Then were the horsehoofs broken by the means of their pransings

In an essay for the Walrus about the pervasiveness of “fake reviews” online, Timothy Caulfield describes something he calls “the opinion economy,” which, he argues, “is built on our desire for a beacon of clarity and certainty in our fantastically chaotic information ecosystem.” Caulfield has a capacious definition of fake:

by fake, I mean not written by a true customer. [The reviews] were either purchased, generated by an AI-powered bot, or written by someone with an ulterior motive, such as wanting to sell stuff or, at the other end of the motivation spectrum, trying to hurt a business, product, or service.

As inclusive as this definition of fakery is, it fails to include the most important consideration, which is that the space provided for reviews is not some version of a public sphere, a neutral space for the free exchange of ideas, but a part of a retail site’s marketing strategy. The review section not to help prospective customers but to help legitimate the platform, make it seem like a normal and trustable venue for commerce. Every review is a sales pitch for the site itself, regardless of what it says about a specific product. Just as every piece of information encountered on a social media platform is “misinformation” by virtue of its context, so is every review on a retail site a “fake review.” “Truth” is not part of the procedure for its production. Setting up a standard for “truth” for this kind of discourse misapprehends what sort of discourse it is.

The same can be said of “true customer.” The concept of a “true customer” is a necessary antithesis, I suppose, if you want to have some grounding for a complaint about “fake reviews,” but it also seems to mark an unknowable void, a kind of black box. What sits at the different ends of the “motivation spectrum”? Isn’t any motivation to write a review at all already a bit suspect, coming from some place that is tangential to fully absorbed consumption? What kinds of motivation would the “true consumer” have to do anything at all but happily consume products? Where does reviewing fit into that pure metabolism? (I wonder what Caulfield would make of Kevin Killian’s Selected Amazon Reviews.)

Lurking in the “true customer” is a fantasy about the sanctity of use value, which can liberate us from the vagaries of exchange and all the deceitful “ulterior motives” it necessarily harbors. The review section beneath any product immediately confronts a prospective buyer with the inescapable reality of the desire of the other, and the implicit struggle that that entails. Do we really want what we want, or do we want to become someone else by wanting what they want? Wouldn’t it be better to believe there was some objective form of “utility” that was the same for everyone in all cases, and that reviews could distill it for us, reassure us that we can escape from our own contingent subject position and become that transcendent “true consumer” ourselves?

When we can orient ourselves to “use value,” we don’t have to get involved with conspicuous consumption and invidious comparison and habitus and other sorts of “inauthenticity.” We can imagine that it is possible to assess the intrinsic value of things, independent of our particular need or capacity to consume them. “Real” reviews presumably are meant to help us believe that such a thing really exists. But actual online reviews seem to serve the precise opposite agenda, dissolving intrinsic value into countless subjective experiences, which are then reconsolidated in star ratings, statistics, or, now, LLM-generated summaries (or outright generated reviews, which are often more or less the same thing).

Instead of respecting the object, the object disappears and is replaced by data analyses that make it even more inaccessible. As Baudrillard claims in “The Ideological Genesis of Needs”:

The empirical “object,” given in its contingency of form, color, material, function and discourse (or, if it is a cultural object, in its aesthetic finality) is a myth. How often it has been wished away. But the object is nothing. It is nothing but the different types of relations and significations that converge, contradict themselves, and twist around it.

The object is always a pretense for these social relations, a proxy. Consumers consume reviews through the object rather than the object itself and its own qualities, and they they are invited (if not expected) to produce their own reviews as the result of their consumption. every purchase ultimately only makes an opinion. Only one use value remains possible, to produce more reviews. Are these “fake”?

“The definition of an object of consumption is entirely independent of objects themselves and exclusively a function of the logic of significations,” Baudrillard argues. The presence of consumer reviews on retail sites reproduces that logic as well as the myth of the “true object” that can be consumed in and of itself rather than as a kind of participation in the consumerist system.

Caulfield cites some research meant to show how much consumers have come to rely on online reviews:

A 2021 survey of 6,000 online consumers found “ratings and reviews have become the most important factor impacting online purchase decisions.” In fact, survey participants placed the value of online reviews above price, free shipping, brand, and recommendations from family and friends. The survey found 94 percent of customers ranked it as the single most important factor, and four out of five said they won’t shop on a website unless it has customer reviews.

A 2023 study published in the journal Nature found that “the most important attribute for consumers in selecting an online shopping service is star rating.” And studies have consistently found that over 90 percent of consumers use online reviews before making a purchase, and 88 percent say they trust reviews as much as or more than personal recommendations.

Nine of ten people trust online reviewers as much as people they know? They care more about other people’s ratings than the price or the brand? I find these statistics hard to believe; it’s as if the researchers polled people who had only ever heard of the internet and have never looked at an actual comment section. But these findings do seem consistent with some idealized notion of reviews that can be purely objective, that can allow us to partake of use value and not branding or posturing or hype cycles or any of the other demonized motives for shopping.

A false sense of inevitable consensus is implied by a review section, a “reality” that emerges from collective judgment. But online retailers (re)produce consumer reviews not to reflect reality but to produce confidence in consumers, and make them more likely to spend. The reviews are part of the vicarious satisfaction of the online shopping experience; they set up a space in which the would-be purchaser can imagine what is like to be satisfied, prefiguring the terms and criteria by which they will eventually recognize it in themselves.

Whether the reviews are “real” or “fake” doesn’t have much to do with whether or not they will invite the reader into that imaginative space. I’d guess that depends more on the retail site’s layout and other purely formal qualities of the reviews and not how much fidelity they have with respect to some actual reviewer’s actual experience. What a reader gets out of a review hinges more on their suspension of disbelief than the argumentative skills of the reviewer — much like their enjoyment of the product itself, if it is not addressing some fundamental need. And if you believe Baudrillard, there aren’t any.

It’s sort of comforting to imagine that LLM-generated “fake” reviews can pollute review sections, as though that discourse space is currently full of people being their real selves, conversing with each other in the spirit of civility. But it reminds me of how Only Fans models use chatbots to simulate one-on-one conversations with their customers, as this Vice report by Luis Prada details. The idea that the chatbot is a step away from some sort of “more real” conversation serves to re-enchant an exchange that was already fully mechanized and transactional.

“The lesson here is that if you’re paying to speak to an OnlyFans creator there is no way of knowing if you’re talking to them or talking to a synthetic version of them,” Prada writes. But the lesson is also that the person paying to speak to an OnlyFans creator doesn’t care who they are really talking to as long as the imaginative space of fantasy remains open to them. They are already committed to consuming a product, not having a reciprocal social experience, and they ultimately just need that product to balance its convincingness against its cost. They don’t care if the other person has “real intentions” beyond what they want to attribute to them, or are paying to believe can be attributed to them.

To put that another way, the buyer is committed to the fantasy of pure use value and its extractability from social relations. They want to believe in the “pure empirical object” as something they can immediately and unilaterally grasp, and they are willing to resort to simulations to preserve that belief. In a sense, the AI-generated version of a thing becomes more real than the real thing. It helps train you in suspending disbelief, in projecting your intentions onto something and misrecognizing that reflection as its true feelings. Reviews that confirm what you are hoping to believe will always seem truest.

Generative models are effective at simulating people in situations where the person has already been turned into an object. Perhaps this is why algorithmic feeds and For You pages seem to work so well. You engage with a simulation of yourself in an environment where you’ve already been made a product.