Vibes and algorithms, or “vibe check” as biopolitics

I’m working on the part of my “Good Vibes Only” book where I distinguish what I’m calling “a biopolitics of frequency” – the kind of biopolitics Foucault talks about in the 70s – from the form of biopolitical governance performed by algorithms today that use vectorial orientations rather than normalized statistical distributions to make decisions. In the stuff I posted below, I work through how exactly algorithms govern via vibes. At the end I explain how algorithmic vibe checks govern sexuality for legitimacy rather than normality.

As Melinda Cooper notes, beginning in the 1990s a form of “neoliberal biopolitics” emerged that was distinct from the kind Foucault theorized in the 1970s. Unlike Foucault’s biopolitical normalization, which “speaks the language of Gaussian curves and normalizable risk,” this new biopolitical logic is modeled on “neoliberal theories of economic growth [which] are more likely to be interested in the concepts of the non-normalizable accident and the fractal curve” (LAS 10). Whereas probabilities track the frequency of aleatory events in past data, neoliberal biopolitics uses speculative tools that orient our perception to possible but presently counterfactual realities like Donald Rumsfeld’s infamous “unknown unknowns.” As Louise Amoore explains, these tools “invit[e] the intuitive and the speculative within the calculation of probability” (PP 45), layering this speculative or possibilistic calculus on top of existing probabilistic math. More specifically, algorithms incorporate feelings and intuitions–what Cooper calls “the essentially speculative…movements of collective belief, faith, and apprehension” (LAS 10)–into numbers computers can crunch. And as work by Amoore and anthropologist Nick Seaver reveal, this invitation occurs as a demand to check the vibe or orientation of data modeled on a spatial axis.

Amoore traces contemporary algorithmic techniques back to Raakesh Agrawal’s invention of “data mining” from studying consumer purchase data from British retailer Marks & Spencer (PP 39-40), and Seaver traces the history of recommendation algorithms like Agrawal’s back to the midcentury invention of non-metric multidimensional scaling (MDS). MDS is a method for quantifying perceived similarities and dissimilarities among phenomena. As Seaver explains, geographic space serves as the epistemic framework for MDS modeling, where “formal and informal senses of ‘space’ are woven together in the pursuit of plausibility” (Seaver ES 48). In his survey of how English-language textbooks introduce MDS, Seaver finds a pedagogical convention where students are taught to read MDS models like they would read a map. These textbooks plot a table of distances between major U.S. cities on an x/y axis, which reveals something that looks like a map of the 48 contiguous states. The fact that this method of data analysis and visualization produces a model that reflects American audiences’ general sense of what the US looks like on a map is then used as both proof of concept and as pedagogical exercise. “Having thus confirmed that the technique works, textbooks move from physical distances to conceptual ones” (Seaver ES 49) where students are asked to use their visual-spatial intuition to find meaning in models of non-geographic data, such as the perceived similarities among different fruits. Vibe checks are the last step in building the algorithms behind recommender and machine learning systems, and these checks consist in interpreting the spatial orientation of data points.

Contemporary algorithms automate this last interpretive step by computing the relative similarity of data modeled as vectors in space. As Seaver explains in a different article, “Vector spaces are the symbolic terrain on which much of the labor of machine learning works, and they provide a widespread metaphorical language across the software industry” (Seaver CS 519). Vectors are objects with both a magnitude and a direction, and they are modeled by a line of defined length pointed in a specific direction. Echoing the language of vibes as orientations, Seaver explains that when “represented as vectors, objects are defined by their orientation” (Seaver CS 518), i.e., by their situatedness in space, what they unfold toward and what they put behind them. Algorithmic systems then compare orientations among vectors to produce decisions about things like what shoppers are likely to buy or whether a student taking an online exam is cheating. As Amoore notes, the relevant metric here is degrees (PP 12), such as the geometric or cartographic degrees used to measure orientation in space. For example, vectors that are oriented at right angles (90 degrees) to one another are held to have no relation to one another. Algorithmic vibe checks clock the degrees of spatial orientation among vectors.

Used across contemporary tech, finance, and security, these algorithms constitute a form of biopolitics whose object of knowledge and mode of governance is spatial orientations or vibes. These vibes are checked not for their ab/normality, but for their legitimacy–that is, for their perceived capacity for private responsibility or private property ownership and acquisition. As Cooper has shown, over the last 40 years US government policy has replaced the forms of social insurance designed to normalize risk across the population with various practices of private family responsibility: “Here we encounter an aspect of neoliberalism that eludes the terms of Foucault’s now classic analysis…The antinormativity of Chicago school neoliberalism is contingent upon a moral philosophy of prudential risk management that leaves no excess costs to the state. This…finds expression in the idea that non-normative sexual relationships must ultimately be channeled into the legal form of marriage.” (FV 174-5). Cooper’s language here is helpful because it tracks the shift away from norms in general and sexual normality specifically towards legitimacy, which is both the legal status conferred by marriage and the expression of private family responsibility. In Foucault’s account, the discourse of sexuality emerged because sex qua procreation was the hinge between the normation of individual bodies and the normalization of the population. However, as population normalization falls out of fashion in favor of algorithmic vibe checks, sexuality is governed in terms of legitimacy rather than normality. 

Michelle Murphy’s discussion of The Girl figure is a perfect illustration of sexuality qua legitimacy. Murphy finds that in the 21st century, social science has shifted from modeling people as populations to modeling them as markets. She calls this “the economization of life.” In this framework, the individual subject appears as human capital. “The Girl” is Murphy’s term for how contemporary biopolitics imagines its ideal subject. As she explains, the “‘Third World girl’–typically represented as South Asian or African, often Muslim–has become the iconic vessel of human capital” for whom “‘chance’ is translated into only two possible paths: the unproductive life and the productive life” (EoL 117). A “productive” life is one where the Black and Brown girls of the Global South reproduce responsibly and don’t rely on support from governments or NGOs. As Murphy puts it, “thoroughly heterosexualized, her rates of return are dependent upon her forecasted compliance with expectations to serve family, to adhere to heterosexual propriety, to study hard, to be optimistic, and hence her ability to be thoroughly ‘girled’” (EoL 117). A “productive” life is one where cisheterogender roles and participation in the patriarchal nuclear family govern sex and reproduction. The issue here is less sexual normality and more the girl and her family’s ability to privately assume the costs of her sexual choices. Murphy’s counterexample of the unproductive life–”married by fourteen, pregnant by fifteen, after which she may have to sell her body” (EoL 117)–makes this clear: to be unproductive is to be unable to privately assume the costs of one’s sexual behavior and to engage in illegitimate sexual behaviors (figured here as sex work). 

Though she doesn’t use the language of vibes specifically, Murphy’s account is especially helpful because she highlights how The Girl’s perceived il/legitimacy is represented as what is easily recognizable as a vibe: “The Girl is represented abstractly as a charged data point, an ebullient cartoon, or an animated icon. Through music, color, and animation, the icon of the Girl is excited with enthusiasm, hope, ambition, and responsiveness to transformation that joins Western liberal feminist imaginaries of empowerment into speculative finance…The Girl is a calculated risk pool that draws together a bloom of possibility, a bouquet of potential, a cluster of affect, applicable to any dispossed condition anywhere, as long as it is ‘girled.’” (EoL 118-20). An abstraction conveying the successful privatization of the costs of reproduction, “the Girl” is less a description of empirical, living people and more a “girled” orientation whose alignment fits with the imperatives of neoliberal feminism and finance capitalism gives rise to good vibes in people and institutions who are similarly aligned. Murphy’s analysis of “The Girl” is helpful because it is an example of the discourse of sexual legitimation that expresses such legitimacy as a vibe. Neoliberal biopolitics governs by checking vibes for their legitimacy.

This shift away from norms and toward legitimacy is also rooted in the difference between normal curves and vectors. Returning to Seaver’s study of contemporary algorithms’ roots in MDS, it’s significant that his primary example of how people talk about modeling relations among vectors centers on the relationship between sex and procreation. As Seaver explains, in the keynote speech at the 2001 meeting of the International Society for Music Information Retrieval, David Huron uses “a striking pair of values: sex and procreation” to argue that vectorial models are advantageous because they allow mathematicians to disarticulate phenomena that otherwise appear mutually dependent. Seaver explains,  

For most of human history, Huron argued, people have valued sex and procreation but have been unable to pursue one without the other. In vector terms, sex and procreation are correlated: they point in the same direction. The advent of contraception, Huron claimed, rearranged these vectors, pushing them to (almost) 90o apart by making it possible to move in the direction of the sex vector without simultaneously moving toward procreation. As Huron put it, contraception decorrelated these values, making them orthogonal to each other. Such was the power of engineering, which could alter the bearing of value vectors “through the manipulation of physical or social reality,” resolving value conflicts by decorrelating them out of existence.” (Seaver CS 519)

The rise of the human sciences and the widespread use of normal curves to model populations lead power to study and govern sexual normality because it treated procreation and sex as dependent variables: the size and health of the population depends on people’s sexual behaviors, so the population can be managed by compelling people to behave normally when it comes to sex. However, now that we can study and manage people with tools that can treat sex and procreation as independent variables to reflect circumstances when material conditions are such that they actually are not dependent, white supremacist capitalist patriarchy is itself less dependent upon normalization and can make use of other tools to cut the racialized, gendered line between acceptable and unacceptable sexual behavior and identities. Biopower has more tools in its arsenal to govern sexuality. For people in circumstances where sex and procreation are dependent (e.g., people who can get pregnant but don’t have access to contraception, queer couples trying to conceive, etc.), good old norms and normalization work just fine. But for people in circumstances where they are not necessarily dependent, practices of legitimation step in to do the work traditionally done by norms and normalization.

My argument here is that evolutions in math since the 1970s have allowed a new form of biopower to emerge. In addition to the disciplinary normation of individual bodies and the statistical normalization of populations, there is now the legitimation of vibes or vectors. Or, more simply: neoliberal biopolitics governs via vibe check. 

As I’ve suggested above, this governance via vibe check has direct implications for how sexuality is regulated and what kinds of queer people are folded into life and what kinds of queer people continue to be left to die. More on that later.