A flip side to the concept that guides Google, Wikipedia, and so many other internet and computer firms–that we need an ever-expanding capacity to store and access informaton–is the idea that computers should learn how to forget. A recently published working paper from Harvard University suggests that there could be good public-policy reasons for implementing this kind of solution:

Why would we want our machines to “forget”? Mayer-Schönberger suggests that we are creating a Benthamist panopticon by archiving so many bits of knowledge for so long. The accumulated weight of stored Google searches, thousands of family photographs, millions of books, credit bureau information, air travel reservations, massive government databases, archived e-mail, etc., can actually be a detriment to speech and action, he argues.

From a technological perspective this sort of thing seems like it could be very easy to implement. Microsoft Outlook, for instance, already has a feature that will delete or archive all messages after a certain date. Companies that have found archiving every email to be either costly, embarrassing, or both are increasingly adopting “email retention policies” to limit their legal liabilities.

These practices seem fairly “dumb” (in the computing sense of the term). That is, it might end up being a lot more difficult to implement “smart” forgetting of information. It seems like it would be very difficult to teach a computer or a program to forget based on multiple criteria that approaches the level of sophistication of the human brain. We could very easily tell a computer to save all messages from a certain person whom we’re especially close to and with whom we share intimate communications. But some of these are bound to be trivial and worthy of forgetting. Still, sometimes a message that looks trivial based solely on its content might have a tremendous amount of emotional significance for its sender or its recipient. A child’s finger painting might not be a Monet, but his mother still saves it for decades.

The idea that computers would be smart enough to discern what to keep and what to get rid of based upon an almost endless number of factors, including less tangible emotionally based ones, suggests sentience and humanity.

Leaving aside the implications for artificial intelligence, the author of the working paper focuses on the issue from the standpoint of public policy. Would it be a good practice to implement? And, how could it best be implemented? My concern here is that future historians researching their past might be deprived of important ways of understanding our society. Then again, they might learn a lot about us by considering what we programmed our computers to forget.

The complete article from Ars Technica: “Escaping the data panopticon: Prof says computers must learn to ‘forget’ “

The complete working paper from Professor Mayer-Schönberger, John F. Kennedy School of Government, Harvard University: “Useful Void: The Art of Forgetting in the Age of Ubiquitous Computing”