Ashleigh Faith, MLIS, PhD, EBSCO 

Search, revise, and repeat. This is the habit in which users often find themselves when searching through mountains of content. To be faced with a repository of information and not find one’s desired knowledge is not a new dilemma, and now with AI, one never knows if what is retrieved is accurate or authentic. But the challenges faced today are not that much different than the issues we as librarians have seen and helped navigate through, before. As with many real-world problems, literature often takes issue and parodies it to highlight the extremes of its vices and virtues. Dating back to the 1800s, science fiction in particular has parodied themes of knowledge, retrieval, and where humans and machine interact, and for a genre that often blends technological mysteries with reality, has touched upon many of the artificial intelligence (AI) conundrums we face today.

Where Science Meets Fiction

As James E. Gunn writes, the mind of a science fiction writer “is concerned with writing entertaining fiction, not predicting the future” but “in the process of entertaining, the science fiction writer may chance upon ideas that make us imagine more dramatically the nature of the problems we face” (“The Road to Science Fiction,” 1979). There are many examples, a few of which can be seen in “The Cerebral Library” by David Keller (1931), in which recording knowledge is accomplished by harvesting human brains; H. P. Lovecraft’s “The Shadow Out of Time” (1936), in which the knowledge of all scholars throughout time is collected in an alien library; Eric Russell’s “The Hobbyist” (1947), in which an alien repository is so large it takes up an entire planet and which adds to its collection only through osmosis of the human mind. Jorge Borges’s “The Library of Babel” (1956) and Stanislaw Lem’s “The First Sally” (1967) in particular describe the wonders of infinite knowledge and the aggravation of not being able to find any of it unless it is organized effectively. Borges writes about a repository that contains the knowledge of the entire universe, but which has no organization, and Lem writes about a digital pirate who is buried under the knowledge they have gathered because there is no way to find anything in the repository. If all knowledge could be recorded and synthesized by machines, John W. Campbell describes in “Twilight” (1934) that without the pursuit of knowledge, if all things are known, humans will lose what is most special about us: curiosity. Vonda McIntyre’s “Starfarers” (1989), adds to this by describing how without human insight, a digital knowledge ecosystem will most likely drown humanity in irretrievable knowledge.

Organizing for Humans, Not Just the Machine

These early science fiction writers were describing a highly digitized world with issues that, although revolutionary for the time, are common in history when new technology is introduced.

For example, as Stefan Decker states, knowledge organization like subject tags and other metadata has a primary focus “on human beings and not on machines…while we create and generate data for machines, the purpose is to support us as human beings (Technology Voice, 2013)” and while science fiction depicts AI as a thinking, sentient entity, AI is not at that stage currently. The information being generated by AI is still meant largely for human consumption. AI can help sort through the vast amount of information in repositories, help humans summarize that information into bite-sized pieces or contextualize the information in relation to the body of knowledge on a topic, and generally help streamline tasks, but if the knowledge that the AI assembles has no contextual metadata such as its information sources or whether the generated output is fact or fiction, it may be misused or never found at all, which we can see ample examples of from science fiction. What’s more, just like David Keller wrote, the pursuit of knowledge needs to be paved carefully and ethically so that humanity can be recognized on the other side of that pursuit. All of these are topics being discussed globally when it comes to responsible AI and the ethical considerations involved with attaining knowledge for the use of AI.

Implicit vs Explicit Knowledge

Whereas much of the past study of knowledge organization focused primarily on explicit information such as the text and images in Creative Works, metadata such as subject tags, and search logs containing the queries of anonymized users, AI now can tease out the implicit, the sentiment and context behind the corpus of information given to it and add “flares of phrase” otherwise known as hallucinations. Hallucinations are the way AI mimics human creativity or turns of phrase to make information sharing more engaging and easier to remember in the form of stories. But just like in libraries of future past, some works are just that: stories. They may be founded on kernels of truth, but they are flights of fancy, nonetheless. For instance, Mark Twain may have depicted a “Snapchat-like” invention with his depiction of the telectroscope which would allow the sharing of “the daily doings of the globe (from “The 'London Times' of 1904” (1898), one would not want to use it as an authoritative source of the history of social media (not the internet as so many counter-histories depict).

The concept of infinite knowledge access is a theme from the past and is certainly a focus of many current, and predicted, AI trends. While explicit knowledge from text and images is easier to gather, gathering tacit knowledge, where shared experiences and wisdom is understood, is more difficult. For example, as a librarian speaking to other librarians, we would not need to describe what “walking the stacks” means. It is a shared understanding. Often, people aren’t even aware of the knowledge they have, which Mather and Leeds (2014) call "unconscious competence." This means people do not think to share what they assume others already know. Mezghani, Exposito, and Drira (2015) developed a simple two-step process to capture this kind of hidden knowledge: first, gather shared knowledge through a survey, then organize it into a knowledge graph, with participants reviewing it for accuracy.

Now think of AI and its assemblage of all knowledge (or at least that is what the intent is behind Artificial General Intelligence -or AGI). The knowledge is assembled from explicit human behavior and information, but the second step of the process is missing, that of verifying the knowledge is accurate from authoritative sources and from the humans who originally generated it. Thankfully, fact verification and Retrieval Augmented Generation (RAG) tactics are now allowing more human oversight into the outputs of AI, which once adopted by those using AI to increase accuracy with AI tools, and which librarians are leading the charge to establish frameworks for ethical and subject matter expert review of AI training and outputs.

Knowledge Discovery and Retention

And that’s the final point. AI is just a tool. A very smart tool at parroting human behavior, but still a tool to be directed and guided by human users. Humans have a finite capacity for retaining knowledge, and the time it takes to find the right question to ask, seek the resources of knowledge, understand what has been learned, test the knowledge in real-world applications, and take those learnings and pass them down to the next generation, all of which stresses the ultimate human constraint: time. But speed up time too much and most of the activities in the pursuit of knowledge dwindle to the point of insignificance. The answer is not always the best end result. Oftentimes the journey is the more meaningful interaction. The cautionary tale of being buried in knowledge is one thing, but never having to seek knowledge again is quite another.

The Timelessness of Search

But similar to the AI in science fiction, AI has a need for an infinite stream of information, and just as these early science fiction writers speculated, finding the knowledge in a sea of information, knowing what is truth vs what is fiction, and having enough discernment to wade through the information available to find what you need amongst the noise, is a constant battle. But a battle that is all too familiar. These challenges of progress remain no matter how far we have progressed because, at least for now, humans are still the primary consumers of knowledge and no matter how far technology progresses, we will still need help discerning fact from fiction, finding what we need, and struggling to retain what we learn. All things that librarians happen to be good at helping folks navigate through, from the past and so on into the future.

This was a fun play on libraries and the challenges of knowledge organization and retrieval. If you would like to learn more about EBSCO’s journey with AI, visit our website.

SPONSORED BY