What happens when a bus, a gun, and an algorithm walk into a search bar.
By Kat Chen
Like many of Columbia’s New Yorkers who flew here (as opposed to those who grew here), I take the M60 bus from LaGuardia to our campus. Arriving at the city and leaving it at the bookends of each semester, I often find myself out on the airport sidewalk in the winter, when New York morphs into the concrete tundra. Just as the chill seemingly multiplies any given distance, it too converts the minutes spent outside into hours and so my regular homecoming ritual with the city plays as follows: With numbed frozen hands, zip down the chest pocket of puffer. Dig past loose bills and receipts to find phone. Huff a visible fog of breath over fingertips so the phone recognizes my hand as human. Open Safari and Google the M60. Scroll past the machine gun until you find the bus schedule. Wait until it feels like you’re part of an immersively staged Beckett play.
My bus-stalking routine upon touching down in New York has become integrated in me like a constant algorithm, one of those mundane, unchanging parts of life that can be placed on autopilot.
Usually, after finding my spot on the bus, I look out the windows until teal patinated rooftops dominate the skyline. I know it’s Columbia long before the operator can crackle out “116th and Broadway.” On my last arrival in the city at the beginning of the spring 2022 semester, however, I was inspired by an antsy whim while returning back to campus. Seeking proof that we were moving, I decided to track the M60 all the way down. The smart approach would have been to continually refresh the same web app tracker, but instead, I opted for the nervous approach: to close and reopen new trackers at every stop. Though this fidget was inefficient, it was unexpectedly insightful; as the bus ran down, its place in Google’s search results went up. The M60-SBS bus line serves, on average, 14,778 riders on any given week, and as one of the main options for ground transport from LaGuardia into the city, the M60 is one of the first—and likely one of the most uncelebrated—introductions its straphangers have to New York City.
The bus shares its place at the top of Google’s search results with the M60 machine gun, also known as “The Pig,” a Cold War darling whose claims to fame are bulk and force. According to the Google Trends data for the term “M60,” New York City claims a score of 100 out of 100 on the regional popularity scale that ranks search interest. I wondered whether my initial experience of online exposure to the Pig and the bus was universal among the ridership of New York City natives, incoming college students, vacationing families, and daytrippers on business. By the time I arrived back at school, I wanted to know: What were we all finding when we searched for the M60?
The plan was supposed to be simple. Dedicate a day to riding the bus as many times as I could, log the search results at each stop, corroborate with information about Google searches I was sure existed somewhere on the internet, to deduce the formula that could explain this dance between the two M60s on my search page.
Admittedly, this was an imperfect experiment. I quickly learned that I was not playing in a virtual sandbox, but rather, a blackbox. Google’s 24-year-old search algorithm is undoubtedly the canal through which the world’s online traffic flows. There’s hardly a choice. In the search engine market as of June 2021, Google owns 92.47% of all shares; Bing, the runner-up in the search engine marketplace, holds only 7.2%. When it comes to Search, Google has no competitors in the playing field sown by its very hand. What we do know about the search mechanism is what Google itself has published, material which is hardly more detailed than the recipe to create the Powerpuff Girls.
On the process of delivering search results, Google offers: “When a user enters a query, our machines search the index for matching pages and return the results we believe are the highest quality and most relevant to the user. Relevancy is determined by hundreds of factors, which could include information such as the user’s location, language, and device (desktop or phone). For example, searching for ‘bicycle repair shops’ would show different results to a user in Paris than it would to a user in Hong Kong.”
In essence, I was provisionally feeling around the edges of just one out of an infinity of virtual faces provincially available to me on my phone with as many round trip rides up and down the M60 line as my sandwich bag of quarter rolls would afford me—which would turn out to be three.
During my three trips, I realized some link between my search results and my location. I was thankful for the moments of air I could collect at the ends of each journey as I stopped to slot my coins, nine at a time, for my receipt. I typically ducked my head as I flashed my ticket to board, but around my fourth embarkment, the driver started to shoot a too-tired-to-ask look at me. I must have seemed like a NPC spending her afternoon taking hour-long rides to nowhere. Suddenly self-conscious of the private oddness I shared with the driver, I remained determined to follow through. As consistently as I was able to determine, four stops of the 18 that the M60 takes bump the bus schedule to the top of my page: Amsterdam/West 120th Street, Broadway/West 120th Street, Broadway/West 116th Street, and Broadway/West 106th Street. These neighboring stops comprise the most inwardly Manhattan leg of the bus’s journey and generally represent the corners dotting our community of Morningside Heights.
Just one stop prior to entering the MoHi internet bubble at Amsterdam and La Salle—a five-block, three-minute walk, to be precise—the Wikipedia entry for the M60 machine gun reoccupies the prime real estate of the first search result. The furthest the bus and the gun ever sat from each other was at Hoyt and 31st Street, the last stop the M60 takes in Queens; there, to arrive at the bus schedule, I needed to scroll past the gun, a tank, several tactical video demonstrations of both gun and tank, and a firearm enthusiast’s blogpost. In our online climate, with its endless capacity to adopt and repurpose real-world mis/dis/information, isolating any singular group of components that could be responsible for the variations returned at each stop is an impossible errand. Instead of trying to put a flashlight to the face of this mystical dragon in the machine, I discovered the value in catching its smoke, the consequence of its machinations that conclude, for one reason or another thousand, that the difference of five blocks and three minutes is the difference between a harmless commute and engines of war.
In actuality, the bus and the gun were never too far from one another. On my phone, at least, they more often than not sit adjacent, separated by no more than half an inch. The natural impulse is to question my motivation to investigate a seemingly trivial difference: Why take this half-inch and stretch it into literal miles of bus riding? The impetus lies in our psychological tendency towards convenience. According to 2020 data collected by Sistrix, an internet analysis company, the average click-through rate for the first result shown on Google is 28.5%. Following that, the second result plummets to a 15.7% CTR, or roughly half the amount of attention attracted by the first. The exponentially increasing dearth of interaction through progressing search results is perhaps best exemplified by Google’s second page: The same Sistrix study demonstrates that every result on the second page generates a CTR of less than 1%. The difference between a single keystroke or click is measured not in inches but in magnitudes of viewership, and a position several results away from the top is the search algorithm’s analogue to a death sentence: invisibility.
Yelp CEO and President Jeremy Stoppelman said as much when faced with the hammer of Google’s omniscience. With the rise of Google Reviews in 2011, Stoppelman had mistaken Google’s tone of voice when they informed him of their strategy to build the back of their review system from data trawled from Yelp: Google was not requesting permission from Stoppelman, but declaring their intentions before doing as they pleased. In essence, Stoppelman faced the bitter choice between obedience and obsolescence: to comply with Google’s self-determined “fair use” of information publicly accessible on Yelp’s platform, or to remove said information from the grasp of its search engine. Stoppelman remarked that the ultimatum constituted a threat: Offer up your platform’s data or “take yourself off the internet. That would have destroyed the company, so it was a false choice.”
In the conflict that ensued, much of Yelp’s energy has naturally gone toward accruing the necessary endurance against Google Reviews’ parasitic practices. The untouchable industry giant, in the early stages of its feud with Yelp, had poached enough from Yelp to make the reviewing platform its subordinate—but not so much so as to eliminate the database as a resource from which it could draw. In an egalitarian vision of the internet, Yelp and Google Reviews would stand on relatively equal footing in attracting consumer attention, not unlike how we scrutinize products in the same aisle at the grocery store. The discrepancy between that ideal and our current online reality is the outright lack of objectivity in the production of search results.
If one could trust the process of exchange to guarantee desirable and accurate results, such a lack of transparency might be acceptable. But with how the consumer-corporation relationship stands now—with Google exempt from the same level of skepticism currently lobbed against an Amazon or a Facebook—Google benefits from consumer trust conditioned almost to the point of instinct, triangulated perfectly with its status as a vertically integrated monopoly immune to the challenges for accountability, from competitors or regulators alike. So while Yelp would pay for a place as the second result—and even for its exile to the second page—chance glimpses remained better than nothing. Time and time again, our collective internet habits illustrate the rule that what first comes is what we overwhelmingly choose.
“Don’t be evil.”
Three words that compose a philosophy so simple it feels as if it were born from some book of ancient truth, not some Y2K draft of Google’s code of conduct. To Google’s credit, the precise origin of this cheeky mission phrase is mired in legend, a song in which the suspected heroes have swapped their spears for software engineering. Irrespective of the semantic details of who said what—long lost to Google’s early history—one of the proposed coiners of “don’t be evil,” Gmail founder Paul Buchheit, explained that it was “a bit of a jab at a lot of the other companies, especially our competitors, who at the time, in our opinion, were kind of exploiting the users to some extent.” At some point, Google believed in this commandment enough to codify it, and it continues to reside in the final line of the company’s code of conduct.
The presently challenged milieu surrounding Google and “don’t be evil” was a conversation I found myself returning to throughout my endeavor to reconcile the Pig with the bus. To be clear, I don’t believe that obfuscating the mystery of how Google Search works is, in and of itself, “being evil.” I do, however, understand the mechanism as it is publicized as a fig leaf, one that Google won’t tear off any time soon. While the majority of the elements directing my experiment lay outside of my observation, I could still maintain control over my interpretation of the results: I knew exactly what I was trying to find (the M60), and my target was obviously distinct from the rest of the results populating its page (the bus from the myriad artillery). But how does this complete opacity over search results bode for those going in completely blind?
On Feb. 15, Google tweeted a fun fact: 15% of its received queries are novel. That is, from the estimated billions of searches Google processes every day, never-before-asked questions come in daily by the hundreds of millions. Genuine curiosity backed by our trust in first-page convenience functions as an open invitation for bad actors seeking to warp the lines of a search engine’s design.
For a while, Google tacitly welcomed this species of malicious interference, for the profit successful runaway disinformation schemes generated for the host platform. Some of the most recent high-profile examples of such online campaigns deal with hoaxes about climate change and the 2016 presidential election. While Google has more recently paired some safeguards with results concerning these topics—such as third-party fact checker pop-ups and source verifications—there of course remain critical dark spots on the site without such protections, including life-or-death matters like abortion and LGBTQ+ rights. Even meme culture has exploited the flimsiness of assigning relevance to webpages: A 2018 tweet that Sir Thomas Running invented the eponymous exercise in 1612, when he attempted to “walk twice at the same time,” trended on Reddit and Quora forums so vigorously that screencaps of Google returning the lie as its first search result when asked “when was running invented” became an entire meme genre of its own.
As coveted a position as the first is, it’s fragile up top. When falsified sources and red herrings can so easily hack page indexing, I’ve come to understand that one of the only effective measures we have against a product crafted for minds on autopilot is to deliberately view more than just the first and second sources on our screens. Every time we perform a search, Google essentially hands us a dossier of up to billions of sources in under a second. Maybe we should rifle through more of it.
I’ve since found a new way to kill my time in the cold. When I’m standing by for the bus, I pull out my phone and look for whatever crosses my mind. From the generated results, I scroll and flip through the internet’s unseen pages. Sometimes, I think of the bus. That search generates the usual suspects of more gun profiling. Other times, I find myself surrounded by minor league baseball teams, “Visit Our City!” advertisements, and local Missouri newspapers on the page-six equivalent on “Columbia.” “Bacon, egg, and cheese” floods my screen with the anticipated mountain of recipes and greasy spoons, but travel past it and find a sea populated by bread boats and waffle melts and other more exotic breakfast eats. Instead of falling into the dull noise conducted by a biased automaton, through this new ritual, I find my own chord of autonomy on the internet. Now, anytime I sift among these buried headlines and links, I get the small, satisfying sense that I’m making the process of waiting feel like finding something human.
コメント