How Technology Helped Me Cheat Dyslexia

Dyslexia, my constant companion, occupies a taboo place in my personal narrative. Like my breath, I often forget it’s there. Sometimes I delude myself into thinking I’ve outgrown it. When I told friends that I was writing this article, several advised me to back out of the contract. One didn’t even believe me when I told her I was dyslexic. How could I be a writer? They were concerned this assignment might be my last.

But I’ve never thought of myself as having a disability. Instead, I see it as a glitch, and one I've gotten good at masking. I've been able to hide my dyslexia for decades simply because I live in an age of technological wonders. Microsoft Word spell-checks most every syllable I write. When my dyslexic mind mangles a word so much that it's rendered un-spell-checkable, I'll deploy an arsenal of workarounds. I might reverse-engineer a word by typing an easy synonym into the thesaurus, or I might paste my best attempt into my browser bar and let the search engine offer the correct spelling as a suggested query.

These "cheats" are ingrained in my writing process; I hardly notice doing them anymore. But something happened a few months ago to break me out of my familiar routines. I began writing with the help of an AI-powered browser plug-in so adept at correcting my linguistic missteps, it ended up sending me on a quest to discover what life might be like in a technologically enabled post-dyslexic world.

When I was really little, I tried to see words—the actual orthography—as pictures. For the word “dog,” I would think: There’s a circle then a line, then a circle, then a circle with a hook. Knowing the specific letters and decoding them wasn’t part of my process. Thinking in pictures was how reading worked, I thought.

My dyslexia was discovered in grade school, where I had the benefit and luck of attending a well-funded institution equipped to respond to my obvious signs of trouble. By the end of second grade, I was enrolled in an intensive summer school program for dyslexics. My class used a slide projector-like device known as a Controlled Reader. Even back then, it was a relic; when the teacher flipped it on, the stuffy room filled with the aroma of an electrical fire.

The Controlled Reader projected text onto a screen at the front of the class just like a regular slide projector, but with one difference. Light would shine only through a narrow horizontal slit, allowing only a single line of text to be illuminated at any one time. Each line of text would flip into view for a second or two, then get replaced with the next one. The teacher could crank up the speed of the machine using a dial, forcing the class to read at speeds up to 130 words per minute.

After each reel, we were given a test, and over the weeks, the speed would be increased. While I was missing out on normal kid stuff—my morning swim time, horseback riding at summer camp—something happened to me in that overheated classroom. Reading began to click. I eventually found myself in honors classes, though I did have to advocate for my placement when teachers assumed my difficulty reading meant I should be kept apart from the smart kids.

I later attended NYU’s film school and set out to make a documentary about my dyslexia. My seventh-grade English teacher even gave me his old Controlled Reader machine so I could use it in the film, but I lost my nerve and never finished the movie. I feared I wasn’t established or successful enough, and I believed in the trope that a personal story about overcoming a reading disability needed to accompany an outsized achievement. Like my dyslexia, I keep that speed-reading machine, an artifact from childhood, hidden away in the back of a closet.

At this point in my professional life, I’m only outed when writing by hand in a public setting, which was the case when I went on a book tour to promote my memoir about new motherhood and wrote my inscriptions with an unforgiving black Sharpie. I'd keep post-it notes and a pen by my side. “Could you put down what you want me to write? And if you have a fancy name like Margaux, well, jot that down too.”

While it is agreed that dyslexia is a language-based learning disability, there is no universally accepted definition of the phenomenon, nor is there a complete understanding of its cause. But with the arrival of functional magnetic resonance imaging to measure brain activity, scientists in the last few decades have been able to study the brain activity of dyslexics. What's striking is how the dyslexic brain does not utilize areas usually engaged in reading. In addition, the brain can be seen jury-rigging other areas to form words in the same way a stroke victim might during recovery, harnessing plasticity—the brain’s ability to rewire itself.

A hallmark of dyslexia is the inability to discern phonemes, distinct sounds represented by specific letters. I struggle with this. I can hear the sounds, but I sometimes can’t translate them to letters on the page. The other day, I wanted to write the word "agitated." This is a word I know. I’ve said it aloud countless times without mispronouncing it, and I’ve read it often as well. And yet, when typing it, even sounding it out as I go, I hear a “d” and a “j” in it. So fishing around in my brain’s Bermuda Triangle, I typed out the word adjetated. I can remember short words—most of the shopworn workhorses come easy—and a bunch of longer ones too. But there remains a large subgroup of words I cannot phonetically master or remember.

Then, a few months ago, I discovered Grammarly, a free cloud-based software extension that you add onto a web browser. The plug-in is billed as a "writing assistant," but I mostly used it as a spell-checker, a task at which it proved nearly omniscient. Grammarly could help me spell even the words that regularly flummoxed MS Word and Google.

Those first few weeks with Grammarly, it felt like I was like falling for a crush. In the browser, it works like any other spellchecker. An elegant light green box (Pantone 2240 U) appears when the cursor hovers over a red underlined word. But my infatuation quickly grew. Even the name, "Grammarly," sounds like the benevolent hero in a Jane Austen novel: Good Mr. Grammarly! The software seemed to get me—and my scrambled misspellings—in ways that no other had before. Grammarly always knew the right word. It even seemed to understand the way my dyslexic brain thinks—a maze of patched and redirected connections zig-zagging around my gray matter—and could come up with exactly what I was trying to say, even though I couldn't fully spell it. It was only then, using something so seamless, that I wondered if technology could soon bring an end to my dyslexia as I knew it.

Since launching its freemium service three years ago as a browser extension, Grammarly claims to have 10 million daily active users around the world. The company also offers a premium service ($30 a month) that comes with bells and whistles like a weekly email listing your performance stats and an option to upload documents for a full grammatical scrubbing. Grammarly only works on English-language documents, but with more than 1.5 billion English speakers worldwide, the company is running a growth business.

I wanted to visit Grammarly to find out if there was a reason its plug-in was able to understand the dyslexic brain better than any other of software I'd used. I was curious if the reason the service's AI was so good at catching dyslexic misspellings in all their muddled glory was because it was, in fact, learning from the world’s dyslexics each time they typed in and corrected their mistakes, accepting or refusing Grammarly’s suggestions. While poor spelling is not unique to dyslexics, maybe there's some knowledge Grammarly's AI was learning by eating our own unique brand of brain food. And maybe Grammarly had brought us all together—through this process of clicking on the offerings in the green box, we were helping each other out.

While Grammarly has offices around the world, I got lucky: Joel Tetreault, the company's director of research and development, is based in New York City, where I live. I enter a coffee shop off Delancy Street in the Lower East Side and walk to a door at the back where I'm buzzed in. I feel like I'm being admitted into a secret laboratory, like in some noir science-fiction story. I head up to the second floor. Tetreault meets me in the lobby of The Yard, a co-working space where Grammarly is a tenant, and takes me to a mostly empty office. He wears glasses and a black hoodie with a discreet Grammarly logo. Tetreault joined the company nearly two years ago after leaving Yahoo Labs, where he worked as a senior research scientist developing algorithms to identify hate speech in comments.

I tell Tetreault my theory of why Grammarly is so good at gaming the dyslexic mind—that it's learning from the input of dyslexic users—but he downplays the role of user feedback in building the AI's smarts. “There’s a little bit of that in there for sure,” he says, but I am disappointed he doesn’t give us dyslexics more credit. And why do I have this irrational need to be more than just a customer? Perhaps it is because the software gives me the keys to my post-dyslexic existence? I mean, I'm not getting all weird and fixated on an AI like the redhead actor in Ex Machina. It’s not like Grammarly is going to pass the Turing test anytime soon. But here I am, sitting in an office on the Lower East Side gushing about a spell-checker.

Those working in the field of natural language processing view spell check, which has been around for nearly half a century, as a pretty remedial problem. In a world full of exotic grammatical mistakes, there are more complex and interesting things to tackle. A quick look at Grammarly’s demo document in my premium account shows an array of things it can do: suggest new words, replace weak adjectives, point out incorrect progressive tense use, suggest the appropriate placement for an adverb, underline modifiers when typed in the wrong order.

Tetreault tells me the genius behind Grammarly's writing assistant, including the “spell-checking functionality”—those at Grammarly do not refer to it as a spell checker—comes from many places. The AI learns from studying millions of documents and other language-based data sets, along with computer generated misspelled words, and, yes, user feedback. For example, with any given spelling error, Grammarly presents one or a few possible corrections. As Grammarly studies the behavior of a subset of users, it sees which replacement spellings users accept and ignore. That information is incorporated into the options offered up to users in the future.

I would have discovered Grammarly sooner had I bothered to visit the many websites for dyslexics that thrive on the internet. Though users of all stripes love Grammarly—the extension has a four-and-a-half-star rating from more than 30,000 reviewers in the Google Chrome Store—it has the status of a beloved rock star in the dyslexic community. Grammarly is up there with Dragon Dictation, the speech-to-text software which, as a fast typist, I never took to. One friend who discovered I was writing about Grammarly outed herself as dyslexic and waxed poetic throughout our brunch about how the software transformed her working life.

The website Dyslexic Advantage named Grammarly as its top dyslexia app for 2016. Unsurprisingly, Grammarly is also a topic of vigorous discussion on Reddit, where the extension receives high praise from dyslexic users and endures bitter criticism from Redditors dissing the app and those who rely on it. Dyslexic-hating seems to be a hobby on Reddit, where the erroneous belief that dyslexia is a sign of lower intelligence crops up frequently. In fact, I have an uneasy thought while sitting with Tetreault. It’s one I have whenever I reveal my dyslexia: I wonder if he thinks I’m stupid.

I suppress that thought as Tetreault shares the thing he is most excited about, his North Star. In the world of natural language processing, it's called "style transfer." In the future, a user could be offered the option to upload a document and, with the push of a button, transform the style and voice of the document without altering its meaning. A friendly note can be rendered in the more formal voice one would use in a cover letter for a job. Or a starchy essay can be transformed into something with the casual familiarity of a chatty blog post.

My immediate response is a small Luddite seizure: “Wait, this is going to put me out of a job.” As Tetreault describes it, style transfer is more like a friend who knows you, knows your voice, and can help you craft a piece of writing. I can see such technology having a substantial positive impact on dyslexics. Get the meaning down, errors be damned, then run it through the software. Words come out the other side clean and tone-perfect. Still, it makes me queasy.

Because of my dyslexia, this thing I do—writing—has been hard won, and I feel oddly resistant, almost miserly about the advent of style transfer. While the current version of Grammarly can make corrections to style with the option of "formal" or "informal," a blanket style-transfer feature does not yet exist in the plug-in. "It's the next frontier," Tetreault says.

I was nervous about feeding my personal documents into Grammarly's machine mind, fearing I was striking a Faustian bargain by giving up my privacy. But the company tells me that after writing suggestions are provided to the user, any uploaded text is disassociated from the user's account, and if it's kept in storage for further study, it's anonymized, with names, email addresses, and other identifying information stripped out. (A security bug discovered in February was quickly remedied.)

I had a deeper concern too: Would the the easy fixes I found with Grammarly dull my faculties since I wasn’t “working” for the word?

I first heard Guinevere Eden, director of the Center for The Study of Learning at Georgetown University, on an NPR segment about rewiring the dyslexic brain. Her research focused on studying the fMRI brain scans of both dyslexic children and adults before and after they participated in an intensive, weeks-long remedial reading program. She discovered that previously under-activated areas in the dyslexic brain are more active after the course. Whatever they were reading in the study, and however they were doing it, their brains were changing.

In her NPR interview, Eden described how the human brain was not initially designed to read. That was fascinating to me, and somehow comforting, too. So in our first phone call, I ask her to elaborate. Reading, she says, is a relatively new activity for humans, maybe only five or six thousand years old. Our brains were designed for language, but we were storytellers with an oral memory. The task of reading takes place in the parts of our brain initially tasked with recognizing and identifying objects, she says.

I told Eden I was curious about what was going on in my brain and asked if I could get a scan at her center. After inviting me down, she asked me, “Aren’t you concerned about coming out about this?” I understood what she meant: Wasn’t I afraid people would think I’m stupid? Hadn’t I hidden it for that very reason? I hated that Eden’s question echoed my friends’ concerns. Maybe keeping my dyslexia on the down-low is better.

When I hung up, it occurred to me that perhaps I hadn’t thought this all the way through. In truth, my dyslexia had been weaponized against me by family members, classmates, and exes. Keeping it a secret to the outside world gives it a certain amount of power that’s out of my control. I often write about the things that make me uncomfortable—giving birth, breastfeeding, divorce, online dating, my sex life, and taking Ayahuasca to name a few—but in the process, I understand how I truly feel and surrender any lingering shame or embarrassment. Would I regret writing about how I see words?

It’s a misty morning in April, and I am on the train from New York to Washington, DC, to get my fMRI brain scan at Georgetown University Medical Center. I meet Guinevere Eden outside. She’s a tall woman with light eyes and blonde, shoulder-length hair. She speaks with the faint lilt of a German accent that I mistake for Dutch. I imagine her calm presence is helpful in putting her young subjects at ease. She walks me back to her office. Its large windows illuminate rows of bookshelves and a small anatomical model of the human brain on her coffee table.

I show Eden a list of words that I often misspell—maintenance, unfortunately, definitely, abominable—and the variations of their spellings I use. I printed them all out on a page in large letters. I want to know why I am still unsure of the correct versions. She looks them over. “Well, I think this is really interesting because it speaks to the fact that their visual word forms are very similar, right?” She says, “a skilled reader represents this visual word form in the visual cortex. There's an area down there called the visual word form area. Research has shown that there are neurons that are literally tuned to this word. They own this word. It's your dictionary in your brain, and these neurons, they fire when they see this word.”

I tell her that I think all of the words on the page are correct—even the misspellings. Eden tells me that it's my dyslexia that's keeping my "visual word form" neurons from firing when I see the correct spelling. I find this explanation comforting. I live with this thing day-in and day-out, and yet speaking with Eden in the safe space of her office elicits a feeling I never experience—compassion for myself.

I also want to know if my reliance on Grammarly to correct my spelling is dulling my brain. Just the opposite, she says: “If the program is always reiterating with certainty, 'This is the correct spelling,' and you see it, it may actually help you to begin to represent this word form in a way that it is actually correct." Hearing this, I’m surprised by a rush of relief. Maybe I don't need to feel guilty about my reliance on technology.

Next, we head to the basement, where my brain scan will happen. Eden explains that while in the scanner, I will perform a series of tasks. I will be shown both real letters and a made-up font of nonsense letterforms, and I'm to push a button if I see a word—real or fake—with a long vertical line in it. So if the word is shaped like an “l” or “d,” I'm supposed to click the button on the right. If I don’t see a letter shape with a vertical line, I push the button on left. This test is intended to activate the parts of the brain usually used for reading. There can be no metal in the fMRI machine, so I remove my earrings and slip out of my heels as Eden runs a TSA metal detector wand up and down my body.

The fMRI scanner looks like a white airplane engine with a conveyor belt sticking out of it. As I lay down, Eden’s colleague gives me headphones to protect my ears from the racket the machine makes while doing the scan. Then he places a white plastic “Birdcage Coil” over my head, a contraption that looks like a birdcage with a circular mirror at the end of it so I can see the computer screen. This is how I see the reading test. I am handed a rubber emergency squeeze bulb in the event I feel claustrophobic, and for a brief moment as I look out at the room from behind my plastic cage, I worry I might need to squeeze it. I'm also given two buzzers—one for each hand—as the conveyor belt feeds me into the giant hole of the fMRI machine. I look up at the mirror attached to my head, and see the image of a smiley face come into view. It is time to start my scan.

My performance anxiety makes me press the wrong buzzer for the second image, and I wonder if mistakes will show in my results. When I finish the test, a voice from the control room instructs me to stay perfectly still while I watch a video. As I take in seven minutes of the children’s animated show The Magic School Bus, Eden's team captures a 3-D scan of my brain, along with a colorful Diffusion Tensor Imaging scan that exposes my white matter—the tendrils of nerve fibers running through the various regions of my brain. The idea is that this will show me how the different parts of my brain are wired together.

As I step back into my shoes, I see my scans on the control room’s monitors. Along with the slices of my brain is the most hideous image—the 3D scan of my brain that includes my scowling face, looking very much like a demon baby. After the scan, Eden shares how emotional parents get at the end of their children’s six-week tutor study when they see the difference between the initial "before" scan and the more activated "after" scan.

It shows the kids making real progress, their brains kicking into gear. Sometimes the parents cry, she says. I understand this. I have three children, and one is dyslexic. (Yes, dyslexia is highly heritable, and I was on the lookout for signs of it in my children. My son, now a high school freshman, benefited from being identified early on and has always attended schools that teach to his way of learning. If the options were to struggle or to thrive, he’s thrived.)

When I leave Georgetown, Eden says she will send me my fMRI scans. On the train home, I find myself wondering what my brain scan would have looked like back when I was 8 years old during that summer I spent with the Controlled Reader machine. Back then, I think, I would have shown excellent results in the "after" scan.

Perhaps it’s my own diagnosis at a young age, and my son’s early intervention, that made me reach out to Lexplore. The Swedish company, which launched in the US last year, uses eye-tracking software that promises to identify a dyslexic reader in minutes.

The foundation for Lexplore’s algorithms comes largely from data collected by the Kroneberg project, a study that ran from 1989 to 2010 and followed 2,165 Swedish students into adulthood, tracking their reading development and the progression or regression of their disabilities. The Kroneberg project gathered data by recording subjects' eye movements using tech-enhanced goggles, sort of like very early-stage smart glasses.

I meet up with Janine Caffrey, Lexplore US's new CEO, next to the Empire State Building in an office booked through Breather, the hourly office space rental app. I am there to demo the eye tracker. I want it to look like the Voight-Kampff Machine used to identify Blade Runner's replicants. Instead, it looks like large iPad with an inch-wide piece of black plastic hardware clipped along the bottom. The strip, made by the Swedish company Tobii, contains three eye-tracking cameras.

Caffrey has more than three decades of experience in education. She started out as a special education teacher and most recently worked as superintendent of the school system in Perth Amboy, New Jersey. As she shows me the proper way to sit for the eye tracker—one arm folded on top of the other on the table to keep the head stable—I easily fall into the role of student to Caffrey’s teacher. My eyes need to sync to the machine. With the three red camera lights pointed at my pupils, Caffrey tells me to move my eyes, not my head, and stare at the bullseyes in the four corners of the reading window. I sync. This is easy.

First, to test fluency and ability, I read a paragraph about a dog. I read it aloud, as instructed, perfectly, though I do not pay attention to the text. The text disappears from the screen and the question comes up, “Did the dog bite?’

“Yes.” I answer. I am wrong. I was not told there would be questions. Caffrey laughs and assures me most adults get this part wrong, but I feel stupid anyway. I mean, I think I trained myself over the years to be a fabulous reader.

Next, Caffrey plays back the recording of my eye scans. As I listen to my voice recite the text, small purple dots dance across each word as I read it. The dots connect to each other with a thin purple line. It isn’t so much that my eyes are burning holes into the screen with lasers, but more like my pupils are uncapped purple markers drawing on a white tablecloth. Each time my eyes pause, perfectly round purple inkblots appear. The longer the fixation on a word or letter, the larger and darker the circle grows.

Next, I'm asked to read a paragraph silently. It is a short passage about Emily wanting a horse—easy. This time I pay attention in case there is another “anchor question,” as Lexplore calls it. There is, and I get it right. When Caffrey runs the video with the eye-tracking data overlaid, it looks different than the previous one. This time, I see that as the dots grow and purple lines connect the words, my eyes jump backward to certain words.

These slight regressions, these back-and-forth lines, are caused by me going back to certain words to make sure I am reading them correctly. I have no idea I do this, but shown the dots and lines of the eye-tracker, there it is. And it is something that could be flagged as a sign of dyslexia. How did I not know I had this tic? Did I do it all the time?

Unhappy with this result, I ask Janine if I can take the test again. “You can’t beat the machine,” she says.

Caffrey shows me the recordings of children who screened as "low readers" or as dyslexic. I watch purple circle after purple circle grow larger with the child’s fixations. As the fixations grow along with the zig-zags of regressions, the screen fills with purple. If a parent or teacher could see into the child’s mind’s eye, they would appreciate how hard the child is working to decode the words. That kid wouldn't be called lazy or slow.

The effect of seeing dyslexia in action is profound and unexpected. Throughout my life, I’ve encountered disbelief—revulsion, even—when a typical reader witnessed my struggle. As a child, I asked an adult to help me spell a word only to have them say the letters back at me while doing an impression of Lennie Small from Of Mice and Men. But with Lexplore, the child would ideally get the intervention and resources for the deep work that lay ahead. Perhaps more importantly, the child would be understood. “This is why I call it the empathy machine,” Caffrey says.

I leave the demo stunned.

Lexplore already has a foothold in Sweden and is looking to grow in the already crowded American market for early reading assessment. The company claims a 95 percent accuracy rate in identifying “at risk” readers. Its portable screeners can be either leased or purchased, and Lexplore can train a facilitator in three hours. The test itself takes only a few minutes. Caffrey estimates kids could be screened for around $15 to $20 a student. The company's tech could identify kids early enough in life to make a real impact.

Not long after my Lexplore test, I hear back from Eden. Being a good scientist, she doesn't offer any analysis of my brain from only a single scan. But she shares three images from my fMRI. It appears that all of my reading activity is squatting on the right side of my brain. The left side, where humans typically process language, looks completely abandoned. How am I even talking?

Like the Lexplore test, my immediate reaction is to hide the images. They’re not hacked nudes from my iPhone, but I feel exposed. She also shares the colorful DTI scan of my brain—the image captured while I watched The Magic School Bus. Eden explains that the image illustrates the white matter pathways in the brain. Green is from front to back, red is left to right, and blue is top to bottom. I don’t understand exactly what the DTI scan is showing, but I am less alarmed by the colorful scan.

I thought back to her question during our first conversation: Wasn’t I afraid of people finding out I was dyslexic? I’ve never doubted my intelligence. And yet I find myself feeling self-conscious about my wayward neurons. I am unsteady in my conviction of living my post-dyslexic life. I want to shed this stigma. Wasn’t Grammarly the last bit of digital scaffolding I need? But I wish I hadn’t done those tiny regressions when I read. I wish my test results showed something different. I wish I didn’t take such a hard look under the hood. I am brain vain.

Perhaps it isn’t technology that’s going to bring us to the post-dyslexic world, but our broader perception of dyslexia. Maybe this thing I’ve covered up—my kryptonite, the thing I last admitted out loud on a seventh date in 2017—actually gives me an advantage. I thought about Eden’s observation that possibly, in a few thousand years, humans may not be getting our information from reading, but from some different method. Instead of letting my dyslexia leaving me feeling vulnerable or exposed, I wondered if my true post-dyslexic life was really one where I embraced it.

It's my hope that when we talk about dyslexics, we won't keep pulling up the old list of famous ones like Einstein, Picasso, or Charles Schwab, but add somebody current. And while we’re at it, we could move away from oft-repeated dyslexic version of the Horatio Alger story of the white males (dyslexia is more common in boys) who, once they reach the pinnacle of success and wealth, are free to announce to the world, “Hey everyone I’m dyslexic!” Perhaps the real shift is when the estimated 5 to 17 percent of the population that has dyslexia comes out too.

Maybe it wasn't the technology that made me feel safe to come out, but rather the fact the benefits of neurodiversity have become more broadly recognized. Richard Branson recently opened the world’s first dyslexic sperm bank, TedTalks on dyslexia and its virtues are proliferating online, and the Gershoni Agency, a creative services shop in San Francisco, markets founder Gil Gershoni’s dyslexia to its clients, promoting it as a creative advantage. The way I think—the fast connections, the ideas that come out of nowhere, the quick fixes to storylines, like the way I'm plotting the novel I'm currently writing—maybe it’s been dyslexia that’s powered the whole show.

I'm reminded of a scene from Harry Potter and The Sorcerer’s Stone. Hagrid is sent to take Harry to Hogwarts School of Witchcraft and Wizardry. It’s clear that Harry’s evil aunt and uncle never told him about his true nature, and that they made him feel his gifts were a curse. Hagrid, angry at this injustice, tries to inject a sense of pride into the boy by telling him: “Harry… yer a wizard.”

And yes, I just spell-checked the word sorcerer.