Page 24 →1 Manufactured Processing, Ritual, and Expert Systems

Aping particular forms of science communication can have potent effects on the persuasiveness of claims about health. Quantification, for example, can offer the effect of appearing more factual by rendering content into numbers.⁴ Similarly, including trivial charts and graphs can add persuasiveness to truth claims.⁵ In a study run by Aner Tal and Brian Wansink, sixty-one people were given the same information about a fictional drug designed to reduce susceptibility to the common cold.⁶ The researchers split participants into two groups. One group got the original text, and the other group got the original text plus a graph that reiterated some of the text in a visual aid. That is, the text told the participants that colds occurred in nearly 90 percent of people who did not take the fictional drug, compared to just under 50 percent in people who took the drug. In the graph condition, the visual aid merely reiterated those numbers. With posttest survey data, the authors found that “while only two thirds of the people believed the medication would reduce illness without the graph, all but one participant in the graphs condition believed this.”⁷ After replicating the study with fifty-six participants to control for possible repetition effects on information understanding and retention, the authors also found that persons with stronger beliefs in science were those who demonstrated the most considerable effect when the graph was present. In a third study of fifty-seven persons, the authors gave each participant a description of a fictional drug, which is “carbon-oxygen-helium and-fluorine based,” but gave only one half the chemical formula: “C₂₁H₂₉FO₅.”⁸ Similar to the first two studies, participants in the formula condition reported more confidence in the drug. Despite the graphs and the formula being unnecessary for communicating the drugs’ effectiveness or composition, the inclusion of such trivial elements increased the confidence of study participants. In their conclusions, Tal and Wansink compellingly describe the effects of including trivial components in one’s science communication as science “signaling,” which works by playing off of Page 26 →elements—like graphs and formulas—to associate claims made with scientific objectivity, in turn, augmenting the authority of those claims.

In the context of automation, making one’s argument “move” by retrieving data and returning calculations to a user, in other words, can be done trivially, and, as I will demonstrate in the case of Vaccine Calculator, machines can perform as agents of legitimate science and public health, but without actually being a part of those spheres. In such cases, the machinic agent instantiates a manufacturing of legitimacy by way of emulating the rituals of interface one might commonly associate with scientific expertise, enlivened by the rhetorical energies of computational performance.

Theorization about visual rhetorics of science and technology offers a useful proxy for adding depth to understanding manufactured processing—namely, its implications for enhancing ethos through ritual. Lynda Walsh, in her Scientists as Prophets: A Rhetorical Genealogy, illustrates the “prophetic ethos” that attends climate model visualizations. In Walsh’s frame, worn into the grooves of historical trajectory, since the days of the Greeks, we have looked to “oracles” to see the truth—even in science.⁹ Within the mimetically repeated trope of the prophet, there are also technologies of prophecy, such as climate model visualizations, which help to discern the ether of data and dynamism that constitutes climate change. It is in this way that climate change modeling plays off of the longstanding trope of the prophetic ethos. When we look at a climate model, we are looking at the work of seers, reporting back, after they have “read the bones.”

Within this frame, we can grasp that computational performances represent “more” than just the data; they instantiate the animation of ecologies of ideas, shaped by practices and habits entangled with historical echoes of expectation that still resonate with practices of science, like the trope of the prophet as it resonates with the automation of data retrieval and analysis. To further account for the deep end of computing that matters to the case of Vaccine Calculator, I will take up what Clay Spinuzzi has articulated as interface archaeology.¹⁰ In this analytic approach, one dives into the meanings and effects of a given interface, by retracing the evolutions of similar interfaces by locating antecedents of cultural practice within, and around, those interfaces. For example, analyzing a geographic mapping interface not simply by focusing on the interface itself, but also by accounting for the genre ecology that matters to that interface (e.g., by looking to historical designs of mapping interfaces and how they have evolved over time, in step with other technologies that preceded them, like colored pins or paper maps).¹¹

Page 27 →Because the current study is interested in the deep ends of computing—the energies of computational performance, manifest not just in the front end, but between the front and the back ends—I will place attention on the interface of Vaccine Calculator, but also its back-end processes. By doing so, we will be doing not just an interface archaeology, but an archaeology of computational performance, for genres of computational performance are constituted not only by visual representations but also by manners of movement (as in when a machine reacts to input, based on its programming on the back end). Furthermore, I will show that the computational performance of Vaccine Calculator is entangled with cultural practices—rituals—associated with science communication and human–computer interaction, enlivening claims about health science by animating them with machinic movement. Specifically, I will draw attention to the procedures of Vaccine Calculator and how they mime the rituals of legitimate health science while performing as an “expert” system, inviting users to attune to “expert” knowledge with a misplaced contentedness, empowering them to find false legitimacy in pseudoscience. For now, we turn to begin describing some characteristics of the genre ecology of knowledge-based systems.

It takes only a small interaction with Amazon’s Alexa, who answers your voice-based query about what the proper name for a group of unicorns is—“A group of unicorns is called a ‘blessing’”—or a brief session with IBM’s Watson, which, according to its automated sentiment analysis, informs you that your academic writing is low on “joy” but high on “anger,” to realize that we look to machines for answers.¹² The disappearance of machine communicators into the background of our everyday assemblages of public life is striking, despite their relatively recent history existing only in practical terms as mockups and props displayed on the soundstages of sci-fi programs, like Star Trek: The Next Generation.¹³ The resources required to make such machines are minor, especially in contrast to the teams of experts, expensive equipment, and trial and error traditionally necessary to creating such systems.¹⁴ In other words, the genre of software agents that is referred to as expert systems is a genre that is increasingly accessible to the wider public, beyond the traditionally exclusive realms of technical action and discourse.

All one requires to create a “knowledge-based” system (sometimes called an “expert system”) is a knowledge base (derived from content experts) and an Page 28 →inference engine (a computer program that follows “scripts” to make use of the content).¹⁵ Although this might appear daunting to some readers, and while it might require some working knowledge of coding, it is important to underscore the ease with which one can make such a system. If someone already has even an entry-level understanding of coding and they were interested to make their own “expert system,” they could easily have an inference engine up and running in a couple of hours, after reading a few of articles, watching a You-Tube video, and copying and pasting someone else’s code (e.g., see Chatterbot below). And, using that, one can design a system to work from any knowledge base, including knowledge bases that are not derived from actual experts. The scripting of a given knowledge-based system can be simple or complex.¹⁶ For example, one could build a more straightforward knowledge-based system with a few simple “conditionals” that simply return values prescribed in a database, but which adjust to inputs (e.g., taking a user’s input of “age” and adjusting a vaccine schedule to give suggested dates of vaccination). Or one might build a system that uses machine learning to “discover” and illustrate patterns in a large data set, outputting suggestions for closer inspection by a human user. While knowledge-based systems that employ machine learning can exhibit more complexity, building one is made easier with the use of third-party services (e.g., see IBM’s Watson above) which provide pre-built machine learning tasks in an on-demand fashion, without requiring the coder to know how to make their own machine-learning system.

Technically speaking, because knowledge-based systems replicate the thinking and analytic processes of humans, they instantiate artificial intelligence. However, as computer scientists Rajendra A. Akerkar and Priti Srinivas Sajja explain, there are two types of artificial intelligence. Some systems instantiate symbolic artificial intelligence—preprogrammed rules designed to emulate human thinking and action in a rigid way.¹⁷ “How much philosophy is too much, as per my body-weight?” There are also systems that use statistical probabilities to map “fuzzy logical” relationships between variables. These would instantiate connectionist artificial intelligence. “Computer, can you tell me if this statement is from a person who has had too much philosophy?” It is imperative to acknowledge that, in actual practice, knowledge-based systems are often mixtures of symbolic and connectionist elements.¹⁸ For example, someone could foreseeably “train” a neural network—a connectionist model—on medical textbooks and retrieve information from that database in the case that the preprogrammed rule set for a given response was unable to retrieve an answer. Conversely, one might symbolically preprogram specific responses Page 29 →that are facilitated by connectionist natural language processing to get the gist of a user’s inquiry, retrieving content based within a threshold of confidence (e.g., see Cox’s Chatterbot as an example of such a system).¹⁹

Vaccine Calculator is a pseudoscientific knowledge-based system that can be categorized as a simple instance of symbolic artificial intelligence in that it operates on simple rules, rather than connectionist mathematical models. Because Vaccine Calculator does not exist in a vacuum, but rather as emerging from existing genre ecologies, it is useful to retrace similar web applications as they are used within the spheres of legitimate health science. An example of a web application representing legitimate health science is the Catch-Up Vaccination Scheduler, a first-of-a-kind tool born out of a collaboration between the Centers for Disease Control (CDC) and the Georgia Institute of Technology.²⁰ The tool was designed to help simplify and streamline the personalization of vaccine scheduling by drawing on current guidelines of vaccination, and automatically coordinating inputs about a particular person, such as age and current vaccine status, to output a schedule of vaccine recommendations for that person.²¹ The Catch-Up Scheduler largely instantiates symbolic artificial intelligence in that it involves referencing current data and then sculpting an output of that data based on user input. Take for instance a snippet of the JavaScript program that runs the Catch-Up Scheduler (replete with the programmer’s comments embedded within the code, following “//”s). Basically, it describes the retrieval of vaccine data, designated as the variable vdb, and then parses that data so it can be returned to users, based on their inputs.

$http.get(vdbFile).then(function(response) {

var vdb = response.data

//initialize vaccine data list so that we can populate it fresh with data from the vdb file.

factory.vaccineDataList = [];

//loop through the vdb json and get the data we need

Object.keys(vdb).forEach( function(key) {

var vacJson = {};

vacJson[‘vacId’] = vdb[key].id;

vacJson[‘vaccine’] = vdb[key].shortName;

vacJson[‘description’] = vdb[key].longName;

vacJson[‘numDoses’] = vdb[key].doses.length;

vacJson[‘vaccineUrl’] = vdb[key].vaccineInfoUrl;

factory.vaccineDataList.push(vacJson);

Page 30 →//Only override the vacAdminDates if they do not exist

if(!(factory.vacAdminDates[vdb[key].id])){

factory.vacAdminDates[vdb[key].id] = [];

}

});

};²²

The Catch-Up Scheduler was designed to help parents and physicians update missed vaccines for both children and adults. The particular sets of data points (e.g., types of recommended vaccines), as well as the processing of those data points as per particular inputs (e.g., the age of a child, which vaccines have already been administered) can be nebulous. As such, the application was designed to help simplify the otherwise complex and potentially overwhelming sets of variables involved in creating a vaccine schedule, which includes timings of vaccination, sequencing of vaccination as well as the possibilities of discretionary movement of those timings and sequencing for a particular patient.²³ Furthermore, the processing of those data points is done in line with rules, garnered from content experts. Specifically, the Catch-Up Vaccination Scheduler has been designed in accordance with the “Childhood and Adolescent Immunization Schedule,” derived from deliberations among members of the Advisory Committee on Immunization Practices, The American Academy of Pediatrics, the American Academy of Family Physicians, and the American College of Obstetricians and Gynecologists.²⁴ Eventually, we will see that the rhetorical strategy of Vaccine Calculator, a similar, but pseudoscientific web application, is to emulate the performance of such an expert system, but independently of mainstream science, offering science ritual without scientific credentials.

For now, we can say that the rhetorical effect of the CDC’s Immunization Scheduler is in the fact that it affords agency to users to sidestep the “all at once” anxiety that some experience when it comes to the complicated barrage of variables involved with coordinating vaccine schedules. For instance, some parents are not necessarily dubious about the effectiveness of vaccines, but they might question the health implications regarding how many and at what time to administer a vaccine. The CDC’s Immunization Scheduler empowers users to participate in vaccination, offering further clarity, realized through the movements of a computing machine, encouraging a contented attunement to vaccination, wherein the parent is making choices, not about whether they will vaccinate, but rather when and how, by receiving personalized information for their child, based on expert data regarding immunization schedules. Differently, for physicians, the application might represent a time-saving tool for generating vaccine schedules with patients. The Catch-Up Vaccination Scheduler takes the user through a series of steps. The first solicits a child’s age and name.²⁵ Then the application returns a list of vaccines for the user to select as already administered vaccines.²⁶ Finally, a report is generated that includes a color-coded schedule for administering vaccines, including retroactive vaccine doses in red as “catch-up doses” or “CD” (figs. 1.1 and 1.2).²⁷

Extended Description

Landing page of the Catch-Up Immunization Scheduler. The page depicts a doctor and young child smiling during an office visit alongside a statement that reads, "Get a personalized 2018 vaccination schedule for children ages birth-18 years old." Accompanying that statement are a series of steps: "1. Enter your child's name and date of birth; 2. Enter your child's vaccination history and generate a vaccine schedule; 3. Print and save vaccination schedule." Just beneath the steps is a button that reads, "Get Started."

Extended Description

Calendar view of Catch-Up Immunization Scheduler output. The image depicts a timeline from 0 weeks to 15 months, over which time periods for adminstering the hepatitis B; rotavirus; and diphtheria, tetanus, and pertussis vaccines are indicated at different periods over the course of the timeline as "administered dose" (in blue), "catch-up dose" (in red), and "on-time dose" (in yellow).

Physicians who possess the wherewithal to “double-check” the outputs, might treat the web application as a tool for streamlining the decision-making Page 32 →process with patients. For patients, on the other hand, who might be less privy to current vaccine schedules and might not be programming savvy, the web application’s actual calculations and database “under the hood” are mostly inaccessible, meaning that these users have to take the app at its word. And it is precisely because of this that the effect of the application, might invite an ELIZA effect, an effect that Noah Wardrip-Fruin describes as existing in those moments where users ascribe complexity to a system, based on front-end indicators of a given application, rather than understandings of what is actually happening on the back end. The effect is named after ELIZA, an early, famous chatbot that could engage communication in a seemingly open-ended manner, adapting to user input as a given conversation evolved over time. Drawing on an array of scholarly accounts of ELIZA from within human–computer interaction, Wardrip-Fruin explains that, despite being relatively simple in its programming, users who interacted with the bot nonetheless demonstrated a tendency to ascribe complexity to it. But, as Wardrip-Fruin elaborates, the ELIZA effect is subject to breakdown, based in the design parameters of the interface: “the illusion that something much more complex was going on inside the system (a human considering her problems seriously and answering questions thoughtfully, rather than random yes/no answers) [was able to be sustained] because the scope of possible responses was so limited. If it had been expanded only slightly—say, to random choice between the responses available in a ‘magic eight ball’—almost any period of sustained interaction would have shattered the illusion through too many inappropriate responses.”²⁸

After one submits their data inputs to the Catch-Up Vaccination Scheduler, they are, for all intents and purposes “shaking the eight ball,” interacting only with its outputs, seemingly inviting “more” than the impression of simplicity. Between the inputs of the user and the outputs of the system is the performance of the machine, carrying out its processes, in this case, vetting the ritual. The computational performance of the Catch-Up Vaccination Scheduler represents the energies of medical institutions and software engineers, wrought through the movements of a computing machine, carrying out calculations and referencing the most up-to-date data. “You will converse with an automated agent, who, of course, you associate with the deliberations of experts.”

Clifford Nass and Youngme Moon offer the framework of mindlessness a for accounting for how, counterintuitively, humans tend to apply social scripts to computers, even when they are fully aware that the entity they are interacting with is a machine.²⁹ Saying “please” and “thank you,” to a machine, moreover, is normal, based on mere routine and basal cues of interaction. Conversely, more Page 33 →involved interactions that take forethought—like dealing with a complex emotion on the part of one’s interlocutor—are less likely to invite mindless engagements of social script on the part of a human interacting with a computer. Important to Nass and Moon’s contribution is their answer to anthropomorphism as a useful explanation for the phenomenon of applying social scripts to machines. Persons do not say “please” and “thank you” to a machine because they think it is human. Rather, people treat computers with polite interaction because they are not thinking about it—they are operating in rote, being mindless. Such rote scripts can be leveraged in ways that draw on deep genre ecologies, such as those characterized by knowledge-based systems and health science, carrying with them rituals that punctuate the energies of computing to offer a feeling to the user that the information being presented is reliable. Web applications, like the Catch-Up Scheduler, moreover, enliven communication at a nonconscious level, by speaking to human concerns with more-than-human energies.

After running from 2008 to 2020, the Catch-Up Scheduler was decommissioned.³⁰ In its place is the Child and Adolescent Vaccine Assessment Tool, a similar but simpler application still in use at the time of this writing (fig. 1.3).³¹ Although its interface is similar to the Catch Up Scheduler, the newer Adolescent Vaccine Assessment Tool suggests vaccine names rather than output a vaccine schedule, alongside advice to consult one’s doctor about obtaining those vaccines. In either case, the Catch-Up Scheduler and the Child and Adolescent Vaccine Assessment Tool both demonstrate examples in which the energies of computational performance can be leveraged to enliven legitimate health science. The recommendations of experts, in other words, are animated by the computational movements of these web applications, shaped by the rituals of interface interaction associated with the genre of public health knowledge-based systems as well as the longer standing trope of technologies of prophecy.

Although the apps referenced above are aligned with institutions of legitimate health science, one can enact the rituals traditionally associated with the league of scientific experts without actually requiring initiation as an expert. Experts in computer science and programming are not necessarily experts in anything beyond that. Even then, one does not have to be an expert in programming to cobble together existing libraries of code or to copy and paste others’ computer coding projects, tweaking a few minor details to call it one’s own. Despite this, one can use programming to create an automated agent that performs rituals of mainstream science, but without credentials to forward an argument to undermine mainstream science, beyond words, and beyond the human. In doing so, the ritual of interacting with a knowledge-based system can be mimed within a computational performance, in turn cultivating a contented attunement to false medical advice based in pseudoscience.

Extended Description

Input screen of the CDC Child and Adolescent Vaccine Assessment. The screen includes multiple-choice options and dropdown menus for inputing child age, gender, travel plans, and immunological status (e.g., if they are immunocompromised). At the top of the screen is a list of steps for the user to follow: "1. Answer the questions below; 2. Get a list of vaccines your child may need based on your answers (this list may include vaccines that your child has already had); 3. Discuss the vaccines on the list with your child's doctor or healthcare professional."

What I hope to show in the following section, by untangling the social and technical scripts built into the pseudoscientific Vaccine Calculator, is that one can construct a mindless experience, which leverages the rhetorical energies of computational performance while emulating the ritualized procedures of interaction attendant to the tradition of knowledge-based systems, giving one’s arguments a manufactured, but forceful, sense of legitimacy, spoken quietly through the movements of an automated agent, designed to feign a performance of legitimate health science.

Vaccine Calculator is a web application, designed by a volunteer at the National Vaccine Information Center, an organization dedicated to raising suspicion Page 35 →of the safety of vaccines, framing it as “informed consent.”³² According to the application’s homepage, Vaccine Calculator is “a simple tool to help make the calculations. . .. as a fundamental part of [peoples’] research to make the best decisions for themselves and their families.”³³ It bears a striking resemblance to the CDC’s Catch-Up Vaccination Scheduler.³⁴ But where the user experiences the Catch-Up Scheduler as an opportunity to more clearly understand the advice of experts, Vaccine Calculator emphasizes nonexpert understanding, an emphasis on patient-empowered “research.” Where the CDC’s Immunization Scheduler is built from the input of panels of medical science experts, Vaccine Calculator is the result of a single person who is aligned with the National Vaccine Information Center, a notoriously anti-vaccination leaning organization.³⁵ This person might possess expertise in web development but lacks credentials as an established expert in vaccine science. Nonetheless, according to the promotional copy written into the landing page of the application, Vaccine Calculator offers information for “hundreds of thousands of families from around the world” to use “as part of their informed vaccination decisionmaking research.”³⁶

Written in JavaScript, Vaccine Calculator generates outputs dynamically based on input data from the user, including “name,” “age,” “weight,” and “allergies” to “egg” and “gelatin.”³⁷ After the user has input their data and initiated an onclick event by clicking the submit button, a file loads containing thousands of lines of code. The main script for the application references a structured data set, which outlines specific vaccines and their corresponding values and then dynamically outputs ingredient data to the next page in correspondence to the data input by the user.³⁸ The script works from a data set, which (from looking at a file entitled “vic.csv” located on the “VaxCalc-Labs” public GitHub profile) seems to break out individual vaccines by “vaccine_type,” “brand,” “manufacturer,” whether a vaccine has been “discontinued,” the “name” of a given vaccine ingredient, “units,” and units of measure, labeled as “uom.”³⁹ Included in table 1.1 is an extract of the Hepatitis B vaccination entries. After the user clicks the submit button, they are brought to a second page and given a list of twenty-three checkboxes next to specific vaccines. Table 1.2 depicts this list.⁴⁰ Some of the vaccines in the list are accompanied by small .png image files, depicting an unborn fetus or a viral spore, meant to designate the presence of “human protein/DNA” or “live virus”—these are also indicated in the table.⁴¹

When the user clicks the checkbox next to a given vaccine (e.g., the “Hepatitis B” vaccine), Vaccine Calculator generates a report toward the bottom of the page.⁴² The report corresponds with the data input by the user. For a thirty-year-old person weighing one hundred pounds who selects the Hepatitis B vaccine, the application generates a bar graph, depicting “injected aluminum versus possibly safe for [name of user],” presented to the user in the form of red and green bars: “500mcg injected” vs. “180mcg possibly safe.”⁴³ The very same input data with only the weight changed to two-hundred pounds and the same selection of the Hepatitis B vaccine gives this result: “500mcg injected” vs. “360mcg possibly safe.”⁴⁴ Changing the age to 5 years old, 100 pounds, changes the output for the Hepatitis B vaccine to “250mcg injected” vs. “180mcg possibly safe.” The results track along with the data set entries for the “pediatric” and “adult” Hepatitis B vaccine aluminum contents of the “vaccines-ingredients-data” (table 1.1). Based on testing, the pediatric results appear to stop at nineteen years old, and the adult results begin with entries of twenty years old.

After analyzing and testing the inputs and outputs of the system, I have found it is evident that the results are not very sophisticated. The only Page 39 →meaningful shift in the information returned is in the difference between pediatric and adult doses. This information could easily be communicated by way of a table in a brochure or something of the sort; automation is not required to return these results. An answer to why someone might write a web application where one is not necessarily required is in the rhetorical energies of the computational performance. Enhancing the content of the table is the performance of an automated agent, “calculating” the results and returning them to the user in a manner that emulates the knowledge-based systems of legitimate health science. This is a pseudoscientific argument, given legitimacy by the movements of a machine.

Technically, Vaccine Calculator is not reporting false aluminum contents. According to the Engerix-B (Hepatitis B vaccine) insert, “Each .5-ML pediatric/adolescent dose contains 10 mcg of HBsAg adsorbed on 0.25 mg [250 mcg] aluminum as aluminum hydroxide.”⁴⁵ What is deceptive is the interpretation of that number. As the personalized Vaccine Calculator report states: “It is possibly safe for [user name] to receive 180 mcg per FDA recommended maximum daily aluminum dose of 4 to 5 mcg/kg/day to prevent accumulation and toxicity.”⁴⁶ Anna Kata, in her study of the common argument points of vaccine denialists—and with regard to Vaccine Calculator specifically—explains that such an interpretation is misleading. Vaccine Calculator, by “comparing aluminum from a one-time vaccine dose to the daily estimated safe dose based on chronic, long-term exposure” is “making the vaccine dose appear dangerous.”⁴⁷ Because Vaccine Calculator implies the movement of a machine in the background “running the numbers,” the user is primed to believe that they are getting “facts” from which to draw their conclusions.

The process of inputting the data, clicking Submit, and awaiting a report on that data is a ritual associated with the practice and legitimacy of science. In this way, Vaccine Calculator invites the user to participate in the rituals of science, to manipulate outputs and to try different permutations, “including the ability to compare ingredients for the different brands of the same type of vaccine.”⁴⁸ Being able to run different simulations and to be able to examine the outputs equips the user with just enough information to be able to feel competent enough to make a conclusion. In reality, the user is being given facts, reported out of context. Contrasted with the CDC’s Catch-Up Scheduler, which was largely about relying on experts to “filter” out the clutter of possible data points so as to suggest a coherent plan of vaccination, Vaccine Calculator presents data to the user as a “wall” of information, affording the impression that Page 40 →the user is receiving “unfiltered” data, further exposing the general suspicion of expertise that undergirds the experience of the application. “I can figure this out on my own! I don’t need your agenda!”

In light of Heidi Yoston Lawrence’s insight that “fears of vaccination operate as more than just a set of discourses that express worry or conspiracy theories among the public but rather reflect the unknown as a material exigence,” the leveraging of the rhetorical energies of knowledge-based systems in Vaccine Calculator can be seen as making an embodied appeal to an embodied concern.⁴⁹ Moreover, the concern is one of fear regarding the unknown threats of vaccination. As Brian Massumi articulates: “What is not actually real can be felt into being . . . fear is the anticipatory reality in the present of a threatening future. It is the felt reality of the nonexistent, looming present as the affective fact of the matter.”⁵⁰ In the same way that a fire alarm can charge the body to be vigilant of the threat of a fire (even when there is no fire), Vaccine Calculator offers a “performative,” or something that strikes the body as an indicator of threat.⁵¹

Put differently, the rhetorical effect of the Vaccine Calculator exists in its procedural reification of the existence of a threat (e.g., by repeatedly pointing to vaccines as dangerous), while augmenting with the energies of “calculation,” inviting regard for the threat of vaccination as an affective fact. “If the data is returned, then the threat must exist!” Moreover, Vaccine Calculator represents a nonhuman performance, offering the appealing feeling of knowing, activated in the body by the movements of a computing machine, that is apparently just “running the numbers.” This is especially so for those who might otherwise approach the risk / reward calculus of vaccination with a general sense of doubt.

Vaccine Calculator is a rhetorical construction that procedurally encourages the user to rationalize the threat of vaccination as real while augmenting with energies, punctuated by the prophetic rituals of expert systems to afford the feeling of staying “ahead of the threat.” The Vaccine Calculator operates by cultivating an attunement that encourages a Dunning-Kruger effect in the user. The Dunning-Kruger effect is the phenomenon in which persons can overestimate their abilities within a given knowledge domain due to a lack of understanding in that domain.⁵² Vaccine Calculator encourages users to feel false confidence regarding their understanding of vaccine safety by inviting them to “play the part” of the expert. By having an “expert system”—an automated agent—“report” the “data,” the user can then apply their own self-knowledge, enacting their own agency as though they have competence in a particular Page 41 →knowledge domain by endowing them with empty rituals of expertise, enlivened by computational movement.

Justin Kruger and David Dunning explain that some understanding of a given knowledge domain is necessary to allow for the overestimation of one’s own competence. According to Kruger and Dunning, “in order for the incompetent to overestimate themselves, they must satisfy a minimal threshold of knowledge, theory, or experience that suggests to themselves that they can generate correct answers. In some domains, there are clear and unavoidable reality constraints that prohibits this notion. For example, most people have no trouble identifying their inability to translate Slovenian proverbs, reconstruct an 8-cylinder engine, or diagnose acute disseminated encephalomyelitis. In these domains, without even an intuition of how to respond, people do not overestimate their ability.”⁵³

Restated in terms of the energies of computational performance, the movements of the machine activate confidence, supported and punctuated by the rituals of knowledge-based systems and public health but ported to the expert-suspicious logics of vaccine denialist discourse. The user is invited to feel that they are adept at drawing conclusions regarding the safety of vaccination, when in fact they might not possess any actual competence in that knowledge domain. The ritual of Vaccine Calculator is not an authentic one; it is out of touch with scientific consensus regarding vaccine safety, risk, and benefit. Vaccine Calculator exploits instances of “meta-ignorance” about the topic of vaccination by supplanting with the movements of a machine to cultivate an attunement, a feeling of self-satisfaction in knowing that one is ahead of the threat of vaccination.⁵⁴ Built from scripts and databases and resonances with wider commitments to scientific suspicion, Vaccine Calculator performs as a machinic agent that inspires users to imagine themselves as experts, to identify with the rituals of legitimate health science.

Hillary A. Jones’s discussion of identification in interactive, new media environments helps explain the process of identification that makes the Dunning-Kruger effect of Vaccine Calculator possible.⁵⁵ In particular, Jones argues that in online social media platforms like Pinterest, Instagram, Facebook, and Tinder, persons not only identify in the substance of the communications they share but in the practices of sharing—the energies of ritual. “Pinning, gazing, hashtagging, or swiping” are instances of human-computer interaction with which persons can identify.⁵⁶ Moreover, I can identify with my fellow “pinners” even if we are not pinning the same things to our boards. We share in identification, not because we share an “essential identity” but rather “shared Page 42 →forms and procedures”—we share in the ritual of “pinning.”⁵⁷ The procedures of Vaccine Calculator are those that emulate other knowledge-based systems of public health and vaccine science by taking input data and returning a personalized report. By emulating these procedures, Vaccine Calculator enacts the rituals of legitimate public health and vaccine science, blurring the lines of difference between the rituals of legitimate science and a construction built to pantomime those rituals. Furthermore, by having the user engage the rituals of legitimate public health and vaccine science, they are being invited to identify with those realms of practice, equipping them with the (embodied) semblance of understanding needed to overestimate their competence to evaluate the scientific data that they are “analyzing”—to identify with experts of vaccine science, including “expert” machinic agents.

Moreover, because the interaction encouraged by Vaccine Calculator is one associated with the “muscle memory” of practices that can be associated with public health and vaccine science, the user is being primed to respond less cerebrally than if they were to locate the ingredients of a given vaccine, locate its tolerable aluminum content, and calculate the risk themselves. With respect to automated interface design, the interface of a knowledge-based system can call for “analytical cognition” on the part of the user, where connections and reasoning are necessary—the human needs to make sense of the computer outputs.⁵⁸ By contrast, an interface can also ask the user to respond with more “intuitive cognition” wherein the user responds more instinctually, based on recognition of patterns.⁵⁹ As demonstrated by other literatures on the “dual process” model of reasoning, analytic reasoning—also known as “System 2” reasoning—is activated in moments of “experienced difficulty” or “disfluency.”⁶⁰

Because Vaccine Calculator’s performance is moored in the genre ecology of “expert” systems, an ecology that can be easily associated with legitimate public health and science, it can be said to include familiar cues that help the user feel “in the know,” in turn encouraging System 1 reasoning at the instinctual end of the reasoning spectrum, making them comfortable with information they might not actually know anything about. Key to this effect are the rhetorical energies of the computational performance itself, in which the user is asked to participate with, and inadvertently identify with, the practices of experts (including expert automated agents), enlivened in the movements of the machine. Via manufactured processing, Vaccine Calculator is able to convince against scientific consensus, by way of appropriating the social and technical scripts—the rituals—one associates with the authority of scientific consensus and embedding them within a lively computational performance. Page 43 →Put succinctly, Vaccine Calculator is a computational performance, characterized by the rhetorical tactic of manufactured processing, or the leveraging of the energies of computing machines to fabricate legitimacy for claims that are indefensible with respect to scientific consensus. In particular, Vaccine Calculator moves with the energy of an oracle to offer the “expert” advice to ignore the experts, beyond words, and beyond the human.

Aimee K. Roundtree helps us further distinguish the difference between legitimate and manufactured processing in her examination of the rhetoric of computer simulations. In her thoughtful consideration of “virtual evidence,” like visualizations of bumblebee flight and weather patterns, computer simulations manifest not as direct observations, but rather representations that nonetheless offer energy, an experience with the data “at work” (energeia)—the bringing of a phenomenon before the eyes, by way of looking across data points but not necessarily the phenomenon directly.⁶¹ Such modeling is necessary in cases where the data points are so numerous or complex that they are difficult to perceive without first being modeled or filtered. The flights of small, industrious insects are difficult for the human eye to track in any meaningful way. And thus, Roundtree teaches us what is unique and powerful about computational media within the realm of science communication: it can bring knowledge and ideas “before the eyes,” despite complexity. Conversely, as I hope that the current chapter has illuminated, the lively movements of computational performances—“running calculations”—can also be leveraged to enliven science communication—in some cases, illegitimately. Vaccine Calculator, moreover, enlivens pseudoscientific claims with the energies of machinic movements, animated by a deep end of computing permeated by rituals of knowledge-based systems as well as the longstanding trope of technologies of prophecy. Procedurally, Vaccine Calculator instantiates an argument, but vitalizing that argument are the movements of the machine between the front and back ends.

Because manufactured processing relies on the sorts of mindless associations between machines and science, it emerges as a troubling rhetoric of empowerment, wherein users are inspired to exercise their own agency. However, instead of being made comfortable with the conclusions of groups of experts, the user is made comfortable with conclusions that circulate within a problematic discourse ecology known to be suspicious of scientific consensus. Page 44 →Such a happening can be juxtaposed against the phenomenon of citizen science, or the use of digital technologies to facilitate the participation of laypersons in the process of scientific inquiry (e.g., collecting data using sensors or participating in the analysis of data). Citizen science is sometimes conjectured to possess the potential to enhance publics’ capacities for scientific thinking as well as their identification with scientists.⁶² Rhetorically speaking, then (and as James Wynn points out), citizen science might be leveraged to counteract such phenomena as science denialism (although, as he would be careful to note, it is not a surefire strategy).⁶³ Vaccine Calculator invites laypersons to a similar experience—to feel like they are taking part of the rituals of science—when, in fact, they are doing pseudoscience. Consequently, the result can be said to be deleterious to people’s capacities for scientific thinking, by leveraging a strategy that feels very similar to that of citizen science models of communication, but most emphatically lacks scientific credentials.

Manufactured processing is an alchemy of scientific legitimacy, which allows the user to equivocate the realms of self-knowledge with scientific knowledge. In this way, manufactured processing is not merely a reification of the idea that procedures can be used persuasively. Rather, manufactured processing teaches us that the energies of computing, at least insofar as they are associated with the rituals of interaction associated with self-moving machines endemic to particular knowledge domains, can be appropriated to add scientific legitimacy to claims that are in actuality pseudoscientific. It is when we put concepts of computing in conversation with rhetorical practice that we realize that, beyond the front ends (interfaces) and the back ends (databases) of computing systems, there is also a (mindless) deep end of computing, where the rituals of oracles and prophets interact with the genre ecologies of knowledge-based systems, animating some computational performances, such as Vaccine Calculator.⁶⁴

This chapter pursued the epistemic end in an interactive computational performance. In the next chapter I examine the aesthetic end of rhetorical energy in a computational performance that does not involve user interaction per se, but nonetheless imbues sublime energies through the processual magnitude of vast computing.