When access to books is actually a book ban.
The student-run Amnesty International Club at Williams College has invited me to participate in a conversation with them this weekend about technology, AI, and information access as they relate to censorship and book bans (a slightly belated Banned Books Week event). So I thought I would use this space to work out some of the things I want to float for conversation: above all that too much information access can be a form of censorship.
One of the most important things I consider when thinking about AI in its relationship to the humanities and the study of history is access. The term "access" has an almost totemic charge in our culture because of its definitional appeal to openness and inclusion. When we use the term "access" with respect to information, we often think of "gatekeeping" as its opposite.
Calls to universal information access are coextensive with the life of the commercial internet, and if you read the (by now, hopefully universally?, notorious) Magna Carta for the Knowledge Age you'll see throughout it demands for the vaults of knowledge to be unlocked everywhere so that a vast market of open information can be availed to anyone to improve her condition. Of course—and this is all is well-trodden territory—none of this was what materialized. When power is distributed inequitably, then the empowering potentials of information access are distributed unevenly too and those with the capacity to both profit off free information and then use it to reinscribe inequitable power relations will do just that.
The point is that one of the normative assumptions governing American culture is that universal access to information is an unqualified good, and it's the responsibility of the producers and stewards of information to make it available. This sounds nice.
What does any of this have to do with book bans?
It has been almost 25 years since the launch of Google Books. On the surface...what an amazing boon. When I want to verify texts I've quoted in anything that I am publishing and when I don’t have the actual book to hand, I can just go into Google Books and quote-search those lines of text to ensure I didn't mis-copy anything. When I want to see if a book interests me, I look up an excerpt from Google Books. This is great. But the scanning that availed those books to me while I sit on my couch at home, the scanning that's been carried out by countless other institutions and corporations, the drive to get every library and every special collection, every museum, every archive to mount their holdings online, to make them "universally accessible" (and whether access is actually universal when you can't even get reliable internet in many parts of the US and the world is another issue) has brought us to our current moment in which information is alarmingly inaccessible. We have stuff, we have content, we have images, we have all the things. And yet it has become increasingly difficult to access information as such.
For one, the ease now with which information is availed to people by summary means that there is a screen between people and the actual sources of information they seek. That's the obvious part, and that's a kind of low-hanging-fruit argument about how generative AI has mounted barriers between people and information.
But I want to stoke a conversation that takes seriously how
(a) the welter of stuff available online is there because commercial interests drove us towards a political economy that makes disputing the value of universal information access a non-starter (and in which stewards of information and intellectual laborers including academic researchers, librarians, curators, and archivists are often vilified as "gatekeepers" of it)
(b) we are witnessing a manifestation of the weaponization of the backlog, an insistence that the most important thing is just to digitize and process everything at lightning speed and sort shit ("shit" being the quality data and metadata, not to mention further discursive framing) out later, but also often never, so that material is just brazenly being miscatalogued by OCR and Automated Text Recognition (this is like putting a library book on the wrong shelf; you can technically still access it, but lord help you trying to find it until the next shelf audit gets done in five years)
(c) every information ecosystem is a nuclear waste dump not simply of summary and slop but just plain unnavigability and un-useability because bots are busy scraping a library's every bit of data, and sorry, the servers are overwhelmed now
(d) AI has provided the permission structure for firing or not just not hiring the desperately needed caretakers of information and producers of knowledge because, well, you can just ask chat, and chat has access to all the things
(e) the illusory nature of universal access to information obscures not only how partial that information is, but how also stripped of context it becomes through every form of digital mediation and remediation one can imagine; access in this respect entails making context inaccessible
(f) the sheer volume of things being produced by putting all of this information in the AI blender and then lobbing it at people to process and verify—I'm thinking in particular of the Math Dump here—means that there a real people who have to devote their time to wading through this tsunami and not doing their actual work
And so, while we are right to focus on the evils of literal book bans, of demanding that libraries purge their stacks of volumes deemed by one group of people to be objectionable, I want, at this event, to invite us to think about how addition can also be subtraction.
