Delete: The Virtue of Forgetting in the Digital Age
Princeton University Press, $24.95 (cloth)
In 2006 Stacy Snyder, a 25-year-old student at Millersville University in Pennsylvania, was denied a teaching degree just days before graduation. University officials had discovered a photo of her, captioned “Drunken Pirate,” on MySpace. The photo showed Snyder wearing a pirate hat and drinking from a plastic cup, and the university accused her of promoting underage drinking. As Viktor Mayer-Schönberger tells the story in his new book Delete: The Virtue of Forgetting in the Digital Age, Snyder lost control over the photo when it was indexed by Google and other search engines: “the Internet remembered what Stacy wanted to have forgotten.”
Snyder’s story, and others like it, motivate Delete’s plea for “digital forgetting” (though it turned out that the university had other reasons to deny Snyder her certificate, including poor performance). According to Mayer-Schönberger, we have committed too much information to “external memory,” thus abandoning control over our personal records to “unknown others.” Thanks to this reckless abandonment, these others gain new ways to dictate our behavior. Moreover, as we store more of what we say for posterity, we are likely to become more conservative, to censor ourselves and err on the side of saying nothing.
For people like Snyder, Mayer-Schönberger proposes a creative remedy: enable users to set auto-expiry dates on information. Thus, Snyder’s “drunken pirate” photo could disappear from the Internet in time for her to receive the teaching certificate. Even if a third-party discovered the photo, Snyder could adjust its expiration date and destroy all digital copies—including those cached by search engines—with a few clicks. Were she to appear in someone else’s photo, Snyder would be able to negotiate the proper expiration date for this photo with the photographer.
The details of the proposal are a little implausible, but then, Delete is more a romanticist rebellion against technology than a how-to manual. The focus of the rebellion is technologically enhanced remembering, and Delete is an impassioned call for less of it. Unfortunately, this interesting argument suffers from three large and arguably fatal flaws: a very loose account of what memory is, an insufficient appreciation of the value of remembering, and—most important for public policy—an unconvincing effort to distinguish the animating concerns about memory from more conventional (and serious) concerns about privacy.
• • •
Suppose we transfer photos from an iPhone to a hard drive: who is remembering? And is this an act of remembrance at all?
If the transfer succeeds, we may have a faint memory of saving the photos in some generically named folder on our hard drives, but to find those exact files we’ll also need to know how to look for them (e.g., by name, date, approximate contents). Yes, these days we produce, consume, and save more data—a study by researchers at the University of California at San Diego found that in 2008 the average American consumed 34 gigabytes of information per day, an increase of about 350 percent since 1980—but it does not follow that we remember more. Perfect digital memory is useless without perfect digital cataloging. In the absence of perfect cataloging, perfect memory still requires a colossal cognitive effort.
The story of AJ, a mysterious 44-year-old Californian who never forgets anything that happens to her—she recalls details of every day since she was eleven—is a case in point. AJ has invested tremendous effort in remembering so much information. Gary Marcus, an NYU cognitive psychologist who profiled AJ in Wired, found that she kept every stuffed animal she ever had (“enough . . . to completely cover the surface of her childhood bed”), as well as 2,000 videotapes, 50,000 pages of meticulously handwritten diary entries, even a copy of every TV Guide since 1989. She told Marcus that one of the biggest regrets of her life is that no one followed her around with a microphone in childhood.
AJ is one of Mayer-Schönberger’s central examples of the perils of not-forgetting, though he ignores these exertions and, surprisingly, says she leads a “relatively normal life.” Her doctors told her that her brain scan resembled those of people with obsessive-compulsive disorder, and she clearly is obsessed with thinking about the past. The more she thinks about it, the more she remembers it—a phenomenon that psychologists call “elaborative encoding.” But why attribute these phenomena to the easy availability of “external memory”—tape recorders, videotapes, TV Guide, or stuffed animals—rather than, say, childhood trauma (which is present in AJ’s life)?
What we treasure most about our memories is not our GPS coordinates or body temperature at those moments, but the totality of experiences we felt.
Even if AJ’s plight were a product of external memory aids as opposed to obsession and trauma, are digital pressures really threatening to turn us all into AJs? Not so clear. Cheap digital storage has certainly created a nation of electronic pack rats. The shrinking physical size of our hard drives conceals the fact that they store more and more data, constantly challenging us to fill them up with a glut of music, films, and e-books. But Mayer-Schönberger makes a large conceptual leap from compulsive hoarding to compulsive retrospection.
Consider digital photography: we take lots more photos than we did in the analogue age. But is the global photo-orgy—3 billion photos are uploaded each month to Facebook alone—good for remembering? There is a good case to the contrary. The explosion of photos makes it harder to find what we need. A 2008 study conducted by researchers at the University of Sheffield in Britain found that 39 percent of surveyed participants failed to retrieve digital photos of important events that took place only a year before; they couldn’t find them on their hard drives and had no idea how to search for them, as they had not organized and annotated them properly.
In Mayer-Schönberger’s universe, all those photos are simply “remembered” because they are stored on external memory. The reality is strikingly different since our external memory intricately depends on our limited willingness and ability to curate our digital collections. To be sure, the need to actively curate may be mitigated if the annotations that describe our files—known as “meta-data”—are generated automatically. But this future is more distant than Mayer-Schönberger thinks.
Our ability to search “within files”—according to Mayer-Schönberger, another instance of newly easy information retrieval—is also inadequate. We have not fully mastered text searches: try using synonyms to search for text files on your hard drive. And audiovisual search is in its infancy: try singing a tune to your computer and seeing if it can find a matching audio clip stored on your hard drive.
But even if new forms of meta-data and search do arrive, information retrieval is beset by another challenge: our storage facilities are beginning to disintegrate. Most of our digital possessions are no longer stored on just one hard drive. Instead they are distributed across different storage platforms: a combination of email accounts, social-networking sites, laptops, iPods, and mobile phones. It’s impossible, thus far, to conduct a simultaneous search across all platforms.
Powerfully drawn to the myth of easy retrieval, Mayer-Schönberger attacks those who are trying to benefit from it commercially. His favorite targets are some strange folks who defy biological memory by documenting their entire lives through a practice known as “life-logging.” Some eccentric life-loggers wear cameras that take snapshots every 30 seconds or whenever someone approaches them. Others scan all their receipts and broadcast their current geographic location over the Internet. Most opt for some lesser forms of data exhibitionism. Mayer-Schönberger picks a fight with life-logging’s cheerleader-in-chief, Microsoft’s Gordon Bell, who argues that life-logging could help us achieve “total recall” and thinks that forgetting is a bug, not a feature, of human memory.
Defenders of life-logging believe that it can help us get more effective at managing information, improve our productivity (no more lost car keys!), and help us identify health problems sooner. The greatest promise of life-logging tools lies in their use as memory aids by patients with Alzheimer’s disease. For example, using life-logging tools, caretakers could present patients with cues to stimulate their episodic memory. But these caretakers report that the biggest drawback of using life-logging tools as memory aids is precisely their comprehensiveness: reviewing every frame and sound captured by such tools leads to cognitive overload in caretakers and patients.
The uselessness of comprehensive life-logging tools reveals the common flaws in how Mayer-Schönberger and Bell think about memory: perfect remembering afforded by life-logging yields less actual memory than more restrictive and discriminatory forms of capturing an experience. Retrieving the video footage of one’s entire lifetime from a hard drive is simply not the most effective strategy for evoking rich and emotional memories of an experience. Given a choice between using a camcorder or an audiocamera (a gadget that takes photos and records the ambient sounds at the time of the capture) to document an experience, 79 percent of respondents in a 2005 British study chose the audiocamera. Photos set to ambient sounds trigger more intense memories of experiences than do the same photos set to participants’ own voice commentary or even the video footage. What we treasure most about the memories of an important event is not the knowledge of our GPS coordinates or body temperature at that moment, but the totality of experiences we felt.
Reminiscence—not forgetting—faces extinction in a digital age that prioritizes the present over even the recent past.
Life-logging tools do not capture experiences—when it comes to remembering, quantity does not translate into quality. Life-logging is simply too boring and its data too daunting to review ever to become our dominant platform for remembering. And let’s face it, most of us are not exactly gifted cameramen.
• • •
Life-logging may, however, have virtues as a supplement to ordinary memory. What if we could hear the ambient sounds that we heard on the beach every time we lift that seashell from our bookshelf? Or see an automatically reconstructed class schedule for a random day in our school calendar or listen in to the canteen buzz when viewing photos from our university days? What if we could browse through a digital collage of little-noticed ephemera—bus and opera tickets, restaurant bills, wine labels, price tags, and the like—collected on a summer vacation abroad?
Life-logging provides an effective way to produce superior, smarter time capsules that could make our future memories of the past richer and more multilayered. Few of us would take the effort to devise such time capsules for all events in our lives—besides, the importance we attach to such events changes over time—but life-logging could do this automatically, revolutionizing how we reminisce. Appreciating these virtues, however, requires that we understand the value of memory. And a second large problem with Delete is that Mayer-Schönberger makes no room for the positive and therapeutic role that remembering and especially reminiscence play in our lives. As a result, he does not see how technology could actually augment our memories without driving us insane.
Mayer-Schönberger fails to recognize that it is reminiscence—not forgetting—that faces extinction in the digital age. As Facebook and Twitter prioritize the present—our most recent updates always appear first while older ones have a shelf-life of half an hour—our digital lives are increasingly detached even from the most recent past. (The ubiquitous ADHD does not help either; Mark Helprin recently observed in The Wall Street Journal, “With the American attention span being what it is, time capsules are now retrieved 45 minutes after they are buried.”) Intelligent, highly selective, and responsible use of life-logging combined with an effective system of sending non-intrusive reminders could make us more aware of what we thought and did in the past. It would be disastrous to let the fears about digital forgetting blind us to the new possibilities for reminiscence created by digital remembering.
Mayer-Schönberger’s argument for “forgetting” as a precondition for forgiving is also based on an insufficient appreciation of the rewards of memory. As Avishai Margalit observes in The Ethics of Memory, forgiveness requires that we transcend the resentment that the memory of an event conjures up in our minds. Forgiveness requires not forgetting, but remembrance. Repressing memories—instead of acknowledging them as a first step in leaving resentment behind—would benefit only the counseling industry. Furthermore, Mayer-Schönberger does not seem to grasp the full implications of files with auto-expiry dates. Do we really want people to be able to erase their records, so that Van Jones can un-sign the 9/11 petition that cost him his job, and Senator George Allen’s “macaca” moment is negotiated away with an iPhone app? Or would we prefer a world that allows Van Jones to reenter politics despite knowledge of his past?
• • •
Mayer-Schönberger is at pains to emphasize that his anxieties are about memory, not privacy. For him, the problem of digital forgetting transcends the purely transactional concerns over who has access to what data; even if access to private information is uncompromised, our mental balance could still collapse under the burden of too much data.
Whether his concerns are cast as problems of memory or privacy is extremely important from a policy perspective. To regain our ability to forget, we need to fine tune the temporal dimension of information and establish firmer control over the time period that a certain file can stay online. To regain our privacy, in contrast, we would need to fine tune the contextual dimension, establishing stronger controls over who can access a given file and under what circumstances (time, of course, could be one of the many contextual factors to consider here).
There is nothing wrong with digital memory per se, as long as the information is only remembered in appropriate contexts.
The most disturbing anecdotes in Delete all have a social dimension and usually involve a third-party that acquires access to data that it was not supposed to see. This sounds like a problem of privacy, even if Mayer-Schönberger resists the label. His resistance may reflect the fact that conventional understandings of privacy are not very useful in the digital era. Still, once we factor in complications, nuances, and uncertainties associated with digital preservation, disintegration of our storage facilities, and the positive role that reminiscence can play in shaping our personalities, the façade of memory crumbles and the only unambiguous problem left in the book is that of privacy.
The idea of privacy may need to be rethought in an age of search engines and social networks. And for this purpose, the notion of “contextual integrity”—developed by NYU’s Helen Nissenbaum, and brilliantly summarized in her new book Privacy in Context—suggests a way forward. Nissenbaum argues that information revealed in a particular context always bears the tag of that context; no information is context-free and thus no information is “up for grabs,” even if it is revealed in public spaces. Thus, appropriating information revealed in one context (seeing your supposedly straight colleague enter a gay bar) and inserting it into another (sharing this knowledge with other colleagues) would violate your colleague’s privacy, even though the events took place in public view.
Applying this framework to the digital world—dominated by social networks, behavioral marketing databases, search engines, and gossip boards—yields a number of illuminating insights. The problem with sites like Facebook is that they force us to list anyone whom we do not actively hate as a “friend.” Usually, this leads to “context collapse”: unless we are adept at operating the site’s byzantine privacy controls, we have to treat our coworkers, neighbors, and high-school buddies in the same fashion, disregarding the subtle differences in communication that emerge in the course of human interactions in the real world.
Search engines violate contextual integrity as well, often by propelling strictly local situations to the global de-contextualized stage. An interview given to a local college newspaper ten years ago may have a higher placement on Google than one’s current professional biography. As college newspapers are digitizing and putting their archives online, many alumni are increasingly uneasy about the consequences: all of a sudden, their long-forgotten college transgressions pop up on Google, becoming visible to their current and prospective employers. There is no one good way to deal with this problem. Some college newspapers refuse to deal with alumni complaints, some redact the archives, some—such as The University Daily Kansan—tinker with their Web pages to ensure that Google cannot find the articles but they are still accessible online to those who already know about them.
Seen from the perspective of contextual integrity, there is nothing wrong with digital memory per se, as long as the information is only remembered in appropriate contexts. Stacy Snyder may still want her buddies to see those embarrassing photos from their college days even when she is 50. Indeed, for the appropriate audience, those photos will only increase in value over time. Setting auto-expiry dates on them is an overreaction that would only aggravate the problem of self-censorship. Auto-expiry suffers from the same problem of context collapse that plagues Facebook: people who should and should not be seeing these pictures 25 years from now are lumped together as “friends,” or, in Mayer-Schönberger’s proposed solution, “enemies.”
Allowing Snyder to delete the photo whenever she wants to is useless if she has no control over who accesses that photo before she deletes it. Thus, as Mayer-Schönberger himself predicts, those who are concerned with privacy will be unimpressed with his efforts to recast these problems as issues of forgetting.
And they are right to be. Attempts to force companies to adopt more transparent and restrictive policies for retaining and sharing data online as well as to make those policies more visible and better understood by their users would be more effective and productive than a fight to give users the ability to destroy their files whenever they wish. Instead of succumbing to the binary nature of choices forced upon us by technology—with its delete/keep, friend/non-friend, and public/private dichotomies—we should resist them at all costs and strive to introduce more granularity into our decisions. This would require bringing concerns like the ones raised by Mayer-Schönberger back under the umbrella of privacy—perhaps, in Nissenbaum’s re-conceptualized form—rather than finding completely, but not very helpful, substitutes for them.