Tuesday, March 13, 2018

The "Grand Challenges" of Curation and Preservation

I'm preparing for a meeting next week at the MIT Library on the "Grand Challenges" of digital curation and preservation. MIT, and in particular their library and press, have a commendable tradition of openness, so I've decided to post my input rather than submit it privately. My version of the challenges is below the fold.

The challenges of curating and preserving digital information are fundamentally economic. These are things we know how to do, but we don't know how to do at a scale commensurate with the problem we face. Even very optimistic estimates are that less than half of the academic papers or web pages that should be are preserved. The selection of, and metadata for, those that are is the subject of incisive criticism.

Assuming that the goal is for the digital information that is curated and preserved to be open access, two economic analyses are relevant:
  • Cameron Neylon's Sustaining Scholarly Infrastructures through Collective Action: The Lessons that Olson can Teach us applies:
    First, it illustrates that the problems of sustainability are not merely ones of finance but of political economy, which means that focusing purely on financial sustainability in the absence of considering governance principles and community is the wrong approach. The second key insight this approach yields is that the size of the community supported by an infrastructure is a critical parameter. ... Olson describes three different size ranges of groups: those that are small enough to reach an agreement to provide the collective benefit; those that are too large to do so; and those that lie in the middle, where the collective good may be provided but at a level which is below optimal.
    Only in small groups, whether of individuals, institutions, or governments are social pressures strong enough to prevent free-riding.
  • Brian Arthur's Increasing Returns and Path Dependence in the Economy from 1994, which explains how the increasing returns to scale and network effects endemic to digital networks ensure that there will be one, or at most two, winners in each niche in the ecosystem.
But at a more granular level the economic challenges of curation and preservation are quite different.


The two main aspects of curation in this space are selection, and adding value by enhancing metadata, both human activities that don't scale. The Internet Archive's non-selective approach to collection is the foundation of its success. As I wrote back in 2013:
What matters isn't the perfection of a collection, but the usefulness of a collection.  Digital preservation purists may scorn the Internet Archive, but as I write this post Alexa ranks archive.org the 167th most used site on the Internet. For comparison, the Library of Congress is currently the 4,212st ranked site (and is up despite the shutdown), the Bibliothèque Nationale de France is ranked 16,274 and the British Library is ranked 29.498. Little-used collections, such as dark archives, post-cancellation only archives, and access-restricted copyright deposit collections are all at much greater economic risk in the long term than widely used sites such as the Internet Archive.
Although the process of adding technical metadata has been automated, the value of doing so is negligible and even potentially negative. The value of other types of metadata, such as provenance and bibliography, can be significant, but it is doubtful that the cost of staff time to generate them is justified. It can be argued that machine learning could automate these processes. Human processes are biased, but the humans involved are far more diverse than the small set of geeks whose biases would be built in to the machine learning algorithms.

Via its "Save Page Now" feature, the Internet Archive crowd-sources part of its selection policy. Crowd-sourcing selection would be a good way to scale the selection part of curation. Alas, there are risks here too, as the bots and troll farms rampant in social media show. Two years ago Kalev Leetaru wrote:

of the top 15 websites with the most snapshots taken by the Archive thus far this year, one is an alleged former movie pirating site, one is a Hawaiian hotel, two are pornography sites and five are online shopping sites. The second-most snapshotted homepage is of a Russian autoparts website and the eighth-most-snapshotted site is a parts supplier for trampolines.
Just as the cat videos don't impair YouTube's pedagogy, these facts don't impair the Internet Archive's usefulness. Its highly automated collection process may collect a lot of unimportant stuff, but it is the best we have at collecting the "Web at large". But for crowd-sourcing to be effective, it has to happen on a single platform.


Preservation happens in three phases; ingest, preservation and dissemination:
  • Ingest is the most expensive phase, and it is subject to strong economies of scale, both in terms of infrastructure (the Internet Archive sustains about 20Gb/s inbound), and in terms of staffing. Ingest at this scale would be technically challenging even if Web technology stood still. The rapid evolution of the Web from quasi-static hyperlinked pages to a JavaScript programming environment requires a highly skilled, fast-moving team to keep up.
  • Preservation is less expensive, primarily because it is less staff-intensive. But it is subject to even stronger economies of scale, which have powered "the cloud" to displace on-premise storage at scales below about 10PB. The economic and business risks of cloud storage make it an inappropriate way to preserve our cultural and scientific heritage.
  • Dissemination is typically the cheapest phase, at least at scales below the Internet Archive's 40Gb/s outbound. But that is because preserved collections other than the Internet Archive's get relatively little use, because network effects drive traffic to the dominant player in the niche (see Google, Facebook, Netflix, etc.).


Displacing the Internet Archive as the go-to resource for preserved digital content would require not merely building a bigger, better curated collection, but also capturing the mind-share that has kept it in the top 300 sites on the Web over decades. It isn't going to happen at a single institution, since the dollars that it would take would be far more effective in increasing access to preserved data at the Internet Archive.

Could a collaboration among many institutions using decentralized Web technology displace the Internet Archive? Chelsea Barabas, Neha Narula and Ethan Zuckerman's Defending Internet Freedom through Decentralization from last August surveyed the various efforts to decentralize the Web and, despite their efforts at optimism, showed that antitrust is the only feasible way to displace the FAANGs (Facebook, Amazon, Apple, Netflix, Google). Herbert Van de Sompel's Paul Evan Peters award lecture entitled Scholarly Communication: Deconstruct and Decentralize? describes a potential future decentralized system of scholarly communication built on existing Web protocols. But even he prefaces the dream with a caveat that the future he describes "will most likely never exist", and I wrote a long exploration of the reasons for such skepticism.

Even if decentralized Web technology were ready for prime time (it isn't), it is inherently less cost-effective at delivering capacity and performance than centralized systems. Since the fundamental problem we're facing is lack of funds, this isn't a good direction to take.

Instead we should build upon the centralized system we have. The Grand Challenge of digital curation and preservation is to find ways to sustain and enhance the Internet Archive's capacities for crowd-sourced curation, at-scale preservation, and openly-accessible dissemination. An additional $10M/yr would be a big step in the right direction, but running ads against the preserved content, or mining cryptocurrencies in visitors' browsers aren't the ways to get it.

Of course, this all assumes that the Grand Challenge isn't to find ways to curate and preserve a Web locked up by Digital Rights Management, as specified by Tim Berners-Lee and the W3C.


Mark Graham said...
This comment has been removed by the author.
Mark Graham said...

Two comments to your excellent post.

The observation that Kalev made about some of the sites most archived by the Wayback Machine, for a given time period, was caused by a “bug” in our crawling technology… and was not a direct artifact of the provenance of our (crowdsourced) crawl directives. Of course that is not to say that undesirable bias can not come from the crowd (esp. as leveraged by bots.)

And regarding the Wayback Machine’s Save Page Now feature. You might be interested to know that we archive, on average, about 80 million URLs (base URLs and embeds) / week via user submissions to that service.

brewster kahle said...

David-- you are right-- working together is the only way. Lets work together better and more. But the Internet Archive has a start of this going now-- most of our income comes from libraries.

The Internet Archive works with over 500 memory institutions now to create the 3000 web collections through the subscription web archiving service archive-it.org (including most everything being visible in the Wayback Machine). We can do vastly more, and vastly more together.

The Internet Archive works with 500 libraries who have paid us to digitize their books. These go in archive.org and openlibrary.org. Again we can do vastly more and together.

We have worked through the copyright balance through respect and trade-offs.

The Internet Archive is willing and interested. Please contact us

David. said...

As regards the biases in the ingest process,in Comparing the Archival Rate of Arabic, English, Danish, and Korean Language Web Pages Lulwah Alkwai, Michael Nelson and Michelle Weigle report that in their sample of Web pages:

"English has a higher archiving rate than Arabic, with 72.04% archived. However, Arabic has a higher archiving rate than Danish and Korean, with 53.36% of Arabic URIs archived, followed by Danish and Korean with 35.89% and 32.81% archived, respectively."

See also here.

David. said...

Michael Nelson's The Internet Archive Can't Preserve the Web's History by Itself uses the media frenzy about Joy-Ann Reid's homophobic blog posts to make a strong case for the importance of multiple, independent Web archives using different technologies to collect, preserve and disseminate Web content.

In an ideal world there would be many, well-funded Web archives, and users would know to use Memento aggregators such as the Los Alamos National Laboratory Time Travel service as the way to access the history of the Web rather than simply using the Wayback Machine.

But that's not the world we live in. It is noticeable that it took the involvement of Web archiving gurus to find the copies of the posts at the Library of Congress and archive.is, and of a team at Old Dominion to analyze and explain the differences and gaps. If we're going to continue to have multiple Web archives the public and especially journalists need to be educated in how to use them. If they aren't used they won't be funded.