Imagine a world where you barely notice the barriers between digital and physical space. Instead of looking at a TV or phone, you have a pair of glasses that can project a screen anywhere. You can seamlessly pull up translations for any street sign or instructions for any task. You can amplify a difficult-to-hear conversation through an earpiece or highlight a hard-to-see detail in your surroundings.
Now imagine the same world — but your glasses scan every conversation to personalize a barrage of advertising. Some locations are replete with helpful holographic instructions, while in other places, neglect and poor connectivity make them few and far between. A sophisticated facial recognition system tracks every stranger you encounter... and, in turn, lets those strangers track your every move.
These are a few of the best- and worst-case scenarios for augmented reality, a technology that some of the world’s biggest tech companies are spending billions to promote as the future of computing. Over the last decade, AR hardware designers have laid the groundwork for a new generation of mass-market products, even as technical hangups still limit its viability. Over the next one, AR threatens to supercharge existing crises of privacy, trust, and consent. But it’s also a chance to deliberately reset how we approach computing.
AR has populated science fiction, specialized industries, and high-tech experiments for decades. But in the 2010s, major companies — and countless startups — began treating AR headsets as a potential mass-market platform.
Facebook founder Mark Zuckerberg predicted in 2016 that televisions and phones would be replaced by holographic glasses. Apple CEO Tim Cook called AR “a big idea, like the smartphone.” Microsoft envisioned people watching the Super Bowl in its HoloLens headset. Google launched its ambitious Glass platform as a potential successor to phones, then helped propel the AR startup Magic Leap toward billions of dollars in investments. More recently, telecoms have partnered with AR companies like the Chinese startup Nreal, hoping high-bandwidth holograms will create a demand for 5G networks.
These companies’ products — as well as those of other major players, including Snap, Vuzix, and Niantic — often look very different. But most of them promise a uniquely powerful combination of three features.
- Their hardware is wearable, hands-free, and potentially always on — you don’t have to grab a device and put it away when you’re done using it
- Their images and audio can blend with or compensate for normal sensory perception of the world, rather than being confined to a discrete, self-contained screen
- Their sensors and software can collect and analyze huge amounts of information about their surroundings — through geolocation and depth sensing, computer vision programs, or intimate biometric technology like eye-tracking cameras
Over the past decade, nobody has managed to merge these capabilities into a mainstream consumer device. Most glasses are bulky, and the images they produce are shaky, transparent, or cut off by a limited field of view. Nobody has developed a surefire way to interact with them either, despite experiments with voice controls, finger tracking, and handheld hardware.
Despite this, we’ve gotten hints of the medium’s power and challenges — and even skeptics of the tech should pay attention to them.
In 2016, for instance, millions of people fell in love with the phone-based AR game Pokémon Go. When players logged on to catch virtual monsters, many discovered parts of their neighborhoods they’d never thought to visit. But they also found a world built on data that placed more gyms and PokéStops in white and affluent neighborhoods. They forged in-person connections by sharing virtual experiences, but those connections could also allow for real-world harassment.
The effects went beyond people who played the game. The owners of some homes near Pokémon Go gyms experienced a sudden influx of trespassers, leading a few to sue its developer Niantic and secure tweaks to the game’s design. Public memorials like the US Holocaust Museum asked players to respect the space by not chasing virtual monsters in it. Even this early foray into augmented reality exported some of the internet’s biggest flaws — like its ability to collapse context and overwhelm individuals with its sheer scale — into physical space.
Writer and researcher Erica Neely says that laws and social norms aren’t prepared for how AR could affect physical space. “I think we’re kind of frantically running behind the technology,” she tells The Verge. In 2019, Neely wrote about the issues that Pokémon Go had exposed around augmented locations. Those issues mostly haven’t been settled, she says. And dedicated AR hardware will only intensify them.
Smartphone cameras — along with digital touchup apps like FaceTune and sophisticated image searches like Snap Scan and Google Lens — have already complicated our relationships with the offline world. But AR glasses could add an ease and ubiquity that our phones can’t manage. “A phone-based app you have to actually go to,” says Neely. “You are making a conscious choice to engage with it.” Glasses remove even the light friction of unlocking your screen and deliberately looking through a camera lens.
Augmentation also doesn’t just mean adding things to a wearer’s surroundings. It also means letting a computing platform capture and analyze them without other people’s consent.
Take facial recognition — a looming crisis at the heart of AR. Smartphone apps have used facial recognition for years to tag and sort people, and one of the most intuitive AR glasses applications is simply getting reminded of people’s names (as well as other background information like where you met them). It’s also a potential privacy disaster.
In 2020, the Electronic Frontier Foundation sounded the alarm about AR glasses’ surveillance capabilities. Any company offering facial recognition through glasses, wrote Katitza Rodriguez and Kurt Opsahl, “could have a live audio/visual window into all corners of the world, with the ability to locate and identify anyone at any time.”
So far, AR systems have mostly sidestepped facial recognition. Phone-based platforms like Snapchat, as well as Facebook’s Portal, use face identification — which can detect facial features to add special effects but not match them against a database of specific people. Google refused to approve facial recognition apps in its 2013 Glass Explorer Edition headset, although one unofficial app raised lawmakers’ ire by trying it.
But the EFF’s concern wasn’t premature. Andrew Bosworth, an executive at Facebook and Meta, reportedly told employees the company is weighing the costs and benefits of facial recognition for its Project Aria glasses, calling it possibly “the thorniest issue” in AR. And outside AR, some people are pushing for a near-total ban on the technology. Researcher Luke Stark has likened facial recognition to “the plutonium of AI,” saying any potential upsides are far eclipsed by its social harms. AR is a ready-made testbed for the widespread public use of facial recognition, and by the time any potential harms are obvious, it might be too late to fix them.
It’s not clear how AR systems will make money either — and what kinds of behaviors the resulting business models will encourage. Some companies in the field, like Apple, are traditional hardware sellers. Others, like Facebook and Snap, made their names on ad-supported social networks.
Facebook has claimed it’s not yet looking at business models for its glasses, and Snap says advertising isn’t the only option, promoting the value of experiences like AR-powered retail. But even companies with no advertising background see its power. In one patent filing from 2017, Magic Leap imagined Starbucks detecting when a headset wearer looked at a branded coffee cup, then offering a coupon for the nearest Starbucks store.
Even basic AR applications, like mapping an apartment to place a virtual screen, could potentially gather a huge amount of information. (What’s the size of your living space? Which books are on your shelves? How healthy are the snacks on your kitchen counter?) Without robust privacy protections, it will be incredibly tempting to use that data for ads. And the more companies collect and store, the higher the odds you’ll see it used for something even more invasive — like setting your insurance premiums — or fall prey to a security breach.
Some groups are trying to get the jump on broader AR policy issues. The nonprofit XR Safety Initiative offers policy frameworks for safety and privacy in the industry, drawing on existing laws like Europe’s General Data Protection Regulations. In the corporate space, Facebook Reality Labs announced a set of Responsible Innovation Principles designed to allay fears about trust, privacy, and consent that have dogged the company. It also awarded a series of academic grants to study specific issues in AR, selecting proposals like “Social Tensions with Always-available AR for Accessibility” and “Anticipating Virtual Gossip — What are (Un)Intentional Dark Patterns in a Ubiquitously Augmented Reality?”
Early efforts at consumer hardware, though, haven’t navigated its pitfalls particularly well. Google found its 2013 Glass Explorer Edition headset banned from some bars and restaurants because the expensive device was seen as invasive and presumptuous, not futuristic and liberating. That shouldn’t have been surprising: around the same period, University of Washington researchers interviewed people in cafés where someone was wearing a mock AR headset, and the results were a roughly 50-50 mix of indifference and negativity. (Only a single person had a positive reaction.) But instead of planning around some very predictable anxieties, Google CEO Larry Page called privacy fears “not that big a concern.”
There’s been plenty of sci-fi speculation about AR, from the flashy gestural interfaces and personalized holographic ads of Minority Report to the digital caste system predicted in Super Sad True Love Story. But science fiction is notoriously bad at predicting how people will use technology, even when it gets the tech right. Companies have offered hypothetical scenarios that prove AR’s value — Zuckerberg says we’ll invite holographic friends over to sit on our real-world couches, for instance, and Cook imagines visiting live sporting events and seeing stats right on the field. But historically, revolutionary technologies also aren’t top-down systems imposed on perfectly pliant users. Their boundaries are constantly being negotiated in ways the original creators never expected.
“It’s very typical that developers of a technology have one kind of idea about what its uptake in society might be, and then the actual uptake turns out to be something quite different,” says University of Washington professor and human-computer interaction expert Batya Friedman. A good system is flexible enough to adapt to these unexpected uses. “We can make it harder or easier to go in and modify technology. So if we build our technologies from the beginning with the idea that as we understand them better, we may need to make modifications, we can build in those technical hooks.”
AR technology also isn’t going to develop in a vacuum. Despite talking up AR glasses’ novelty, figures like Zuckerberg and Cook still describe people using them almost exactly the way they use smartphones: as devices they carry around casually all the time. But ubiquitous, short-lived electronics like smartphones have imposed a steep environmental cost on the planet, and rolling out AR could add billions more devices that are replaced as readily as phones and powered by vast amounts of cloud computing infrastructure. “What are the environmental implications? And do those make sense?” Friedman asks. “If not, I think really the question for us is, what are the really critical areas in which some kind of augmented reality technology really does bring a substantial benefit to people?”
Either way, the past 10 years of tech have been a long struggle to manage crises once they’re already at a boiling point. Meanwhile, AR glasses’ status as a long-awaited, not-quite-there dream can buy us time to figure out what they can do for the world — and whether we actually want them.
Correction November 2nd, 10:00AM: The researcher likening facial recognition to plutonium is named Luke Stark, not Luke Clark. We regret the error.