1
A Brief History of Good Choices
In Existentialism Is a Humanism, Jean-Paul Sartre tells of how one of his students came to him asking for advice. It was 1940, and the studentâs brother had just died in World War II. He yearned to avenge him by joining the fight, but he also had a grief-stricken mother who needed care. In response to this dilemma, Sartre responded, âYou are free, so choose; in other words, invent. No general code of ethics can tell you what you ought to do: no signs are vouchsafed in this world.â1 While this comment may seem ineffectual and perhaps even flippant, Sartre was a strong believer in the importance of human agency, and any student asking him for advice should have expected just such a response. Reflecting on this episode, Sartre noted that when we ask people for advice, we tend to already know their perspectives and what advice they will probably give (i.e., if the student had actually wanted a more direct answer, he should have asked someone else). As Stephen Priest argues, âTo choose an adviser is to make a choice. It is also to choose the kind of advice one would like to hear.â2 No matter how much we try to divorce ourselves from the many small and gargantuan choices we must continually make by relying on recommendations from those we trust, the decision is always ultimately ours. According to Sartre, this freedom to make decisions and use such choices to continually invent ourselves throughout our lives is less a gift than a burden that we cannot escape; âman is condemned to be free.â3
The question of how constrained oneâs actions are by social structures and how, when, and whether individual agency can really guide our behaviors lies at the heart of much of the humanities and social sciences. While RenĂ© Descartesâs famous phrase, âI think, therefore I am,â idealizes human agency above social structures, others, like Karl Marx, argue that our choices are largely guided by the ideological structures around us. By arguing that we are free and must invent ourselves accordingly, Sartre seems to sit at the far end of this debate. Yet his most noteworthy move here transforms our ability to choose from a pleasurable strength to our greatest liability and the source of our misery and anxiety. Choice and freedom are typically presented as universally desirable and thereby also as the neutral premise of all post-Enlightenment ideologies and the formation of a rational cohesive subject. However, for Sartre, the privileging of freedom and choice is itself the ideology that must be critiqued and deconstructed. How rational can people really be if the freedom they seek is a source of their consternation?
While I am inspired by Sartreâs move to consider choice a burden, I would argue that he actually does not go quite far enough in his critique. For Sartre, while no choice is unassailable, the very act of making choices for oneself still remains a central expression of agency and authenticity. Yet the question of what constitutes a choice is far more subjective than he indicates. What one might call a choice, another might experience as a compulsion, obligation, or ultimatum. As contemporary activism on both the left and right, from Black Lives Matter and Me Too to efforts to âbuild the wallâ and âmake America great again,â illustrates, poverty, racism, sexism, and other forms of exploitation and exclusion limit the number, type, and quality of choices many people can make. For some, making certain choices subjects them to greater scrutiny; circumstances may even make it difficult to perceive some people as being capable of making choices at all. By arguing that we are responsible for all our actions and their outcomes, Sartre ignores the structures within which choices must be made by those who are more constrained in the options available to them.
Indeed, much of contemporary racism and sexism does not just limit the options certain people have but can also turn those things they do experience as choices into forms of constraint. For example, Linda Duits and Liesbet van Zoonen argue that neither the decision of Muslim young women to wear headscarves nor the decision of white young women to wear G-strings is typically represented as a choice, even though each is often experienced as such by these women.4 They argue that these are just two examples of how the choices of young women are rarely respected as choices and that young women are therefore rarely thought of as having agency at all. Rather, they are presented entirely as subjects only capable of expressing the will of external influences. Celebrating choices continues to privilege bourgeois white men, who typically have access to more and better options from which to choose. Indeed, choices do not lead to agency and freedom; rather, individuals must first be imagined to have agency and freedom in order to have their decisions count as choices.
The relationship between choices, freedom, and influence is complex and culturally situated. Here, I am interested in exploring how the recommendationâwhich simultaneously, if paradoxically, privileges individual agency as well as structures of control and influence (which typically appear to limit agency)âcan help us critique how choice currently functions. I begin by comparing the etymology of recommendation and suggestion in order to elucidate current debates about whether outside influences and advisers lead to a perversion or affirmation of personal choices. I then focus on how this debate about conceptions of choices and recommendations influenced the concurrent and overlapping development of both artificial intelligence (AI) and feminism. Historically, much of the discourse concerning the internet and digital technologies has focused on how they enable users by broadening their horizon of possibilities and enlarging their sense of free will. Yet this perspective ignores the multitude of ways in which digital technologies present choice not as liberatory but rather as a burden. I then address how this history has shaped currently intertwined neoliberal, postfeminist, and postracial ideologies, and how these ideologies now frame choice as a burden. Indeed, since the mid-1990s, various industries have used recommendation systems to profit from presenting choices and agency as burdens in order to then sell us a sense of relief in the form of convenience and simplicity. Digital media continually prescribe predefined choices and actions through their algorithmic interfaces of drop-down menus, buttons, check boxes, and other clickable objects. Unless you have a great deal of coding experience and time on your hands, these are typically your only options. Of course, you could always just reject all the choices by not using particular seemingly omnipresent websites and applications (or, for that matter, your computer), but due to economic and personal responsibilities and expectations, that option rarely feels plausible. Regardless, these limited choices are often actually presented positively as a feature of digital technologies rather than their greatest limitation (or, perversely, limitations are now the affordance). For instance, Apple markets its design principles on simplicity, a code word for limiting the number of options its customers are presented with in order to avoid confusion and making those few options they do have appear as desirable as possible. Tracing the long history of recommendations helps to explain how these contradictory discourses and impulses work today.
From even the earliest moments of the World Wide Web, recommendation systems were used to sort through the abundance of information and choices that digital culture created. In my introduction I discussed how Tapestry, the first collaborative filtering-based recommendation system, was presented as the antidote for too much information, too much clutter, and too much choice. To explain how recommendations became a major industry, I end this chapter by recounting how the GroupLens Computer Science and Engineering Research Lab at the University of Minnesota, one of the earliest groups to work on recommendation systems in the early 1990s, conceived of and eventually marketed these technologies. Their work continues to resonate through a vast industry that sells recommender and personalization technologies to companies worldwide. At the same time, I argue that by framing free choice as a burden best alleviated by capitalist forces, GroupLens and the many companies that followed them also frame the act of making unrecommended choices not just as dubious but also as subversiveâwith both the negative and the positive connotations that subversion may entail. This alternative understanding of choice has been deeply inflected by the role that choice and recommendations have played throughout the early history of computing. While Sartre spoke of choices making us who we are, now these industries frame unrecommended choices simply as a waste of time and resourcesâthat is, the ultimate sin, according to capitalists.
Recommendations and Suggestions
While I use the term recommendation throughout this book because that is the most common term used by software engineers and digital industries, some also use suggestions, top picks, and other terms interchangeably. These terms are all now synonymous, but each has distinct connotations and a long history that illustrates current values, debates, and anxieties about algorithmic recommendations. As Raymond Williams argues, the meaning of words may change, but the history of these transformations often haunts them into the present: âEarlier and later senses coexist, or become actual alternatives in which problems of contemporary belief and affiliation are contested.â5 In our contemporary use of recommendation, suggestion, and top pick, we can see the traces of larger battles over the shape and meaning of digital culture.
For instance, Netflix (which I discuss in detail in chapter 3) had long labeled the movies and series it highlighted as ârecommendationsâ but has recently begun calling them âtop picks.â At the same time, it also began advertising that it was no longer just using algorithms to make recommendations but had also hired many people to watch all its media and label them with keywords to improve its picks.6 Here, âtop picksâ implies that someone, instead of some machine, is physically making a pick for you. This distinction foregrounds contemporary debates about the value of the labor of AI versus that of humans, raising the question of whether we care if the advice we get is in some small way shaped by a human, and whether we trust humans or machines more with all the personal data that go into making such âpicks.â
Pundits and academics alike continually debate how digital technology and culture figures ânewâ concerns about choice, privacy, security, agency, surveillance, and relationships, but here I will examine how these anxieties have always been implicit in the way we use two very old and very connected terms: suggestion and recommendation. Our anxieties over taking advice and ceding agency are a result not of digital technologies but rather of the way we rely on them for such advice. Suggestion and recommendation have distinct histories that illustrate how our fears, hopes, and misgivings about these actions have played out over the centuries and into the present. Indeed, such fears have always been a central aspect of suggestions and our ability to make up our own minds.
Suggestion first came into English in the mid-fourteenth century, and its original meaning was âprompting or incitement to evil; an instance of this, a temptation of the evil one.â7 Rather than describing a benevolent act, this definition points to the lack of self-control and agency of the one receiving the suggestion, as well as the evil intentions of the one who suggests, who, in this context, was understood to be the devil. Most definitions for suggestion focus on questions of temptation, self-control, and harm. While Sartre places the burden of choice squarely on the subject, suggestions historically have caused confusion about where agency lies. At times, suggestions stress how rarely we actually have final say over any choice we make. In John Miltonâs Samson Agonistes, suggestions are depressedâif not suicidalâthoughts that âproceed from anguish of the mind and humors black.â8 Here, the uncontrollable unconscious and bodily chemistry become the sources of suggestions and the choices that spring from them. We may make such choices, but only when we are not ourselves. In the works of Geoffrey Chaucer and Thomas Kyd, suggestions tend to be false, and subjects must be âfreed fromâ them through confession.9 Not only evil, these suggestions supplant our sense of personal choice and agency. Further, these examples illustrate that it is within our power to reject even these suggestions, but only if we confess and side with God.
By the beginning of the twentieth century, suggestion had become synonymous with hypnotism and the ability to take full control over anotherâs thoughts, actions, and desires.10 While we now use this term in a much more neutral fashion, it still connotes anxieties about self-control and manipulation. We fear that the suggestions that now surround us online will make us suggestible, or easily led toward bad or even evil decisions. For example, when Facebook announced that it was experimenting on its users by sometimes making depressing or uplifting stories appear more often on their feeds (which is Facebookâs way of suggesting articles that you might like), many, like Jim Sheridan, a member of the British House of Commons, feared that such actions would âmanipulate peopleâs thoughts in politics or other areas.â11 Clay Johnson, the cofounder of Blue State Digital, the firm that built and managed Barack Obamaâs presidential online campaign platform, questioned whether such suggestions could control people en masse to start revolutions and swing elections.12 If Facebook started only showing or suggesting positive articles about an elected official or corporation, we may mistakenly interpret this to mean that there are no negative articles; choice is meaningless if we are ill informed and unable to imagine the many choices that we can actually make. With the Trump campaignâs candid use of social media to spread outlandish rumors, the transformation of âfake newsâ into a ânational strategyâ in the form of Cambridge Analytica, and Breitbart and the ascendancy of the white supremacist Steve Bannon, anxieties over the relationship between suggestions and suggestibility have only become more pronounced. As Ann Coulter puts it in a Breitbart article, âMost people are highly suggestible. Thatâs why companies spend billions of dollars on advertising.â13 Advertising is thus a form of suggestion; recommendation, however, connotes something more benign.
If suggestions come from the devil, recommendations often come from God. In the King James Bible, the grace of God ârecommendsâ that Paul and Barnabas sail to Antioch, where they start a church and where the followers of Christ are first called Christians.14 Recommendations are predictive and connote a placement of faith in the recommender and authority figures more ge...