Monday, November 16, 2026

Success in Mathematics Does Not Put Us on the Path to Generally Truthful Large Language Models

Many people now say that large language models are on a path toward greater truthfulness because they are succeeding at mathematics, software verification, and tightly constrained science. A model that can produce a checked proof or confirm that code meets its specification, the reasoning goes, is learning to track reality. With more scale or better techniques, the same ability should appear in ordinary topics: policy, history, medicine, or how institutions actually work.

That does not actually follow, although you might think it would.

There are two different kinds of alignment for these models.

Objective alignment is possible when an outside checker exists that does not depend on human opinion. A proof assistant either accepts the steps or rejects them. A program either satisfies a formal specification or it does not. A scientific prediction either matches the measurement or fails. In these cases the system can be trained or searched against a standard that sits outside ordinary language. How often a similar claim appears in books or on the internet does not matter. Whether the claim is socially comfortable does not matter. The checker supplies an independent test.

Operative alignment is what most chat models actually move toward. It means matching the language, tone, and expectations that particular groups of people treat as normal or appropriate. The usual method is to collect human preferences—rankings of which answer sounds better—and adjust the model toward those rankings. The result can reduce some obvious harms and help companies manage legal risk. It does not supply an independent test of truth. Different groups prefer different things, and the preferences themselves change.1

The reason success in mathematics does not transfer to greater truthfulness is built into how these models work. A large language model is trained by repeated next-word guessing across enormous amounts of text. Each successful guess slightly adjusts millions of internal dials, called weights. What remains are settings that make statistically common continuations easy to produce. A true statement becomes easy only if it was also common and stable in the training text. When a source later proves unreliable, the model has no separate mechanism that does what a careful person does: lower the weight of everything that source has said, and raise the standard of evidence for related claims. Frequency still holds sway. A story that appears often remains easier to continue than a rarer story, even when the rarer story is better supported once the sources are checked.

This is an old human difficulty made larger. What people say, even sincerely, is often different from what is actually happening. Institutions and communities maintain flattering or protective accounts of their own purposes while the day-to-day incentives and results run differently. Strong emotion frequently defends the account. Established practices that have improved the match between speech and reality—scientific testing, courtroom cross-examination, structured peer review—exist because fluent or popular language is not enough by itself. They add outside friction so that claims must survive evidence and people who are rewarded for finding mistakes.

A language model externalizes the storytelling side of that difficulty. It has absorbed the public record of how people describe themselves and their institutions, then been further adjusted to sound appropriate to its raters. Where the distance is large between the common story and what is actually occurring, information that would shrink the distance is usually less common or less rewarded. The model therefore tends to extend the predominant account. Skill at formal proofs does not change this. That skill came from closing a verification loop that ordinary language topics do not have. More scale or longer reasoning traces can make the model more fluent and locally consistent. They do not add the missing ability to discount unreliable sources or to resist a dominant narrative.

This is why claims that one model is simply “more truthful” than another often describe a difference in operative alignment rather than a difference in truth-tracking. When Grok was described as more truthful, the observable difference was that it produced narratives matching a different set of political and social expectations. The underlying limitation remained the same.

Objective alignment success should not be read as progress toward truthful answers in areas that remain under operative alignment. Those areas stay governed by match to prevailing language patterns. Reducing the distance between the stories we (and the models) tell and what is actually going on still requires the same costly outside structures human institutions have always needed when speech alone is insufficient. Formal mathematics shows what is possible when a mechanical checker can be applied. It does not show that fluent language starts to carry its own guarantee of truth once a model becomes good at proofs.

1 Readers familiar with my earlier work may notice that “operative” is used differently here than in discussions of narrative and function. The uses are related. In both cases the word points to what is actually running and treated as standard—in this essay, the expectations that preference training matches—rather than to an idealized story or an independent test.

```

Thursday, November 12, 2026

New Webinar - "AI Slop in Collections: Practical Strategies for a Growing Problem"

31308122085?profile=RESIZE_710x

AI Slop in Collections: Practical Strategies for a Growing Problem
A Library 2.0 AI Webinar with Crystal Trice

OVERVIEW: 

AI-generated books and other materials are flooding the market, and they're not always easy to spot. Some contain inaccurate information, repetitive text, fabricated references, or questionable illustrations. Others imitate legitimate titles or authors, while misleading metadata and inflated reviews make it difficult to separate reliable materials from low-quality content. Even familiar purchasing channels aren't immune.

For library staff responsible for collection development, this creates some challenging questions. How can you recognize questionable materials before purchasing them? Which warning signs deserve a closer look? How much can you trust vendor listings, reviews, and publisher information? And what should your library do about AI-generated materials that arrive through patron requests, local authors, donations, or digital collections?

This practical webinar explores ten strategies for recognizing and avoiding AI slop in library collections. Through real-world examples, we'll examine what to look for in author information, metadata, reviews, covers, and sample pages. We'll also explore ways to strengthen selection practices, work more effectively with vendors, and address AI-generated content through collection development policies and procedures.

Throughout, we'll consider the gray areas, including legitimate books that use AI-generated elements, the limitations of detection methods, and the importance of professional judgment when deciding what belongs in a library collection.

LEARNING OBJECTIVES:

Participants will be able to:

  • Recognize warning signs of low-quality, misleading, or potentially AI-generated materials during the selection process.
  • Apply practical verification techniques using author information, metadata, reviews, sample content, and trusted sources.
  • Identify opportunities to strengthen collection development workflows, vendor practices, and policies to reduce the risk of acquiring questionable materials.

This 60-minute webinar will provide practical examples and evaluation strategies to help library staff make more informed selection decisions. Participants will also receive a printable Before You Buy checklist to use when evaluating materials for their collections. All registrants will have access to the recording, presentation slides, and checklist.

DATE: Tuesday, November 24th, 2026, 2:00 pm to 3:00 pm US Eastern Time

COST:

  • $99/person - includes live attendance, anytime access to the recording and presentation slides, and a participation certificate. To arrange group discounts (see below), to submit a purchase order, or for any registration difficulties or questions, email admin@library20.com.

TO REGISTER: 

Click HERE to register and pay. You can pay by credit card. You will receive an email within a day with information on how to attend the webinar live and how you can access the permanent webinar recording. If you are paying for someone else to attend, you'll be prompted to send an email to admin@library20.com with the name and email address of the actual attendee.

If you need to be invoiced or pay by check, if you have any trouble registering for a webinar, or if you have any questions, please email admin@library20.com.

NOTE: Please check your spam folder if you don't receive your confirmation email within a day.

SPECIAL GROUP RATES (email admin@library20.com to arrange):

  • Multiple individual log-ins and access from the same organization paid together: $75 each for 3+ registrations, $65 each for 5+ registrations. Unlimited and non-expiring access for those log-ins.
  • The ability to show the webinar (live or recorded) to a group located in the same physical location or in the same virtual meeting from one log-in: $299.
  • Large-scale institutional access for viewing with individual login capability: $499 (hosted either at Learning Revolution or in Niche Academy). Unlimited and non-expiring access for those log-ins.
12435796494?profile=RESIZE_180x180CRYSTAL TRICE

With over two decades of experience in libraries and education, Crystal Trice is passionate about helping people work together more effectively in transformative, but practical ways. As founder of Scissors & Glue, LLC, Crystal partners with libraries and schools to bring positive changes through interactive training and hands-on workshops. She is a Certified Scrum Master and has completed a Masters Degree in Library & Information Science, and a Bachelor’s Degree in Elementary Education and Psychology. She is a frequent national presenter on topics ranging from project management to conflict resolution to artificial intelligence. She currently resides near Portland, Oregon, with her extraordinary husband, fuzzy cows, goofy geese, and noisy chickens. Crystal enjoys fine-tip Sharpies, multi-colored Flair pens, blue painters tape, and as many sticky notes as she can get her hands on.

 

OTHER UPCOMING EVENTS:

November 12, 2026

31276315468?profile=RESIZE_710x

November 13, 2026

31278355890?profile=RESIZE_710x

November 20, 2026

31296954256?profile=RESIZE_710x

December 1, 2026

31301981095?profile=RESIZE_710x

December 9, 2026

December 17, 2026

31276319294?profile=RESIZE_710x