Using How Search Decides What You See in a course

What it covers, and what it deliberately does not

The book is about the retrieval half of a search system: how text becomes searchable, how records become candidates, how they are ranked, how the pieces are assembled into a pipeline, who decides what to search next, how retrieval fails, how you would measure it, and what a library must record and require. It assumes no prior information-retrieval knowledge and no mathematics — formulas are confined to footnotes, and every argument that depends on one is also stated in words.

Four things are out of scope by design, and it is worth saying so to a class that will expect them:

• How language models generate text. The book stops at the point where a model writes prose for a reader.
• Detecting fabricated citations. That is a generation problem; the book is about what was never retrieved.
• Prompt engineering. Not covered anywhere.
• General agent architectures and planning. One narrow slice appears — how an agent selects, repeats and stops retrieval actions — because that slice is retrieval design.

The running examples deliberately use recent internet slang — delulu, rizz, cooked, touch grass. This is not decoration: words a system has never seen, or has seen only in another sense, are where retrieval breaks in instructive ways. Some students will find it either funny or grating, and it is worth naming the reason in advance.

Three course shapes

Chapters are written to be assignable on their own, and each carries a stable link — use the Copy link control in any chapter heading rather than transcribing a URL. Every chapter ends with three Check yourself questions whose answers are hidden until clicked; they make usable reading quizzes, though note that printing a chapter reveals the answers.

A half-day workshop (about three hours)

For practising librarians who need the vocabulary rather than the mechanism.

1. Chapters 1–4 — the whole of Part I. It ends by resolving the category error it exists to correct, so it stands alone better than any other segment.
2. Chapter 10 on retrieval failure, which needs surprisingly little of Part II to follow.
3. Chapter 12 with Appendix D — the procurement and evaluation material.

Skip Part II entirely. Participants will not be able to explain why dense retrieval bridges paraphrase, but they will be able to ask a vendor the right questions, which is the point of a workshop.

A six-week module

One part a fortnight, with the application exercise as the assessment for each.

Part II is the heaviest fortnight at roughly 11,000 words. If that is too much, Chapter 8 on query transformation is the one to make optional — it is the most self-contained, and Chapter 9 reintroduces what it needs.

A twelve-week course

One numbered chapter a week, with the appendices distributed as optional reading. Appendix A (the Transformer) pairs with Chapter 5; Appendix B (tokenisation) with Chapter 5 or 6; Appendix C (inverted-index execution) with Chapters 2 and 3; Appendix D (terminology) with Chapter 12 or, better, kept open throughout; Appendix E (rank fusion and learning to rank) with Chapter 7.

This shape leaves room for the thing the book cannot supply: students running searches in systems your institution actually licenses, and reporting back. Chapters 10, 11 and 12 are considerably more useful to someone who has spent a week failing to make a discovery layer behave.

The application exercises

There are three, one per part. All require access to a licensed database or discovery layer, and all ask for a written deliverable rather than an answer. They are the slowest part of the book and the part that teaches most about local tools. The argument survives skipping them.

Exercise I — Explain a result list without invoking meaning

Tests: whether the student can resist reaching for “semantic” as an explanation. They run a natural-language query, record four observations, and must account for each using only query analysis, the execution rule, or query transformation.

A strong answer names a specific layer for each observation and explains why the other two are ruled out. It also takes the final instruction seriously — listing what could not be explained lexically. That residue is the honest part, and a student who reports none has probably not tested hard enough.

Common wrong turn: concluding that a missing query term proves semantic retrieval. That is precisely the inference Chapter 4 dismantles, and it is worth marking as a failure of the exercise rather than a minor error.

Exercise II — Map a product onto the pipeline

Tests: whether the student can distinguish what a vendor documents from what they inferred. Every stage must be labelled documented, inferred or unknown.

A strong answer has a lot of unknown in it. Most vendors do not document their indexed unit, their cut-offs or their fusion rule, and a map with eleven confident answers usually means marketing copy was read as documentation. The label distribution is the thing to mark.

Common wrong turn: treating a named model as an architecture. “It uses BERT” fills no stage on the map.

Exercise III — Diagnose a failure, then try to document it

Tests: diagnosis using Chapter 10’s workflow, a measurement from Chapter 11, and the record from Chapter 12.

A strong answer applies the remedy the diagnosis implies rather than the one the student would have reached for first, and reports honestly when it did not work. The second deliverable — a list of everything the product would not let them record — is the institutionally useful half, and worth weighting accordingly.

Common wrong turn: treating Chapter 10’s diagnostic lenses as four equivalent categories, or diagnosing every failure as vocabulary mismatch because that is the relationship librarians are most accustomed to seeing.

Discussion prompts

After Part I

• Systematic review search strategies are published so they can be inspected and rerun. What exactly is being reproduced — and what is not, given that indexes and analysers change underneath?
• If a database silently made every query term optional tomorrow, how would your users find out? How would you?
• Early Google was non-Boolean and non-semantic at the same time. Which current product do you suspect is being misread the same way?

After Part II

• Every embedding space carries a theory of relevance. Whose theory is in the tools you license, and would they say if asked?
• A reranker cannot rescue a document the retriever never returned. What follows for how you teach searching?
• If a system indexes only abstracts, which of your disciplines suffer most, and can you demonstrate it?

After Part III

• Agency is a trade. Name a task in your service where a fixed workflow is clearly the better buy.
• Your library builds a 40-query evaluation set. Who judges relevance, and what happens when two judges disagree?
• PRISMA-S concedes that some proprietary and similarity-based operations may not be fully replicable. Is that a reporting problem or a procurement one?

Before you teach it: what will have moved

Every claim about how a named product currently works carries the date it was checked. Re-verify these before teaching:

• Table 3.1 — which scholarly systems run on Lucene-family infrastructure.
• Table 7.2 and Figure 7.1 — the Primo Research Assistant pipeline traced end to end. This is the book's most load-bearing product example; if its documented cut-offs change, the argument still holds but the numbers do not.
• Tables 8.2 and 8.3 — query transformation and natural-language filter extraction across products. Vendors add supported fields steadily.
• Table 9.3 — where academic tools sat by control arrangement, tested April 2026. The book already warns this will date quickly, and the same vendor appears on two different rows, so brand-level generalisations were wrong even then.
• Figure 1.1 — the two-by-two of AI academic search. Products migrate between quadrants.

A useful exercise in itself: assign a student the job of re-checking one of these tables against current vendor documentation and reporting what changed. It teaches the documentation-reading skill the whole book depends on, and it dates the material for you.

Figures, and reusing them in slides

The explanatory illustrations live in images/ in the repository and are CC BY 4.0 along with the rest of the text — use them in slides with credit. They were generated with AI assistance, as the book's disclosure records.

The numbered figures built as HTML rather than images — the BM25 scoring walkthrough, the pruning panels, the WordPiece pipeline — are in the page itself and are easiest to capture by screenshot at a wide window. Each has a stable anchor, so you can also just link to it.

The screenshots of commercial products are not covered by the licence. They are third-party material reproduced for comment and criticism. If you are redistributing adapted material, take your own or leave them out.

Citing it

Cite the version as well as the date. The page is live and will change; the version number is what lets a student say afterwards which text they read.