Libraries Lead!
Libraries Lead! is a provocative podcast about all things information & library hosted by Beth Patin (Assistant Professor, School of Information Studies, Syracuse U), Dave Lankes (Professor, iSchool, U of Texas), and Mike Eisenberg (Dean/Professor Emeritus, iSchool, U of Washington). Information age opportunities and challenges affect every aspect of human existence. We wrestle with such topics as social justice, political unrest, mis- and dis-information, kids, family and adult living; education and learning; work, employment, training and jobs; recreation, entertainment, and play; disasters & emergency preparedness with a focus on libraries & information science, services, and systems. 4 segments in approx. 1 hour: WAZZUP, AI WATCH, MAIN TOPIC, and AWESOME LIBRARY THINGY. For Resources & References for All Episodes please go to: https://tinyurl.com/libleadresources
Libraries Lead!
Episode 52 (Sept 2026): Information Shock!
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Every so often the way we make information changes enough that the field built around it must rebuild. It happened after WWII when a flood of scientific data and the digital computer turned librarianship into library and information science. It is happening again. This time the cause is a machine you can talk to. When learning happens inside a conversation, there is no document left to check, and AI literacy runs out of things to grip. Generative AI triggers a major "information shock"—a fundamental disruption on par with the printing press or digital computer that forces fundamental changes in how we create, store, retrieve, use, and evaluate information. Dave makes the case; Beth and Mike push back.
Greetings and welcome back to the Libraries Lead podcast. If you're interested in our back episodes or getting more Library Leads content in digital print form, go to our partner, the Library Journal website, specifically our website, librarieslead.libraryjournal.org. And we're happy to have Bob's Better Bookmarks back as a sponsor this season, Bob's Better Bookmarks. You made it to page 50 before you realize the author is a misogynist jerk. We have a bookmark for that. I'm David Lankis, the Bowden Professor of Librarianship at the University of Texas Austin, here with Beth Patan, Associate Professor at Syracuse University School of Information, and Mike Eisenberg, Dean Emeritus at the University of Washington's high school. Coming off our summer hiatus, we're going to be talking about how generative AI is changing library and information science as a field. But first, Beth, please remind us of why we're here.
SPEAKER_00Hi, everyone, and welcome to Libraries Lead. This podcast is about life in the information world, the changes happening right now, the challenges we're facing, and what might be coming next. The show is for anyone interested in libraries and information work, whether you're deep in the profession or just library curious. And honestly, we love that so many of you listening aren't librarians at all. People like Miss D of Miss D's Pampered Pets and Syracuse, who does the best job of taking care of Walter, so I can record this podcast without interference. Today's episode follows our usual format. We'll start with a what's up, checking in on what's going on in our lives. Then we'll move to an AI watch where we react to different AI developments, technologies, and research. After that, we deep dive into our main topic and we'll wrap with our awesome library thingies, where we highlight amazing things happening across the globe in the library and information world. So let's jump right into it. Dave, what's up with you?
SPEAKER_01Well, you're going to hear a lot of what I've been doing on my summer vacation, but um more recently, I have new eyes. Um I had cataracts, and so a month ago they did one eye, and then Tuesday they did the other eye. So uh it's been a month of everything blurry, either with glasses part blurry, without glasses part blurry. So I'm looking forward to actually being able to see again. So that's that's been sort of occupying my mind. What's up with you, Beth?
SPEAKER_00Yeah, so surprising no one, I managed to catch a little bit of music this summer. I saw Mo a couple of times and fish a couple of few times. So that was nice to have a break. I I worked a lot. Um, but what I want to spend my uh what's up with me thing on is to take a moment to celebrate that this July, my former doc student, Tyler Youngman, was named the winner of the 2026 iSchools Doctoral Dissertation Dissertation Award. Um he won out of 35 nominations um across the globe. So it was a lot of competition this year. His dissertation is called The Monumental Challenges of Past Making. And the central question that he was asking in his dissertation is how are epistemicide and past making related through storytelling? And past making is the idea that like agents we exercise power when we're using documents to like make truth claims to talk about what is true. And this is kind of about who gets to tell the story of what happened in the end and like which version becomes our version. And instead of looking at documents, he looked at monuments and how they tell stories and how people think about those monuments and whether things are true or not. And he looked at Columbus Circle here in Syracuse and then the Pan Am uh 103 place of remembrance on campus to look at how documents afford or deny experiences of history and how like consensus gets built around these monuments. So he's a professor at uh East Carolina University right now, and he's working on a community archives lab and kind of getting to continue this work. So we are very proud of you, Dr. Youngman. We're lucky to have you in our field, Mike.
SPEAKER_02Wow. Well, um, I'm gonna go back to just the personal side. So uh it was just about a year ago, maybe a little over, that we lost our puppy Penny, who he'd had as a family member for over 15 years. It was really tough, and it was a it was a tough year. I mean, um, Penny was really special, but finally uh Carol came around and we started looking for uh another pup to adopt. We wound up adopting Georgie girl from Mississippi, who uh took a long trip to get here, and I have been spending the past six weeks kind of uh she's only uh well, she's six months now, so we got her just a little before five months, and I have been trying to uh uh help to potty train her and other train her and make her not get up at 5:45 every morning. Um, and uh I will talk a little about it, but the AI has really helped me, I have to say. It's like having a personal assistant dog trainer right there whenever I need it and uh giving me advice. She is barking upstairs right now because she's in her crate for her break and she's out of control, but that will end very soon, according to the AI. Anyway, it's really nice to have her to have a dog in the house again, is uh actually pretty special. So that's mine. All right, let's move to segment number two, the AI watch. Uh, two to three minutes each, okay. And let's start with Dave. What do you got for AI watch this month, Dave?
SPEAKER_01Anthropic Claude will start adding invisible watermarks to AI-generated text. So, what's happened is the European Union has passed a set of regulations, and in there um you have to watermark AI-created images, music materials, including texts. And so Claude is the first to announce it's doing it, but um, within milliseconds, it will also be OpenAI and Gemini and the others. So probably not Grok, but that's okay. Um, and yes, the idea is that uh you should be able to identify what AI is generated. And it's this will be not just metadata, it will actually be in how it statistically chooses the wording and the text. So it's it's it's it's a stenographic way of inputting the text. So um once they start inputting that, it'll be done in models starting after August 2nd, and they'll go back and add it to previous models. They'll then produce a series of tools where you can input text and it will say with some authority whether the text itself was AI generated. This will go, I think, a long way to some of the different policies. This is not perfect in many different ways. Um, one, if you take the text as presented and do a significant amount of editing or potentially move it across multiple uh AI formatting and grammar checkers, it may well lose that statistical signature. Um, it is not something that the you can tell the model not to do because it's actually done in how the information is tokenized and input into the system. Um and it will have a real interesting mark. Right now we live in this world where you know there are all these AI text um identifiers, you know, upload some text and it'll tell you whether it's AI generated of varying levels of sophistication and of accuracy. Um, it turns out if you're a really good writer, sometimes you sound like AI to these different systems. And so it's not going to be 100% fixed. There's still going to be things where it's going to identify a person's word choice as a false positive, and there are going to be things where they'll get false negatives out of it. But it's certainly going to change the conversation, um, particularly as I'm in the middle of working with MIT press and publications. They're all really interested in finding out was were the words AI generated. And so I think it's going to be a really interesting change shift addition to that conversation. So that's what I've got. What about you, Beth?
SPEAKER_00Yeah, I uh read a couple of different articles and um that were all kind of about like, is AI going to replace search or how is AI replacing search? Is is is maybe a better uh summary. So one of the articles, and I'll obviously post everything, but one of the things that I read that was pretty surprising is 37% of consumers are starting their information searches with AI tools rather than traditional search engines. And that number is increasing. And at the same time, we also saw that about 80% of the Google searches are ending without a user clicking on a single website, without a single link. So they're increasingly getting answers probably from the AI summaries that are popping up. Um, and so those are being termed as zero-click searches. And those aren't exactly new because we used to see that with like the info boxes that popped up, but we're seeing this increase, right? And so that means people are asking a question, they're getting an answer, and they're leaving. They're not looking into the source, they're not following up, right? They're not engaging with any of the ads, maybe that are on other places, right? Like if if websites were using ad revenue, right? And these things are scraping that data, that is coming into play. And one of the things that I thought was particularly interesting as we think about AI and and for our deeper dive later, is not only is the way that or like the tools that we're using to search different, but how we're searching is different. So when we're searching on Google, we tend to be typing in a three to five-word query versus when we're going to AI, we're asking, having much longer, much more conversational, multi-turn interactions, refining quests, require requests and interests through dialogue, which is the reference interview, right? Like it's it's essentially having a reference interview with you to kind of figure out what the right thing is as you kind of continue to refine your interest. Um, so I I also search isn't dead. It's very it's it's been stable. So people are still using search, but I think we're starting to see them use search for like what coffee places are near me and like hotels in New Orleans, right? When does fish play in Huntsville? Those kinds of things instead of these, how do I train my dog? And so we're seeing this shift. And I think as we continue to think about, you know, information retrieval, this is one of the things we're gonna have to consider as we're teaching how people find information in 2026 and beyond.
SPEAKER_02Mike. Yeah, so Beth, it goes back to the mothership. Bob Taylor, 1968, question negotiation and information seeking in libraries. Bob was the dean at Syracuse, uh, who really transformed it into the School of Information Studies. He hired me in 1979 and uh was a great guy. And yeah, uh, what goes around comes about comes around. I think that that really relates to what we're gonna talk about today. And it relates to my AI watch, which is um has to do with using AI. And again, the reason is that whether you agree with AI or not, whether you think it's a positive or negative, even if you think it's gonna result in the singularity and enslave us all, um, its use today is a reality. And most of us use it, and it seems to be somewhat, or if not a lot, effective. With that in mind, though, I started to think about it from a librarian perspective. And as a user of AI, as an information and library scientist, what are the things that concern me the most? Well, at the top of my list is the issue of credibility of sources, right? Um, not uh just identifying hallucinations or offering AI slop, but just reasonable exchanges. It's hard to judge the credibility of the response. Uh, you know, so and the and the chat box sometimes will give a reference, but not often. So what I decided to do, and I used Gemini, got into the habit of asking for the sources. And I kept doing that over and over. And then I copied my phrase and I would put it in, and all of a sudden it hit me why can't that be done automatically? Why can't the chatbot always identify sources and give me a list in APA format without my having to type in the prompt every time? And Gemini helped me to create it. So I created a script. I'm calling it the Eisenberg iPompt. And um, it it works incredibly well. You can embed it, and I have instructions and I'll include the uh a paper on a beginner's guide there. But I expanded it after just getting the references, I asked it to give me some more information about the sources. So I asked if it was primary or secondary, if it was peer-reviewed and vetted, or if it was unvetted, if it was descriptive or empirical or persuasive, uh, the authoritative. And then I asked it to make a credibility judgment just on its own, high, medium, or low. And uh after many iterations, I finally came up with a working prototype that all my Gemini searches now have that. And when I use it with uh Claude, um, I had to enter in a uh set of instructions, and then I just have to start with iPrompt, and it'll do it every time. And it really works. And as an example, the other day I looked up a person that I had met and played music recently with, a guy named Barb McCody. And uh he had said when we were playing, he's an old guy and he plays fiddle and some of that, he said he was in the New York State Country Music Hall of Fame. Well, I asked Gemini about it and whatever, and sure enough, he is. And the source they wound up using was uh a YouTube video called Little Theater on the Farm, um, that was done in 2020 because it had in it the uh the uh information about Bob's background and about him being in the Hall of Fame. And then it attributed the source as secondary. Uh it was a direct answer, it was unvetted, it was empirical, it was descriptive, uh, it was uh independent, and they thought the credibility was high. So I thought this was pretty interesting that every now, every time now, when I use uh one of the things it gives me the sources, and you combine that with what Dave is talking about, which is uh the um, you know, the the watermarks and the invisible identification, I think it could be relevant there. The last thing I'll say is that I didn't realize that when the AIs are doing a search, when they get a query, it makes a very high-level decision right off the bat whether to go to the web and outside sources or use its own corpus that it used in training. And it does that. And we never know that, right? Well, now it does. My uh thing identifies whether it came from its training corpus or pre-training or post-training, or whether it came from outside sources. That alone I found interesting. Now I know Dave Dave uh critiqued it a bit and played devil's advocate and said, Well, Mike, how do you know they even use the sources? Well, that would be a great research project. And Beth or Dave, if you have a student who would like to compare whether or not the AIs are using the sources that they cite, that would be an interesting little study. So that's where I have been with my uh AI these days. Okay, so let's take a quick uh collect our thoughts break and come back with our main topic, the information shock. And Dave is gonna introduce his main topic for the change, and we'll be right back to the music. Okay. So I can send you the uh the directive, Beth, if you want to use it and even share it with students, I'd be interested in.
SPEAKER_00Okay. It works pretty well.
SPEAKER_02Um I'm I'm into it. All right.
SPEAKER_00Yeah, I want I for sure, I for sure want to see it.
SPEAKER_02Yeah. And it's so funny when I ask it just a question like about the dog training. Sometimes it'll go to an outside source and sometimes it's from its corpus that it just knows it. That's that's like just saying, Oh, I knew that. I mean, it's it's it's it is kind of weird. Oh, the other thing I did was I it doesn't give me that cutesy little, hey Mike, really good question. Or or uh, you know, with the dog. Good job on the dog training today. You know, you've you've really got it now it now direct, it just gives me the just the facts, ma'am. You know, it's uh all right. Here we go. 10 30 on the nose. We'll we'll do about a half hour of the discussion. All right, Dave, you're gonna start it out. Yep three, two.
SPEAKER_01We're back for our main topic, generative AI AI. All right, start again. Sorry. We're back for our main topic: generative AI and the new information shock. It's 1945 and the war is over. In its wake, two things happened that would shape the world for the next 80 years. The first was a deluge of scientific and technical information that had to be translated, organized, and understood. From the Manhattan Project to the German rocket program, science itself was about to change forever. The second was the arrival of the digital computer. Organizing and sharing, all of that was going to happen in ones and zeros rather than ink and paper. That was an information shock. An information shock is when the production of information or a society producing it changes so radically that an entire field has to change its tools and its identity. Think of the printing press. Now, that was a shock. Similarly, in the decades after 1945, the deluge and the digital computer, librarianship became library and information science. Tools that tracked books became systems that organize digital documents, information overload, misinformation, biased search engines, information literacy, all of this is our field's response to that shock, and we are still working through it. Which brings us to today. I want to put a claim on the table. We are at the start of a new information shock, and generative AI is the cause. We have done a lot of talking about AI in the show, mostly as a tool we need to train people around. Yes, Chat GPT and Gemini can flood the world with fake images and synthetic books. That's real. But it's not what's new. We've been handling that overload and slop for decades. Here's what I think is actually different. Open a chat window and ask it something you want to know. It answers. You push back, it revises. You float a half-formed thought and it finishes the thought with words that you now recognize as your own. Six turns later, you close the window carrying a conclusion that was not there when you started. No citations, no list of links to work through, but just a conversation. Someone just built a belief and there was never a document in the room. The documents that trained the system were shredded into tokens and weighted parameters long before the conversation started. We have spent years bracing for a tsunami of fake news. The harder problem is a system that can build a false belief convincingly in a voice that manages to be friendly and authoritative at the same time and leaves nothing behind to check. So, what is the job of the librarian and the information professional in that world? Mike and Beth, we just started this talking about the idea of people just taking answers and not going through the searches, of having conversations about how to train your dog and such that doesn't say cite something behind it and having to ask the system to make citations if we need them, because that's no longer the default.
SPEAKER_02you think what's going on well dave i i i i think you're really on to something and because it is a uh a to use the word significant you know groundbreaking all it it's it's more than that and it only happens or has happened in history you know a handful of times uh writing itself and like you said the the the the codification in libraries and then the the printing press and that stuff and certainly the digital computer the web and all that but this is all this is different and what I liked about some of your writing and things I've I've uh talked to you about is one of the big changes is that we're it's not a document anymore the result is not a document we do searches it's what Beth was talking about in her AI watch we did searches and we come back with a website essentially a document just in a different form than a print thing that we're used to but it's not a document anymore what we're engaged with is an exchange right and the exchange is dynamic and changing and it develops and when you're done you don't go back and consult the whole exchange you kind of look at the result and and that uh it's one of the reasons I like to have the citations along with it because the other thing besides the exchange is that I'm not alone anymore. It's not me doing the search it's me and something else which is the system itself. So it's not me querying the system it's me and part of the system querying a large language model, right? And that you called it a conversant because we're engaged in a conversation. So those two concepts to me the idea of the exchange and the conversant are replacing the fundamental building block of our field. Not that the building block of documents or books or whatever is going away they will continue. But there's a new sheriff in town and the new sheriff is the exchange and the sheriff's tools are the exchange and but the sheriff is the conversant excuse me and the tool is the exchange and I have no idea what the implications are necessarily of that because I think we're just playing it out. But the role of the librarian as helping our patrons users people that use our that seek out information for for us to to make better use of that exchange may be part of our role. Beth you haven't been uh conversing with Dave all summer about this like I have so what do you think come into it a little more fresh?
SPEAKER_00Yeah so it's one of the books that I always rely on is the black the black swan and by um Taleb Nissan somebody I'd have to look it up but he talks about it it's about risk management and predictability and he talks about and I always thought about this principle a lot in relation to her Hurricane Katrina um in how there are moments in time where nothing after a specific event is the same as what happened before it. And I think it you know so like if you are at Tipentina's in New Orleans in 1999, that's a whole different thing than in 2010 even if it's the same New Orleans is an entirely different place after COVID, right? Like that might be like a at a national level or international level another significant moment. And I think you know in terms of technology tools this idea of information shock thank you the yeah thank you um is uh is similar to this and you know he he also talks about libraries and unlibraries in this book which I also always found fascinating right and Mike kind of like what you were talking about all the things we know and all when you were talking about um the AI systems like is it just checking what it was trained on versus going out and getting external sources and it's like what you're trained on is your library and everything else out there is your unlibrary and maybe you're gonna bump into it but maybe you're you won't right and like if you don't know this resource exists you won't ever bump into it. You know and I think so I I think you're right this is a moment that is fundamentally changing not just what's being created it's also changing what's being destroyed which you know let you y'all know I will want to talk about that but um I do think that this is not just profoundly changing how people are interacting with documents um but how are we going to train librarians? And are the curriculums that we have now sufficient and I'd say no you know and we are training people to do cataloging to do database design to do reference interviews um and it's not that any of those things are wrong but like how else do we need to be thinking about training them do we need to be training them to think about mediation about knowledge construction about what it helps uh what it means to help someone know something rather than just find something which is what AI what we are using AI to do right it is not um it is not a search engine it is like a decision engine that some bad search engine tried to say it was a long time ago but this is really helping you decide right it really is helping you create those belief systems um and it reminds me too of Tyler and I wrote a paper about dis and miss epistemologies and kind of what happens when these misunderstandings or bad information becomes part of our belief systems. Well it becomes very hard to like disentangle that if the information is wrong, right? Because it's now it's not just something I know it's part of who I am and people are walking away I think with like not just information where the closest and best pizza is but some kind of belief I want to say like unearned deep understanding about something that they probably need to be doing more footwork on. But as and especially if you think about the kind of language that AI is using, you know, this is a great question. You're it's so good that you thought about this. You're being a really good friend by doing X, Y, and Z. And so we can also see that it is also like encouraging you know the worst person that we know that they are correct. And so it isn't about how to help people find something how do we teach them how to know stuff and that's a different ball game Dave.
SPEAKER_01Yeah yeah a couple of things one it's interesting talk about an unearned I don't think it is necessarily unearned. I mean the question is if you're in conversation and you're looking and asking it to help you through a topic and tutoring we've seen this right the prompts as Mike says getting rid of prompts that tell us how wonderful we are prompts can also tell us that don't give us the answer help us to work through it. We've seen this um in lots of studies in AI and elementary and university which is absolutely it can be used to cheat and it can be used for the quick easy answer and walk away but it can also be used as a way of helping people think um as I've been talking about sort of submitting papers and it's like what's your AI usage? Did AI write this? And the answer is no I you I wrote the words but honestly it was written from conversations with my grad students and my peers and AI and having these as a conversant as someone who's discussing it. And so right it's not just you know it can help us earn that knowledge in some ways. And the question is where did that come from? And because now you have the knowledge but you didn't necessarily have this traceable auditable form.
SPEAKER_00It might if people are careful about using it the way you are deeply talking about using it.
SPEAKER_01Right. And and we also know this is I mean a lot of this came out of because AI literacy bothers me. And I was trying to figure out why it bothers me. And actually in some ways that goes back to an information literacy problem. We have studies that show when you judge someone's ability to do something, find information credible, you then have them go through an information literacy workshop class, et cetera, and you ask them at the end, did they get better? And the answer is always yes. And then when you ask them to do it, you find out it didn't change their behavior at all. Right? But in essence it built confidence without actually shaping behavior. And so and and that's why I think it's it's worth studying that's why this notion of AI literacy is the idea of well what sources did it train on? Who knows? Did you verify the document it created it didn't create a document. You know it it's not uh I didn't write an essay on who I was going to vote for. I voted right I didn't write an essay on who how I was going to train my puppy I trained my puppy right it's building knowledge in the way we've always built knowledge which is a con is a conversation. You could we're able to talk to idiots beforehand too that would lead you to false conclusions. But it's and that's why I think we're seeing this massive shift to suddenly people talking to these systems and not going to the sites and not going to the sources because that's actually a more organic way in which we come to our beliefs and our understandings. And and so you know the other part that that hit me because I'm sure that at some point if they're still listening some of our listeners like both of them have one of them has rolled their eyes and said oh God AI again and oh it's you know get over it it's just the new it's just a new tool. This idea of a shock is happening now and it doesn't require the development of artificial generative intelligence and the singularity and all this other stuff. It can happen you know be like everyone just pointed out a way that it's happening. We know that people will Pew did a study people read the answers that AI give at the top now just like in the back in the old Eric days in the 90s we knew that they read the the abstract and never read the article but they read that and they walk away from it. We know that you mentioned before there is this sort of market collapse of links that that these AI systems and answers, they're giving the answers and not pointers. And so people aren't getting ad revenue and what's that going to do in general it's happening now. And the question becomes what do we do about it? Including, by the way, what do we do about the unethical way that these commercial AI systems are working environmentally, et cetera, which is if this is going to be a way in which we are learning and how we are interacting with these systems, we need to make sure that the systems are ethical and reliable and there's nothing that says just like we're not quite sure what's going to happen in five years, neither is OpenAI or Google. And so we can be talking about shaping it, right? It two years ago OpenAI or Claude wasn't thinking about watermarking but because of EU regulation it's now going to be watermarking. So there's a chance to change these things to be more ethical but it's still coming down to right now it changes our field because the idea that somehow there's a document that we can interrogate that stands still, that we can audit that we can learn about whatever that goes away. And even someone said well but you'd still get the transcript that's not what actually happened. I mean think about you know the the the conversation Dave Lankus wants to know you know why is David Lankus an asshole right and I'm going in and I'm having this conversation because I'm looking for something but I'm not you know explicitly stating it. We know this all the time, right? Or how do you make a bomb? Well I wasn't asking because I wanted to make a bomb I was asking to see if I could skirt around safety guardrails and et cetera and so just the actual transcript doesn't say what was learned or the beliefs that came out of that. But Beth, I I you brought up a beautiful point that that I think needs to be in this conversation, which is the other thing that these systems are doing are normalizing or standardizing discussions and conversation points, which could absolutely erase um people's you know indigenous knowledge systems, people's unique knowledge within its people's perspectives. And that can either be a way that's going to help break down a lot of the barriers and and um fragmentation we see in society because now they're all getting similar answers from a source or what it does is it absolutely and can do both of these things loses the richness and the locality and unique nature of what happened um of what's going on.
SPEAKER_02You know it's it's not though it's not black or white it's not it is gray there's a continuum there's uh on this and I think what the EU just did and this whole thing which I wasn't aware of about the watermarks and stuff that's a a really specific positive example of regulation and forcing these systems to be more transparent. I think frankly my IPrompt is a small step in that way to make it more transparent and then we can study it and whatever. You know I was thinking of the question well if we could turn back the clock five years I guess we have to go back that far and and not have this happen. Right? No AI doesn't exist we're talking today about Google and other search engines and tools and whatever would we really want that um or is this a natural evolution in our information systems and in the way society interacts with information in some ways are these types of conversations richer and more revealing and uh valuable than simply a Google search that gives you paid listings at the top right or that is biased in some other way so you know uh and I would say no I don't want to go back I I like using these systems I I want them to be better but I find it incredibly valuable in a lot of ways I just want it not just I want it to be more transparent I want to know more about it I want to be able to rely on it more in terms of credibility. And then we turn around to our field and you know if if I were in your guys's I would be calling I would look at our curriculum and our classes uh in terms of this today and whatever the topic might be let's say it is knowledge organization cataloging and those things okay what are the issues of knowledge organization in related relation to the exchange and the conversant right um in uh a class that might be about school libraries and information literacy instruction and stuff how does how do we change information literacy instruction in a way so that it is focusing on the prompt and task definition we would say in big six land or something like that. In other words how do we integrate and become we're create it's an information shock we're shocked okay um what how do we react we should react by studying by developing tools by rethinking the way we do things recognizing that a lot of people still go into libraries in order to get a resource or to get uh some uh you know recreational reading or some other thing um that we're not going to do away with that but there are some fundamental changes in information services and the way we dealt with them and online systems I'd like to know what ProQuest and EBSCO are thinking about right now. I mean what's going on there proprietary database systems and whatever and I've worked with both of them and they're wonderful companies um who are owned by other people now and stuff but um I'd like to know what they're thinking about. In fact that's a that's a really good thing. I'm gonna I'll make some contacts to find out what they're doing. And LexisNexis and all the others right um Dave what a last thing I'll say when we created Ask Eric back in the 90s um and the early 2000s well work through was it weren't we creating a human intelligence conversation and exchange and weren't weren't our ask eric experts um in some ways being that uh system conversant uh so in some way in fact uh your line which I still cite was we're going to use natural intelligence until artificial intelligence catches up and and the question is has it caught up um so uh answering that so so yes there was a human being and the reference transaction sort of right because if you uh right before we began this the idea everything comes around um Beth brought up you know this sounds like a reference interview when people are talking towards the resources and Mike brought up Taylor 68 Bob Taylor's question negotiation and and in that paper he identified four different levels of questions maybe it was five but it goes from you as a human being have this sort of ache something's missing I need to know something then you articulate something very hard you you sort of articulate to yourself this is what I don't know which is a hard thing to do.
SPEAKER_01Then you have to communicate to someone else then they have to articulate in some way that a system can answer it and then you get to the resource and you hope that through that chain there's something that resembled the original question. All of reference the the way we built up question negotiations in the reference interview was still towards a document right the idea was to get to a question that was clear enough that we could then provide it link it to an information resource. What happens when it doesn't end or it ends before you ever get to that document and resource and and so I was speaking at the IFLA conference and we did a panel sort of talking about how all these different things could play out. So there's someone on misinformation and disinformation there was someone on um curation and et cetera and the argument was if we set up a lot libraries to be a they use the word neutral I refuse to a third party that would build their own system their own AI system um that they could then assure that the information that was being torn into these models and such was accurate and was good information and et cetera. And the problem that sounds good except that's an old way of thinking it's an old pre-shock way of thinking what documents how can we array the the documents even if it's arraying the documents for ingest into these large language models you're still focused on the documents and are they accurate and are they real and the whatever and we know that first of all to whom for what right that that the there's great information for one topic that is really horrible information for another topic so who gets to pick that um you know the the the my two dads is a great topic for kids with same-sex parents it's also a great argument for the Christian right to make it so is it a good document or not? Do we train on it or don't we train on it? Right? It does it doesn't it bypasses those questions. When it really comes down to we have the ability to instill belief in these systems. People are going to walk away from these systems either having earned or unearned however we define that new beliefs um how do we deal with a world where those aren't auditable how do we deal with a world where those come from a non-human actor that comes into it and we're seeing it now it allows people to do things easy it allows things to be hard right we see this in the open access movement. For decades now the library community has been pushing the academics to go open access with their publications, provide everything in Creative Commons, put the information out there. And one of the primary benefits of doing this was you'll increase your citations right so you'll get paid in reputation for making these things available open access. Well now what happens is we put all this in and it turns into open access training materials For these large language models that now will provide that answer, that material, that resource to the individual, stripped of its original authorship. And so was that the right move to make? People like Chris Borsch and MIT says it doesn't matter. Open access is to be open access, it still achieves the others. People like Leo Lowe in Virginia are talking about things like, well, if you can't provide provenance, then we shouldn't provide you with documents and materials, right? So two different ways and this philosophy of openness, how do we settle those things? Right. So it's a big question that we have to ask, but I just wanted to start with the notion that it's we have to ask them now. We have to ask them now. We don't have to wait for generalized superintelligence and whatever. Even the idea that we chat with these systems was at some point a design decision. Open AI did not have to come out with Chat GPT. In fact, when they came out with Chat GPT, they did it as they thought a relatively trivial example, not something. They did not expect it to become the biggest thing on the web and such. And so now we're used to it. It sort of escaped the lab. But you know, if we'd gone back five years ago and said, let's focus on medical imaging and analysis and et cetera, and not turn into chat, we would be having a different discussion. But it's out there now. And even if you know the stock valuations go away and open AI goes bankrupt and et cetera, I can run these open source models, Chinese models as well as American models, on my local machine now, right? I I can run these things without having to have the large capitalized companies. They will grow eventually stale. But that ability to interact is now out there and I think changing what we're doing. I'm rambling. Beth, please help me.
SPEAKER_00I I have two points that I want to make. And one is that you said that reference interviews are about the document, but I completely argue that reference interviews are about trust, right? I think that it is about relationship building and it is about getting people to ask the right things and open up and clarify. And I think, you know, if I, you know, I think back to um Matt Saxton's work, people feel better about the interaction if the librarian smiled at them, right? Like the document didn't matter at all. Um, so like in some ways, you know, I'd say let's not just go all in that the document has always been what matters. The output, the production is not the only thing that matters in these interactions. And I think, you know, there are so many times. So I'm I'm finishing the second draft of the epistemicide book, and we reread chapter four yesterday. Um, I cried twice because it's a it's a hard, rough read. Um, and it's the chapter, the name of that chapter is called The Politics of Testimony. And I spend a lot of time talking about the ways in which Black women's intellectual thought is dismissed and devalued and demeaned. And this has been true in academia for a long time, through gatekeepers and publishing, through peer review, through needing to have citations, through the devaluing of like what community work and what public scholarship is worth, as you well know, Dave, right? Like what it means to like focus on these other things. And um so much of our knowledge, because we haven't been allowed in these spaces for long, doesn't happen in this way. It is not about publishing a document, it is about sitting on your grandma's porch and having your cousins braid your hair and learning those family stories and those history stories. It's about the the conversations we have at the kitchen table. It is embedded in the songs and the music and the dances, right? Like it is the ways in which information spreads is so different for communities that have been disenfranchised from these information systems. And so, yes, I worry that like how we get represented in these things and the kinds of biases will be replicated because they were never really working on our things, anyways, right? Um, and this isn't just true for us. I think about the the um project that I've been working on with the 10th Mountain Division. We know that the official documents of the army are not the stories of the soldiers and the service members and their families, those are not the same things, and so we can see what the army puts out and we can see what gets collected. That is not the same as the testimony of them, right? And so, yeah, I I think there is this way where it's not just documents, it's this other stuff too, that information gets built around. Um and and I don't know, and I mean, you know, that can change too. And so like I I think it'll be it'll be important for us. And and again, why we need to be teaching our students, and I this will be the last thing I'll say, and I'll I'll wrap up. Um, I'm designing a new research methods class for our master students. And instead of just teaching them quant and qual and mixed methods and analysis, we've really started it from a place of who gets to know? How do we know? How do we decide when enough information is enough? How can we think about these fallacies that get built into arguments? How can we really kind of destruct knowledge as it's coming across our faces? And so, you know, I hope that that is the kind of deeper reflection we need to be giving our students and how to think about stuff, how to think about what we find rather than just finding the stuff.
SPEAKER_01And Beth, I think you put it beautifully because for me, the the shock is for 80 years we have worked on things like precision and recall and relevance and the notion that you know how do we build a system that brings the information? And it's forced work like yours often to go to to answer to that, right? So the idea of saying, no, no, it's beyond it or it's more than that, or well, the documents that it's building on to do recall precision is precise, right? It's framed the argument around, it's framed the field around stuff, right? How do we find the stuff, how do we organize stuff, who is stuff, who says it's that stuff, how do we prioritize the stuff? And moving forward, the shock is if we're if if we're going back, you know, if if the central organizing way of understanding how people acquire knowledge and information is in transactions, is in exchanges. Everything we know about exchanges, marginalized voices, social cues, how we build trust, how we build credibility, all these things move to the top, and those document things move to the bottom. That's the good news. Is that so so the article, which please follow the links.
SPEAKER_00Fingers crossed.
SPEAKER_01Yeah, the the articles are talking about a new inform a need for a new information science. And and the idea is not to say everything we used to do is bad. It's like, what do we prioritize moving forward? And the shock is going to force us to be real, it's gonna force us to go beyond. Well, we gave them the best 20 documents. We don't know why they didn't find the part about the minority voices, right? Well, right, we you know, we we had it was a highly relevant document. Look at our Sigma score, et cetera. Now it comes down to what is our responsibility to people walking away from these systems that the beliefs they hold hold with a societal notion of better, right? And and so, and which is what I find fascinating is that when you look at all the papers and discussions around generative AI, chat AI these days, very little, I mean, yes, there's a whole technical conversation going on, but a huge part of the conversation is how do we make it behave? Who picks the behavior? How do we build these guardrails, which is behavior, right? How do we build on community knowledge, et cetera, which is fantastic? And to realize that that's no longer sits on the side, that's now right in the middle. And a lot of the work that we have been doing around assuring using documents as surrogates for all of that. Well, we'll make sure the document is representative of this voice, or it comes from a black author, or et cetera. Now we have to say, is that lived experience really being communicated in conversations about this topic? Um, do we take people seriously when they have these conversations, right? We know that, for example, um, people with visual impairment, that's something I'm very much on my mind these days, rely on these generative AI systems to do picture to text. They want to know what's there and what they want to see. Um is it good enough for them? Are they are they hearing what we see? Is it representative of how they might interpret that item or that document, et cetera? So the shock is for 80 years, we have used documents sitting still to determine whether they're mis or disinformation, whether they're authentic or not authentic, whether they're relevant or not relevant, whether they're precise or not, how we organize them, how we look at them. And for the next however many years, we need to stop looking at those documents and start going back to, yeah, but did people learn? Do people do better now? Are people smarter now, more informed now? And that means that you can't do that by saying, here's the result set and here are, you know, here are the sponsored links. You have to get into looking at people's beliefs and cognition. And so that means that even our document-centric work, I'm not saying burn the collections, even though they're being torn apart literally to do these things. But the question becomes do those documents that we store and collect and organize become the reality check, right? So that you know you learn it here, but you check it there. Or it, you know, even Marshall McLuhan, way back in the 60s and 70s, was talking about how the book had become an art object. You no longer needed books as a functional tool. You could learn it all electronically and digitally, and the book became this sort of it feel good and it smells good, and I'd have this affective relationship to it. Is that gonna change as well? And so I just worry that if people don't look at the shock, libraries and librarians are going to say, well, that's right, this part's never gonna change. Just like back in the days of the web, when it was everything on the internet's free and you get what you pay for. And um, if you find it on the web, we can't trust it. It's gotta be in a print volume, et cetera. It took us years to get past it. We are going, we're now in this world of if AI was involved in the development of this belief, it is therefore suspect, right? That's what every journal AI policy is now. What did you, what did it write? What words did it produce? What were the conversations about it, et cetera? And we provide that all to a reviewer. Whereas if it was like, I talk with Beth a lot, they don't want a transcript of all of your conversations to go to the reviewer. Right. And isn't peer review on the blind notion looking at the outcome and applicability, not the necessarily the process other than methodologies being exempt, right? It's this philosophy of document centrality is now it's we're trying to bend old models to a new reality. And I don't know what the new tools are or the new names are, but I know we better start working on them now.
SPEAKER_02And with that, and with I am gonna call it a uh call an end to this part of the discussion. I think we should have part two probably next month, which and and you talked about students a lot, but it's also our information scientists and librarians in the field that we need to talk about and bring them in because they're the ones who are on the front lines and they also um have tremendous knowledge and experience. So I'm gonna end it now. I'm gonna remind everybody that our citations and links to the content provided in this episode are on our website, libraries lead dot libraryjournal.com. We're gonna take a short break and come back with our final required segment, the awesome library thingy. Cue the music. Okay, we're back. And now for our awesome library thingy. Uh, let's start with Beth. What do you got for our awesome library thingy?
SPEAKER_00Yeah, my awesome library thingy this week is the United Way of Central New York's book buddies program, which just expanded to public libraries here in central New York. And it is such a cute and great little program. Before the pandemic, the book buddies had about 100 kids that were involved, and now the number has grown to 600. And what happens is community members volunteer their time to help kids improve their reading scores. The program pairs volunteers one-to-one with students and they come in during lunchtime so that they can practice literacy skills with an adult from the community. And so now the literacy coalition of Onondaga County is using that model and they partnered with the county library systems to host that program there. Um, the volunteers work with kindergarten through third grade in the Syracuse City School District for just about an hour, uh a half an hour to an hour a week. Um and so here in Syracuse, after this whole deep AI conversation that we had, uh there are 600 kids sitting across from human beings, holding a book, building a relationship embodied with a real thing and real community. Um, and so maybe, you know, we can think about a volunteer, a kid, a library branch, and a book, and that like, you know, not all hope is lost. The kids are okay. Dave, what's your awesome library thingy?
SPEAKER_01My awesome library thingy is an awesome library person. Um, Eric Bukstein um is the he's a technically now he works for the the uh KB, the Royal Dutch uh National Library as a consultant. And but really he's sort of the library uh whisperer, working internationally for years and years, helping build innovative services across the globe and design new services. This uh last week was the IFLA World Congress, the International Federation of Library Associations, and Eric was awarded the IFLA Medal for his distinguished contributions to IFLA and to advancing public libraries, connecting people, and inspiring creativity and innovation in libraries and in professional practice across an inclusive global library field. That was in Busan, South Korea. He is amazing. He's unfortunately talking about retiring uh at some point if we let him. Uh he'll keep he'll keep we'll keep him busy. But um, I just wanted to give a call out to Eric because ever since he's been working on the Delft Library, when it was the best library public library in the world, and et cetera, he's just an amazing individual and such a well-deserved recognition. What do you got, Mike? Way to go, Eric.
SPEAKER_02Yeah, well, mine has to do with uh it's something from every library, but it had to do with the recent uh elections in August. Um that showed tremendous grassroots support for libraries across the country. Uh, even in places like Missouri, the Wester Groves Public Library in Missouri had an 82% of the voters, actually 83% if you round it up, uh, supported them uh for their library budget and their measure. Um the Pawpaw Library, gotta love that, right, Beth. The Pawpaw Library District in Michigan had 68% voted uh a strong mandate for their library uh and uh supporting it there. And then in uh my favorite place, Seattle, um Seattle had a huge, huge uh uh levy for public libraries. We're only talking about 479.76 million dollars. And it and in a very difficult time financially, and even a lot of people in the Seattle area, 72 percent uh uh approved the library uh levy, which uh accounts for one-third of the library's overall funding. And we see this all over the country. So I really think that you know these book banning and closing library initiatives and all this other stuff, they all get pushed back when you actually get right down to it. Communities love and support their libraries, and that's um, I think the message. And every library has been very involved with helping and training people on um working with their communities, on uh getting the vote out and providing support for their information campaign so that the community realizes what they have in their libraries. So that's my awesome library thingy for this month.
SPEAKER_01And that's a great way for us to conclude this podcast session. Thanks again to our partners, Library Journal, for services and support. LJ is available on the web at libraryjournal.com and social media platforms, and of course, Bob's better bookmarks. You got books, we got marks, let's get together. Please subscribe to us. That's right. It's a whole nother season. It's been off for the summer. I'm back with the same plea. Please, please subscribe to us and rate us on Apple, iTunes, Stitcher, Spotify, and wherever you listen to your podcast and or contact us via email at info at librarieslead.org or post uh on our Facebook group, Libraries Lead. And again, our resources and references are available through our website, librarieslead.libraryjournal.org. Bye, Mike. Bye, Beth. Bye bye.
