories of users based on their choice motivations and repertoires, such as Lazy Choosers – they typically do not read critics’ reviews or articles about films or series, or consult online user forums; they tend to prefer shows that do not require much attention because they use VOD as a kind of ‘wallpaper’ while multitasking.

This book comprehensively explains how real users interact with VOD recommender systems based on their motivations in a world saturated with films and series. It reasserts the agency of audiences while highlighting the persistent social need for curated choices and critical comments, particularly in movies and series. As Frey aptly states: ‘Criticism did not die: the need for gatekeepers, attention focalizers, context, and information persisted.’ (p. 204). The book also reveals that algorithmic recommender systems’ widespread trust and usage are not as pervasive as commonly believed. These systems are not unique, inevitable, or without alternatives, and past forms and practices still influence their development. The fallacies surrounding VOD recommender systems and the concept of technological determinism are debunked by the thorough analysis and empirical findings presented in this book.

Certainly, Frey’s argument would benefit from also considering what VOD recommendation systems designers and owners, policymakers, film critics and relevant stakeholders think about this issue. Additionally, considering that Netflix provides services almost all over the world, the environment, scenarios and purposes of users in different regions using the VOD system will be different due to the actual political, economic and cultural conditions of each country (especially in non-Anglophone countries, such as Japan and South Korea), so I suggest that future research related to Netflix and other VOD systems is necessary, and the arguments and data in this book can serve as a solid foundation and enlightening references for these studies. Nonetheless, this book will greatly interest anyone concerned with cultural recommendation, audience’s media consumption choice, VOD platforms and their algorithmic recommendation systems.