appreciation for the risks to our data posed by “solutions” to other problems (such as DRM), and understand that data preservation is becoming a struggle with active adversaries— malware authors, political partisans, and scammers conducting phishing attacks. Commercial organizations have a mixed record as long-term custodians of personal artifacts and of cultural works.

So in the light of all this, what are some approaches designers and other stakeholders may be interested in exploring? After all, service, application, and interface designers will be the ones implementing the experience now, and thus have a direct impact on the future of our personal and collective digital memories. And who are the stakeholders whom we need to be talking to and designing with, for, and around?

Here are our top five clusters of points and questions on this emerging area. These are overlapping, and there are more, so consider these a seed list.

1. Guide users between backups, archives, and collections. Good design for archival services can help users make decisions based on anticipated future uses and perceived risks.

For starters, it is helpful to distinguish between archiving and backup. Apple’s Time Machine, which is part of Mac OS X Leopard, is an interesting step in the right direction. People report learning that a backup is not the same as an archive when old (but important) versions of files have been overwritten by backup software whose check boxes were clicked (or not). The options the check-

boxes offered required knowing the distinction. Perhaps systems need to ask questions like the following: “Are you sure you want to overwrite this file with all future versions?” Yes, that means overwrite it. Not store another version and keep track of all that you have done with the file.

Users must choose between a wide range of file format and compression options (think of ZIP, TAR, JPEG, MPEG, PDF…). Some are proprietary, some may be unsupported in the future, and some are “lossy,” meaning file sizes shrink by reducing resolution. Purists in the archival community rule out the use of lossy compression (MP3 or MPEG 2) altogether when there are non-lossy options available (FLAC or JPEG2000). But for personal collections of audio and video, lossy algorithms may be the best way to limit storage costs. Systems that allow users to preview the difference, or that explain the implications of loss, may help.

As professional librarians and archivists know, you cannot have archives without cura-tion. At a more personal level, psychologists view strategic forgetting as what constructing a (more or less) stable sense of self is all about. In this case, a question posed to the user might be, “Are you sure you want your kids to see this when they go through your archives?”

The importance of forgetting should not be lost on us. However, we need to guide users through these concepts with intelligently designed systems and interfaces if people are not going to inadvertently lose the digital materials they want to

keep. Unfortunately, the consequences of bad decisions may be felt only days, months, years, and decades later. It is hard to learn best practices when there is this lag, so once again designers need to surface the results of choices and knock-on effects at the time of action.

2. Be involved in conversations about the differences between algorithmic search and human memory. Over time we may be able to follow Google’s directive, search don’t sort, because improvements in search algorithms and applications will eliminate the need to file content manually. This search-don’t-sort perspective is also reflected in David Weinberger’s book, Everything is Miscellaneous, in which he explains how the ordering of our collections can be reworked on the fly, as the situation demands. This argument is most compelling if metadata is well designed and standardized. So, for this approach to work, we should be active in communities where forms and standardization of metadata are discussed. Simply asserting that people can be less careful about providing metadata because search is improving is an unacceptably risky approach for materials that are worth saving.

A complementary approach is to leverage our understanding of the way in which human memory works—by recreating context to facilitate retrieval. This would entail providing time frames punctuated by memorable events (salient or regular events), congruent activities (“I was working on the Rosebud project when I took that picture”), and so on. The point is,