To browse the internet is to subject oneself to sophisticated and unceasing techniques of attention-capture, of which the pop-up advertisement is only the most crass and vexatious example. This paper describes the development of Nightingale, a web browser extension that fights distraction with distraction. It does this by injecting the web with pop-up ads consisting of semantically-relevant fragments of the poetry of Keats. Nightingale represents an attempt to engage in “noöhacking”—that is, repurposing the cognitively-destructive aspects of contemporary digital media in order to care for one’s own mind.’
Pharmacological Design
So you have realized, or admitted to yourself, that digital media really have reformatted your mind. You feel ill at ease in those few moments when you do not have your smartphone. You have a stack of books on your bedside table; if they are not dusty already, they soon will be. But the phone in your hand is also a book, after a fashion—an unending book that seems to adapt to your desires.
“You” in the preceding paragraph is in fact me. With effort and the proper conditions, I can still find myself in a sustained state of “deep attention” (Hayles 2007), yet the fragmentary and hectic “hyper attention” solicited by digital media seems more and more to be my default cognitive mode. But “you” may also very well be you. At the very least, the phenomenon of feeling distracted is a common enough malady that a cottage industry of books promises us ways to take back control of our attention.
Take, for instance, Jenny Odell’s How To Do Nothing: Resisting the Attention Economy (2019).1Other examples of this sub-genre include Cal Newport’s Digital Minimalism (2019) and Nir Eyal’s Indistractable (2019) This book recommends a program of resistance against the attention economy’s devious machinations. Odell’s book admirably eschews any quick fix; rather it counsels us to do the difficult work of training ourselves out of internet addiction and into a deeper appreciation of one’s own local ecology and community. What do you do, however, if you are simply not the kind of person who is predisposed to mindfulness? What if your brain has already been irrevocably reformatted by the jittering rhythms of the social media, by the smooth sublimity of the endlessly unrolling feed, by the recommendation algorithm’s actually-quite-keen suggestions?
I raise such questions not simply because I suspect that the cognitive habits that we develop as we use digital media are difficult to break. As philosopher of media Bernard Stiegler (2010) observes, all media are “pharmacological”—that is, variously beneficial or poisonous, depending on their specific quantities, qualities, and uses. Stiegler is in one sense among the most straightforwardly pessimistic analysts of contemporary technology and its effects on attention. For him, the “programming industries” use digital media to steal our attention away from all of the traditional regimes of learning that society would otherwise use to pass down to us knowledge and skills. Schools as well as families, religious institutions, and other “programming institutions” are no match for Facebook et al. and thus are no longer able to direct our attention. While this state of affairs may seem liberating, Stiegler argues that it is only through weaving oneself into and against “long circuits” of intergenerational attention that we can keep from being “short circuited”—our sense of the world and of our relationship to it contracted and impoverished.
Despite this Jeremaic gloom, Stiegler is also a prophet of hope. His vision is not that we “do nothing” but rather turn that which vitiates our intelligence into that which nurtures it. Just as the pharmacological technology of writing itself makes possible new forms of intergenerational memory even as it (according to the Platonic critique) threatens to allow us to forget, so could digitally-mediated intelligence work against the distraction and misery that digital media themselves inflict upon us. Stiegler frequently refers to this project as the “battle for intelligence,” yet this warlike language belies the fact that this battle is to be waged through acts of stern kindness, of “taking care, of oneself and others” within our current “pharmacological context” (34). Put another way, the tools of “psychopolitics” (industrial control of cognition) must become tools of what he calls “noöpolitics” (the care and cultivation of the mind/spirit, from the Greek νόος or noos). 2A related Stieglerian term is “noötechnics,” which Anaïs Nony defines as “investigat[ing] the relationship between memory and digital platforms to reevaluate the psychic milieu in which human and machine coevolve” (136). Nony provides an insightful overview of the relationship between noos and techne. To the algorithms of social networks such as Facebook, we are merely our attention, a quantifiable resource that can be bought and sold. By invoking the concept of “spirit” (something that exceeds mere cognition), Stiegler remains hopeful that the dehumanizing logics of digital media may nevertheless be turned towards the purpose of helping us to become more deeply human.
This essay describes an attempt to develop a piece of software that tactically intervenes in the ways that the internet captures attention. Some measure of autobiography is inevitable, as the software in question—a literary browser extension called Nightingale—is designed by me and for me, fitted to my own mind the way a bespoke suit would be fitted to my body. 3Nightingale could thus be considered an instance of “bespoke code” (Geiger 2014). Furthermore, what follows is not “human-computer interaction” research insofar as that discipline requires one to study the generalizable efficacy of technology, typically by testing it with various users. Rather, in the vein of “critical making” (Ratto 2011), I am less interested in presenting Nightingale as valuable in its own right than I am in recording and thinking through some of the conceptual challenges of designing tools that solicit the finite resource of my attention but that do so in order to take care of the part of my mind that is anything but quantifiable—that could be called “spirit.”
Ad-Tech for Poetry
Available for download in the Chrome Web Store, Nightingale 4Nightingale's code and instructions for running the extension on iOS can be found at github.com/kbooten/nightingale. is a browser extension that injects “literary pop-up ads” into the web. Traditional ad-tech, such as the software that tracks what websites I visit and fills portions of the page with advertisements allegedly tailored to my interests and desires, is on the side of the various firms who have paid for my attention, be they sellers of shoes, kombucha, heartburn medication, or something else. These ads are most noticeable and distracting when they occlude the staid and still prose of an online newspaper with a looping animation or video—a “pop-up.” With Nightingale, I aimed to use the nefarious techniques of ad-tech on the behalf not of capital but of poetry. If traditional web-ads distract me in the hope that I might buy something, Nightingale would distract me in the hope that I might read some poetry that I would not otherwise go out of my way to read. Since a large amount of the time I spend online is in a state of semi-intentional distraction—scrolling idly in order to space out—a distraction from a distraction could open up space to reorient my attention toward more worthwhile expenditures.
This vague intuition phrased with more technical specificity: a browser extension would somehow pull my attention from whatever I am reading on the web and redirect it toward poetry—specifically John Keats’s six 1819 odes, “Ode to Psyche,” “Ode on Indolence,” “Ode on Melancholy,” “Ode to a Nightingale,” “Ode on a Grecian Urn,” and “To Autumn.” Why these poems? First, because I wanted to spend some time with them, and I knew that I would not do so in an idle moment. Second, because I had recently seen Roman Holiday (1953) and was struck by a particular moment of banter between Audrey Hepburn’s Princess Ann and Gregory Peck’s humble reporter. Hepburn’s character, drunk, recites a bit of her “favorite poem” by Keats; Peck’s instantly corrects her, “Shelley.” In a film filled with all sorts of whimsical coincidences, evidence of this kind of familiarity with canonical (and thus, allegedly, well-known and deeply-known) authors seemed to me to be the most far-fetched occurrence.
This scene confronted me with the limitations of my own literacy—my “stupidity,” an unavoidable term in Stiegler’s philosophical work. For Stiegler, stupidity is not truly a characteristic of the individual, just as intelligence itself is not truly a characteristic of the individual. Rather stupidity and intelligence are not within our minds but are extended to our technological milieus and how we have been trained to think (and to avoid thinking) within and through them. While the history of memorizing poetry is a long and complicated one, surely my own inability to remember verse is in part a function of growing up in a context in which it is often said, even by educators, that the ubiquity and ease of Googling obviates the need for one to hold facts and even precious words inside one’s own memory. Yet stupidity itself is not without its pharmacological aspect, its potential curative component. As Stiegler observes, riffing on Deleuze, “[i]n the face of stupidity I am ashamed, and this shame…forces me to think” (32). And of course the connection between shame and technology is an old one, going back at least to that first repurposing of fig leaves into rude garments.
My fundamental intuition—admittedly a vague one—was that spending time thinking about Keats would be of obvious value to my mind/spirit/noos. One might object that Keats, of all poets, does not in fact need any more attention directed toward him. But if Stiegler is right that the traditional cultural institutions—presumably including the institutions of criticism and literary education that have shaped the canon—are losing what he often refers to as a “battle for intelligence” waged against the much more powerful forces of technocapital, for whom Keats is nothing but another authorial last name in a database of millions, perhaps Keats’s cultural significance is more tenuous than it may seem. Still, that Nightingale is a tool for re-proliferating Keats in particular is entirely a product of my own aesthetic disposition. Its mechanisms, described below, could be just as easily applied to any author or collection of authors. Since the code is open source, others are free to do just that.
Alternatives to Automaticity
Ad-tech algorithms solve an information retrieval (IR) task: given a user’s patterns of behavior (e.g. sites visited), retrieve from a list of possible advertisements the most apposite one, serving it to the designated portion of the user’s browser in real time. Sometimes the connection is immediate, as when one receives an ad for sandals on a site for booking a hotel on the beach.
I imagined an alternative information retrieval task: given a span of text on a web page that the user visits, return and display a relevant span of text from one of Keats’s poems.
The most obvious strategy for accomplishing this task would be to focus on individual words. A reference to the words “beauty” or “truth” (and certainly to both) should summon the famous peroration of “Ode on a Grecian Urn.” The major impediment to this approach is the stark difference in vocabulary between the language of the web and the language of Keats. Browsing Twitter, one is not likely to come upon terms such as “Lethe-wards” or “Hippocrene,” “beechen” or “blushful.” Likewise, common elements of contemporary verbiage (let alone terms drawn from online vernaculars) will be absent from Keats.
Such problems may seem at first well-suited to more advanced techniques of Natural Language Processing. If the language of Keats and the language of Twitter, the New York Times, etc., are too different, translate between them. Vector-space models of language, such as those produced by the word2vec algorithm (Mikolov et al. 2013), perform such a translation. For any two words in its vocabulary, the model can return a real number describing their semantic propinquity. While some specifically-Keatsian words may still not be in its vocabulary, a vector-space model may still help to bridge the gap between the two discourses of verse and general web-text. For instance, the word “casement,” rare in web-text, is similar to much less rare words such as “window,” as measured by cosine similarity between their vectors. 5The cosine similarity between “casement” and “window” is 0.43. For reference, the similarity between “casement” and “friend” is -0.004, meaning much less related. Here I use the Google News vectors (https://code.google.com/archive/p/word2vec/) via the Python library Gensim (Radim and Sojka 2010).
Then again, an information retrieval system need not focus on individual, isolated words. Given a query (for instance, a sentence from a web page I happen to be visiting), techniques of IR could also be used to summon the sentence from Keats’s poems that is overall most semantically similar. For instance, Word Mover’s Distance (Kusner et al. 2015) calculates the quantity of change that must be made to two sentences—transformed into a sequence of vectors using a vector-space model of language as mentioned above—so that they become (mathematically and, one hopes, semantically) equivalent. The lower the Word Mover’s Distance score between two sentences, the more similar they are, at least according to the algorithm.
However, just because one can use an algorithm like Word Mover’s Distance on poetic text does not mean that one should. Techniques of statistical IR such as Word Mover’s Distance may suffice as “machine reading” but they are unlikely to suffice as “machine close reading.” A good close reading of a poem luxuriates in the discovery of oblique, subtle, ambiguous meanings of a text. These meanings are often surprising and thus are “low probability.” Reading in this way is inherently antithetical to the logic of information retrieval, according to which all possible meanings can be ranked from most to least likely, with the latter discarded.
As a foil to the automatic techniques for statistical information retrieval, consider the humble regular expression, a nearly-ubiquitous technique of pattern matching. IR algorithms such as Word Mover’s Distance are automatic; one need only prepare the algorithm by giving it a query and a collection of texts to search against, and it will surface the most relevant result. To write a regular expression (a “regex” for short), on the other hand, is notoriously tedious, stubborn, and time-consuming. For this reason, there exists an ecosystem of tools that help the programmer compose them, from simple “cheat sheets” that furnish reminders of their opaque syntax to more sophisticated interfaces that provide syntax-highlighting and even machine-generated explanations of what a particular textual pattern or patterns regular expression is attempting to match. But these only dull the pain.
At the heart of Nightingale is a series of painstakingly-composed regular expressions—136 of them to be exact. Each of these is associated with a line or a few lines from Keats’s odes. Each span of text from these poems is associated with at least one regex.
From any kind of normative engineering perspective, writing so many regular expressions would be a sign of misallocated resources. However, regexes have their advantages. With them, it is possible to perform a machine reading that is much more like close reading—and in fact is built on close reading.
For instance, take the lines from “Ode on Melancholy”:
Nor suffer thy pale forehead to be kiss’d
By nightshade, ruby grape of Proserpine
One part of this line’s matching regex goes like this:
\b(nightshades?|eggplants?|
tomatoe?s?| [Xx]ans?|[Xx]anax(es)? |[Bb]enzos?)\b|\bOcean Spray\b
Eggplants, tomatoes, and potatoes are all quite literally nightshade plants, from the family solanaceae. Yet in this line, the nightshade is a poisonous thing, the “deadly nightshade” (atropa belladonna) to which eggplants and tomatoes are related. Once consumed, this deadly plant begins to shut down the parasympathetic nervous system and can induce dangerous narcosis. Benzodiazepines such as Xanax (colloquially pluralized as “Xans”) produce similar effects. But Keats is referencing not a pill but a beverage—a “ruby grape” one. Perhaps the most natural referent here is wine, but several other (regex,line) pairs already made reference to wine. (For instance, the line from “Ode to a Nightingale” beginning “O, for a draught of vintage!” will be matched to references to varietals such as pinot gris, pinot noir, merlot, and zinfandel.) Better to focus on what is distinct about this reference, its “ruby grape” juiciness, hence the reference to the Ocean Spray company and, implicitly, its grape juices as well as its ruby red grapefruit juice. The regex is thus a condensed, notational version of this sort of close reading—a reading that includes oblique connections that an algorithm is unlikely to draw (nightshade ≅ Xanax) while also ignoring other possible readings (such as the lapidary significance of “ruby”). The regex, encoding my own particular interpretations, is necessarily idiosyncratic, prejudicial, and selective—not fair and neutral as a statistical information retrieval system aims to be.
As the references to Xanax and Ocean Spray evince, writing regexes also provides an opportunity to predict the points of semantic relatedness between Keats and the more demotic and contemporary vocabulary of the web. I say “predict” because writing regexes that will match whatever web page Nightingale’s user happens to visit requires some guesswork. According to one (regex,line) pair, a “beaker full of the warm South” may match references to certain refreshing beverages as well as to name-brand vessels to hold them, such as Nalgene and Kleen Kanteen. “And for that poor Ambition!” is paired with a regex matching references to to-do apps and the Pomodoro method, a popular time management technique. “How came ye muffled in so hush a mask?” is linked with a regex that would match a reference to the N95 respirator. Other (regex,line) pairs more directly refer to the sorts of words and phrases I predicted I would encounter online by dint of my specific interests. For instance, knowing that I spend too much time on the tech-focused forum Hacker News, I encoded a connection between Keats’s line about “the wreath’d trellis of a working brain” and any reference to neural networks or related concepts:
\b(neural[ -]?network
|cognition| neurogenesis|synap\w+| dendrit\w+|decision trees?)\b
Matching specific words is a rather rudimentary use of regular expressions. With regexes it is also possible to encode complex similarities at the level of syntax or even prosody. A handful of examples:
-
- The regex
\b(those|her|his|my|our|
- refers to the line from “Ode to a Nightingale”: “Where Beauty cannot keep her lustrous eyes.” This regex matches strings like “my brown eyes” or “his blinking eyes”—a possessive pronoun, followed by a word, followed by “eyes.”
- The line from “Ode to Psyche” that references “lucent fans” is, in Nightingale’s collection of (regex,line) pairs, connected to a regex beginning \byour (bright|glowing|
shining) [b-df-hj-np-tv-z]{0,} [aeiou]+[b-df-hj-np-tv-z]+\b. This will match strings such as “your bright tooth” or “your glowing laugh” but not “your bright laughter”; this is because the bit that runs [b-df-hj-np-tv-z]{0,}[aeiou]+ [b-df-hj-np-tv-z]+ specifies a string of characters that begins and ends with a consonant with only one contiguous sequence of vowels in between. This targets words that are, like “fans,” one syllable." - From “Ode to Psyche”: “But who wast though, O Happy, happy dove? / His Psyche true!” The regex \b(\w+(y|Y|ie|IE)), \1\b matches “silly, silly” and “exactly, exactly”—a word repeated and separated by a comma, if and only if that word ends in “y” or “ie.” Here again the goal is to match the specific sound a bit of verse creates in the voice or the mind’s voice.
- The regex \b(blue|green|
black|white| silver |gold|green| red|orange|teal |gr[ea]y|pink| purple|yellow) \b[ -]\b(blue|green| black|white| silver|gold|green|red| orange|teal|gr[ea]y |pink|purple| yellow)\b matches pairs of colors such as “pink purple,” “white gold,” or (from “Ode to Psyche”) “silver-white”: “Blue, silver-white, and budded Tyrian….” - From “Ode on Indolence”: “My sleep had been embroider’d with dim dreams.” The regex \b([b-df-hj-np-tv-z]+)
\w+ \1 [b-df-hj-np-tv-z]{0,} (ea[nm]s?|oo[nm]s? |ai[nm]s?)\b matches sequential pairs of words like “dim dreams” and “silent screams.” The second word in this sequence must begin with the consonant that began the prior one (assonance). The second word must also end in “eam,” “eans,” “oon,” or another letter sequences corresponding to a small number of phonetic combinations. - From “Ode on Indolence”: “Farewell! I yet have visions for the night, / And for the day faint visions there is store.” The regex \bnights?
[\w ,'\n]{1,30} \bdays?\b matches any string containing the word “night” (or “nights”) followed by “day” (or “days”). “Night” and “day” need not be immediately subsequent: [\w ,]{1,30} specifies that they may be separated by as many as 30 characters (with some limitations placed upon the characters in order to try to capture only those spans of text that are within a single sentence, not spread out across two or more). - From “To Autumn”: “Where are the songs of spring? Ay, Where are they?” The regex Where\b
[^?.!]{5,50}\?[^?.!] {0,6}\b[Ww]here\b [^?.!]{5,50}\? tries to match two contiguous sentences, each of which begins with “Where.” (In the case of the second sentence, an allowance is made so that it can begin, for instance, “Yes, where….”)
The regular expression as a tool for performing a close reading, or an amulet in which to preserve one, has its limitations. It is not possible to reliably search for patterns such as “the word ‘heart’ followed by a present-tense verb,” which would require part-of-speech tagging. Yet the specific affordances of regular expressions can also push one to draw connections that one would not otherwise. Regular expression engines typically interpret \d as a digit. [3-6] means any digital between 3 and 6, inclusive. The regular expression \b[3-6]:\d\d (a.?m.?|
The fact that I have been able to give several distinct examples of how regular expressions can be used to match lines from Keats with substrings of other texts is itself indicative of a way that Nightingale’s mode of information retrieval is distinct from the workings of statistical information retrieval techniques. In the case of Word Mover’s Distance, for instance, the algorithm has only one way of comparing two sentences: by comparing the vectors of the words that comprise them. When writing regular expressions to hand-sew a connection between lines of poetry and some other textual snippets, one can begin naively with each new line. In this line from “Ode on a Nightingale” certain words may be important; in the next, a sound; in the next, a syntactic structure. Due to the limitations of regexes, it may not be possible to encode everything that strikes the reader/regex-composer as important about a line, but at least it is not necessary to decide what features are important a priori.
Next I will describe what Nightingale does with these (regex,line) pairs in the context of an interactive Chrome extension. However, had I somehow been derailed from this project before getting the extension itself to work, writing these regexes would not have been for naught. I began the project under the blithe assumption that I would delegate the reading of Keats to an algorithm. By the time I realized this would be insufficient, I already had enough momentum to begin the work of writing all of these regular expressions (promising myself that I could always give up after finishing only one of the odes). And so the plan to escape having to read Keats redounded as an occasion to read him with a sometimes-exhausting attention to detail.
Nightingale
Browser extensions can be used to customize the functionality of a web browser and even to modify the content of web pages the user visits. Popular desktop browsers such as Chrome and Firefox allow developers to create extensions (sometimes called “plug-ins” or “add-ons”), which can either be installed manually or distributed via online repositories. On a technical level, a Chrome extension can consist of additional .html, .css, and .js files, media resources (such as images), and a manifest.json file that defines what kinds of powers the extension has (its “permissions”) and on what pages it will operate, among other things.
Some extensions, like Adblock Plus, offer the user some helpful functionality (in this case, blocking or erasing advertisements). Others offer no obvious utility and instead aesthetically transform the experience of reading the web. Some of the most popular artistic web extensions are those that humorously replace all references to some person or concept, such as “millennials” to “snake people” (an extension with the droll name Millennials to Snake People [Bailey 2015]). The Deletionist (Montfort, Borksuk, and Juul, 2013) is an extension that cleverly deploys regular expressions to delete words from a web page, producing an “erasure poem.” Rafaël Rozendaal’s Abstract Browsing (2014) removes all text and images from a page while also changing the background colors of elements into garish hues. Thus the only information left on the page is its layout, now transformed into a crisp and blocky abstract “painting” in a Washington Color School style. Both The Deletionist and Abstract Browsing seem to anticipate the weary complaint of the internet user who scrolls and scrolls, who bounces between links, addicted yet longing for some other experience. One cannot become over-invested in some pointless argument on Twitter if the argument itself has been half eaten or fully demolished. As it often does, “art” (here, the procedurally-generated poem or painting that negates the page’s value as information) represents an escape from the weary tedium of real life, or in this case the real internet. Like Abstract Browsing and The Deletionist, Nightingale also redirects the user’s attention toward art. Nightingale, however, does not procedural-generate a poem or other artwork; rather it channels the user’s attention toward works of art that already exist, Keats’s poems.
When a page is loaded, Nightingale scans the page’s text with the intention of inserting a Keatsian pop-up ad. Iterating through its 136 (regex,line) pairs, it searches for text that matches the pattern described in one of the regexes. Should a match be found, Nightingale will modify the matching text on the screen by wrapping it in a special html <span> element. This accomplishes two things. First, the new <span> is associated with css style that makes the text glow purplish-black—a hint that there is something peculiar about these words. (In some video games, a glow effect is used to let the player know that an object is special and should be investigated.) Hovering the cursor over these words will trigger the revelation of the corresponding line or lines from Keats. This text appears in a dark square in the center of the page, obscuring the text beneath it and demanding the user’s attention. Simultaneously the user’s cursor is replaced with a small image of a dun-brown bird, the titular nightingale. Moving the cursor from the glowing words and hovering over them once again will reset the cursor’s appearance and trigger the Keatsian pop-up to slowly fade away.
The glowing text of the pop-up-triggering words is meant to make it feel special to discover this text on a page. Likewise, the avian cursor is another attempt to make the app more charming. The more charming it is, the less likely I am to get annoyed by it and uninstall it. These are not the only design decisions meant to prevent annoyance or boredom or to make triggering the pop-up sufficiently alluring. Since that which is rare is more beguiling than that which is common, at most one pop-up appears in any one page. Nightingale randomizes the order of the (regex,line) pairs so that the same page may, in subsequent visits, be filled with different pop-ups, avoiding a potential source of monotony. It also searches through the text on the page in a random order so that the glowing text can appear anywhere on the page (rather than preferentially near the top, at the first possible match).
Immediately after building the first version of the extension, however, I began to notice a problem—not with the code or its functionality but rather with my functionality as a user.
Nightingale is an extension for Chrome’s desktop browser. This meant that to use it I had to be on my laptop—an object that, not unlike its owner, had become quite sedentary during the Covid-19 pandemic, typically hooked into a large monitor and external keyboard. With Nightingale, I had hoped to intervene into those moments of digital media use when I am least active and intentional; those moments when flitting mindlessly from link to link feels most like a form of low-grade self-hypnosis; those moments that feel most wasted; those insensible moments that add up to hours. Perhaps it should have been obvious to me that most of these moments pass not on my computer but on my iPhone, an imbalance no doubt exacerbated by the pandemic. No longer able to visit the bar or the coffee shop, I replicated these spaces by lurking from the comfort of my sofa amidst the parasocial hubbub of Twitter spats—unseemly yet entrancing.
Though iOS browsers, much more locked-down than their desktop counterparts, typically forbid extensions for the mobile versions of Chrome and Firefox (one reason why the browser extension is almost entirely a desktop phenomenon), recently an extension-friendly browser was released for iOS. In fact, the browser is so new that it did not exist until months after I had finished the desktop version of Nightingale; during the intervening time, I assumed that a mobile version was impossible.
Porting the desktop version was primarily a matter of accommodating the touch interface.
Phenomenology of Re-Hijacking
Even as I painstakingly wrote Nightingale’s regular expressions, figured out how it would intervene in a web page, and packaged its code together as an extension, I still had only a dim sense of whether or not it would have any beneficial effect on my mind at all.
While slumped on the couch scrolling through some Twitter feed, darkly glowing words will sometimes catch my eye. It is far from guaranteed that I will take the time to click on these words to activate the Keatsian pop-up.
It is often said that the morsels fed to us by social media are “junk food”— delicious but (in excess, at least) unhealthful. If the outrageous tweet or click-bait headline is a potato chip, Nightingale’s interventions are more like a vaguely-healthy, but still sweet-enough, granola bar. (Though it may be a close cousin of the candy bar, one can feel virtuous about the protein, the fiber, the omega-3s.) I have noticed that I am most likely to click on Nightingale’s links when I have already binged and when the initial pleasure has faded into a dull ageusia. I am lost in a web of intentionally-provocative opinions or gossip about the peccadillos of celebrities—but then I see the glowing words. They come as a relief, snapping me out of the spell. And it promises me that I (without much effort, merely by clicking on this and not that), can recuperate the wasted time, if only a little bit. I can say without any prevarication that this was a day in which I “read some poetry” (which I construe as virtuous) rather than merely “wasted time online” (which I experience as shameful). So the first benefit accrues on a psychological rather than purely cognitive level: self-esteem is produced through the act of conforming to one’s “ego ideal” (Freud 1914).
The nature of the glowing text, promising a link (whether semantic or prosodic) to Keats’s poetry, also affects whether or not I activate the pop-up. These links may be said to be more or less mysterious. A less mysterious one: when the word “prophet” glows, I am fairly certain that clicking on it will reveal some fragment of an ode that has something to do with prophets, oracles, or other such figures. And sure enough:
“Thy shrine, thy grove, thy oracle, thy heat / Of pale-mouth’d prophet dreaming.”
Even without remembering/predicting this precise quote, I am prepared for it, especially since the language of prophets fits into my schema of the sorts of words and images that appear in Keats’s verse (along with tendrils, wine, satyrs, etc.). Other connections are more challenging to predict. For instance, seeing the word “neural network” aglow poses a question: surely Keats did not literally write about computer science, so what could the connection be? I must click to remind myself of the link my mind made between the word “neural network” and that line about the “wreath’d trellis of a working brain.” Even more mysterious are those connections between web-text and Keats that have something to do with prosody or syntax. On a tweet, the words “really, really” glow. This sounds rather informal for Keats, so what could be the connection? Triggering the pop-up satisfies my curiosity: “really, really” echoes the “happy, happy” of “O Happy, happy dove.” In general, the more mysterious the “advertisement” for Keats, the more likely I am to pay attention to it. 6As I used the app, I also noticed that it was more difficult than I would have expected to guess from which ode a pop-up was sourced. I added a new feature to aid this guessing game: after tapping on the glowing words to trigger a pop-up, tapping on the pop-up itself opens a new window containing the version of the ode published on the Poetry Foundation’s online archive of poetry.
When we say, as is common, that our attention is “hijacked” by digital media, the metaphor of terrorism suggests that these media are nefarious but also competent, even clever: the successful hijacker does not just exert force but plans ahead, lays in wait, and acts only when their target is most susceptible. With the concept of “re-hijacking your mind,” I likewise do not mean “exerting more force than the entities that have already hijacked it.” Rather, re-hijacking my mind demands planning just a bit further ahead, laying in wait just a little bit longer until both my current captor and myself would not suspect it.
Once I do trigger the pop-up, however, it is not a guarantee that I will actually read the words from Keats that now occlude the page. No doubt I will pass my eyes over them, but shifting in an instant from reading web-texts that are designed for ease of consumption to (a stray fragment of) one of the odes feels a bit like moving from the actions of the autonomic nervous system to intentional, directed movement.
That the verse is just a fragment does and does not make consuming Keats easier: reading just a bit—not a whole poem—seems manageable in the moment, yet the lack of context impairs my ability to get my readerly bearings. Here, however, the glowing words on the web-text help me to orient myself with respect to the verse. Each snippet of verse in a pop-up comes “pre-annotated” with my interpretation of it, manifested somehow in the linking text on the page. I have already done the work of analyzing Keats (way back when I wrote the (regex,line) pairs). What I must do now is simply remember the connections, “reactivating” the thoughts I have already thought—“Ah, yes ‘wreath’d trellis’ does sound a bit like a neural network” or “‘The winged boy I knew’ does put one in mind of Derek Jarman.” This process can also be recursive, as when I determine that the connection between some web-text and a line from Keats is too tenuous, leading me to go back and revise the relevant (regex,line) pair.
When I originally wrote those 136 (regex,line) pairs to match web-text to Keats’s odes, I experienced this as an extravagant use of prime work hours over several weeks. Yet now this expenditure—beneficial in its own right as a form of close reading—appears to be a kind of “principle” that continues to gain interest. When I not only pass my eyes over one of these snippets of Keats but actually reactivate the interpretation of Keats embedded in the connection between this poetry and a word or phrase below it on a web page, I allow my present self to benefit from the attention exerted by my past self, my past self to care for my present self. Encoding my close readings into regular expressions poses a trade-off: by accepting the rigid limitations of the regular expression as a tool for encoding meaning, I gained the possibility of not just preserving the fruits of my expenditure of attention but of automatically projecting them into the future, my future.
Nightingale thus transforms my original act of carefully reading and interpreting the poetry of Keats into what Yuk Hui (“Archives of the Future”) would call a “tertiary protention.” As Hui describes it, the tertiary protention is not a mere recording (which is passive and requires human intervention to reactivate it, such as when we pick up a book and decide to read it); rather, tertiary protentions such as algorithmic recommendation systems activate themselves “to the point of displacing or marginalizing active directness”—that is, to the point of usurping the human’s own autonomous use of attention.
Yet, as Hui further argues, “protention” exists in a loop with “retention” (memory). In fact, I have already begun to be able to recognize/remember certain connections between the web-text and Nightingale. For instance, coming upon the glowing phrase “obviously crazy,” I heard in it an echo: “cloudy trophy,” as in “among her cloudy trophies hung” (“Ode on Melancholy”). By reactivating my interpretation of Keat’s again and again, Nightingale creates the circumstances of repetition that are the precondition of memory. Ironically, externalizing my thoughts about Keats’s verse may (slowly, ambiently) allow me to more deeply internalize them. A morsel of evidence for this potential: when a colleague described something as “nicely artsy,” I heard an unbidden echo of “cloudy trophy” (“And be among her cloudy trophies hung,” from “Ode on Melancholy”). Perhaps one day I will be surfing the web and see the word “neural network” and think “Ah yes, neural networks, those ‘wreathed trellis of the mind.’” Or perhaps I will see a sentence beginning “Truth is,” and my mind will know that, whatever the next word may be, the writer alludes to “Beauty is truth, truth beauty.” In the theory of learning and development of Lev Vygotsky (1980), the “Zone of Proximal Development” describes those functions—“buds of development”—which a learner can perform at first with assistance but will later be able to accomplish unaided. For Vygotsky, the learner’s learning is thus first external, accomplished with the aid of social interaction and collaboration; only through this process does the learner internalize the requisite cognitive skills to be able to perform the activity solo. The “bud of development” (a function or capability in “embryonic state”) is related to but distinct from the function into which it will eventually blossom. Whenever I am able to anticipate or remember why a particular span of web-text glows, what Keatsian idiom it conjures, I—with the assistance of externalized knowledge—accomplish one function that may perhaps develop into a mature one, a pop-up of the mind composed without html or javascript. 7Hui’s notion of the “tertiary protention” is a riff on Stiegler’s concept of “tertiary retention,” itself an expansion of Husserl’s distinction between retentions and protentions. Insofar as Nightingale helps me to internalize echoes of Keats that I begin to notice without the software’s help, the tertiary protentions of the software are converted into what Stiegler would term “secondary protentions”—anticipations of perception that are within my mind, not in some external apparatus.
For Hui, as for Stiegler, that the algorithms of companies like Facebook get to preempt our own determinations of what is important enough to warrant our attention is a clear threat to our capacities to care for our own minds. However, Hui also follows Stiegler’s refusal to turn away from the digital as inherently harmful:
The resistance against the immanent pre-emptive operation of the digital society, is probably not to destroy these algorithms, but rather to consider automation beyond its exploitation in consumerism and governmentality (in the sense of Michel Foucault). (150)
Hui endorses and reframes Stiegler’s noöpolitical call to rebuild “long-circuits” of attention, thereby resisting the way that digital media companies tend to “short circuit” our minds:
A ‘long circuit’ [is] also a cultivation of an archival skill that will allow us to resist the kind of determination that is now taking shape in the treatment of archives, and to do so through new forms of archival and protentional invention. (150)
I began the process of building Nightingale by thinking that it could be a tool of “resistance” simply because it showed me content—Keats’s poetry—that was beneficial to me at the level of my mind/spirit/noos and that was unlikely to be recommended to me by any of the corporate, industrial-scale algorithmic systems that broker my attention. As a critical making project, however, developing and testing the extension was first and foremost an opportunity to think more clearly about slippery concepts, including what it would mean to use the tools that spoil attention to cultivate and nurture my own. My primary conceptual finding is this: if resistance is to take place, it must not just be in terms of content—being made to read poetry instead of tweets. It must also take place through the cultivation of “long circuits” that weave together patterns of machine and human cognition, including by projecting my cognition into the future so that it may more thoroughly become my own.
An Update: Recalibrating the Microdose
In the five months since writing the above discussion of Nightingale, I have encountered the neon purple glow of its textual injections nearly every day. Over this extended timescale, I have read more lines of Keats than I would have otherwise read, and I have grown a bit better—though not dramatically so—at internalizing and thus predicting the connection the extension has found between a web page and Keats’s poetry.
Still, I also felt the novelty wear off. Nightingale’s pop-ups became, much like the paid links tacked on to Google search results, easier and easier to ignore.
Earlier in the development of this extension, I had begun to experiment with limiting the frequency of any particular Keatsian pop-ups, assuming that seeing the same common ones too much would breed contempt. But I now began to think that encountering even a diverse set of Keatsian quotes too frequently discouraged my clicking on any of them, since I could trust that another one would appear later that hour if not much sooner. Yet the problem was not just that any one pop-up did not feel “special” enough (and thus not worth taking the time to trigger, let alone read with any care). I also began to resent the fact that the app itself did not notice or care whether or not I clicked on any pop-ups whatsoever. Whatever feelings of virtuousness I derived from pausing to read some Keats (“This is good for me. I am taking my vitamins, antioxidizing my soul against the toxic drivel of memes and gossip.”), it was not enough.
My reaction was twofold: the app would need to dramatically throttle the injection of Keatsian pop-ups, and it would need to give me some kind of “credit” for clicking on them and reading them. After some further tinkering, the app will now try to inject pop-ups only once every five days or so. Once enough time has passed, it will keep inserting them until I click on one and keep the pop-up itself open for at least ten seconds (rather than hastily dismissing it as soon as it has appeared), at which point I might as well read it. Nightingale will then wait again before attempting to serve me another pop-up.
In web-advertising, “click-through rate” (CTR) describes the number of clicks from users divided by the total number of impressions of an ad. Though Nightingale does not currently calculate such statistics, my general impression is that the revised version of the extension has so far achieved a much higher click-through rate than the previous version. When Keatsian pop-ups are rare, they feel more fleeting and thus more alluring. Furthermore, when I trigger a pop-up and read some lines from Keats, I can be contented by the fact that I am satisfying a weekly obligation, accomplishing something while simultaneously getting it out of the way, just as running on a treadmill for half an hour might allow the gym-goer to feel that they have earned a lazy weekend. At the same time, as Nightingale’s CTR has gone up the total number of impressions has gone down. I speculate that the trade-off between the rate at which pop-ups appear and the likelihood of my clicking on them would look something like this:
A much more computationally-sophisticated version of Nightingale could use machine learning in order to automatically titrate the appearance of pop-ups, maximizing the total number of click-throughs per week.
Noöhacking
Borrowing the vocabulary of Bruno Latour, Matt Ratto explains that “critical making is about turning the relationship between technology and society from a ‘matter of fact’ into a ‘matter of concern’” (259). This is to say that critical making should refuse to falsely naturalize what technologies do through us and what we do through them. The forces of technocapital have their own way of regarding our minds and our attention spans as mere matters of fact. Attention, to them, is defined operationally as click-through rate, active page dwell time, scroll rate, and other ad-tech metrics. 8These and other stats are provided by Google’s Ad Manager platform. See: support.google.com/admanager/table/7568664?hl=en. Building Nightingale continues to be an opportunity to regard myself as an ad-tech algorithm might regard me while integrating attention-economy metrics within a larger project of caring for my own mind. As this narrative of my design process has shown, self-care is also self-combat; while I have revised Nightingale to circumvent my premature boredom with it, there is no certainty that I will not become inured once more. My acclimation would demand further revisions of the software so that it would keep my attention.
As a technology or a medium, the browser extension lends itself to the practice of critical making since it makes clear that the attention-grabbing mechanisms of the web are mutable—matters of concern rather than of fact. Though Nightingale is open source and readily available online, this software is not for everyone. Some (though not all) of my interpretations of Keats’s verse encoded in the software would seem opaque to others. They are almost “inside jokes” with myself. Other people may not feel that microdosing Keats would ameliorate their digital woes. Thinking about this web extension as a countermeasure in the “battle for intelligence,” it is resigned to at most a local effect. However, what may be more useful in this battle than this piece of software itself is a hopeful recognition of the advantages that individuals have in taking care of their own minds/spirits with their own small-scale, bespoke digital interventions—an activity that we might call “noöhacking” in order to nod to its provisional, experimental approach. To digital media companies, I am an “end user.” But this assumes an ending, one predicated upon my willingness to be satisfied. Unlike in chess, here the one who moves second has the advantage; after the programming industries roll out some new technology for capturing minds, one can assess the damage, regroup in obscurity, and try to respond tactically and nimbly with new tools and practices that sneak through whatever interstices these industries have left unguarded.
References
Bailey, Eric. Millennials to Snake People, 2015. https://chrome.google.com/webstore/detail/millennials-to-snake-peop/jhkibealmjkbkafogihpeidfcgnigmlf?hl=en-US
Eyal, Nir with Julie Li. Indistractable: How to Control Your Attention and Choose Your Life. BenBella Books, 2019.
Freud, Sigmund. “On Narcissism: An Introduction.” 1914. The Standard Edition of the Complete Psychological Works of Sigmund Freud, Volume XIV (1914-1916), edited and translated by James Strachey, Hobarth, 1957, pp. 67-102.
Geiger, R. Stuart. “Bots, Bespoke Code, and the Materiality of Software Platforms.” Information, Communication, & Society 17.3 (2014): pp. 342-356.
Hayles, N. Katherine. “Hyper and Deep Attention: The Generational Divide in Cognitive Modes.” Profession (2007): pp. 187-199.
Hui, Yuk. “Archives of the Future: Remarks on the Concept of Tertiary Protention,” Inscription, pp. 129-151.
Kusner, Matt, Yu Sun, Nicholas Kolkin, and Kilian Weinberger. “From Word Embeddings to Document Distances.” In Proceedings of the 32nd International Conference on Machine Learning Research, Lille, France, July 2015, pp. 957-966.
Mikolov, Tomas, Ilya Sutskever, Kai Chen, Greg Corrado, and Jeffrey Dean. “Distributed Representations of Words and Phrases and their Compositionality.” In Advances in Neural Information Processing 26, Lake Tahoe, UT, Dec 2013, pp. 3111–3119.
Montfort, Nick, Amaranth Borsuk, and Jesper Juul. The Deletionist, 2013. https://thedeletionist.com
Newport, Cal. Digital Minimalism: Choosing a Focused Life in a Noisy World. Portfolio/Penguin, 2019.
Nony, Anaïs. “Nootechnics of the Digital.” Parallax, 23:2. (2017): pp. 129-146.
Odell, Jenny. How to Do Nothing: Resisting the Attention Economy. Black Inc., 2019.
Ratto, Matt. “Critical Making: Conceptual and Material Studies in Technology and Social Life.” The Information Society 27.4 (2011): pp. 252-260.
Řehůřek, Radim, and Petr Sojka. “Software Framework for Topic Modelling with Large Corpora.” In Proceedings of the LREC 2010 Workshop on New Challenges for NLP Frameworks, Malta, May 2010, pp. 46-50.
Roman Holiday. Directed by William Wyler, performances by Audrey Hepburn and Gregory Peck, Paramount Pictures, 1953.
Rozendaal, Rafaël. “Abstract Browsing,” 2014. http://www.abstractbrowsing.net
Stiegler, Bernard. Taking Care of Youth and the Generations. Translated by Stephen Barker, Stanford University Press, 2010.
Vygotsky, Lev Semenovich. Mind in Society: The Development of Higher Psychological Processes. Edited by Michael Cole, Vera John-Steiner, Sylvia Scribner, and Ellen Souberman. Harvard University Press, 1980.