Method: The reading list — five tests before a book gets recommended

"The Common Shelf" is Toronto's public reading list — a curated set of books meant to give people shared vocabulary for civic problems. A book earns a place on it the same way a claim earns a place in our research: independently checked, never taken on faith or from memory. This page explains how a title actually gets onto the list, and how we keep it honest.

What it produces

A public reading list with two visible tiers: a short, high-bar core list of the books we think are most worth a reader's limited time, and a wider tier of additional titles for people who want to go further. Every entry on the visible list carries checked bibliographic details — title, author, year, and a working way to actually get the book, whether that's a library, an e-lending service, or a bookstore.

How it works, step by step

We start by pulling candidate titles from work we've already done — books already referenced in our own research on a given issue, or already surfaced by earlier reading recommendations — rather than inventing a list from a fresh, ungrounded search. Every candidate has to trace back to a specific reason it's on the table.

Every candidate then has to pass five tests before it earns a place on the core list:

1. It contributes real vocabulary — named ideas a room full of people can actually reuse in conversation.
2. It's readable in about two weeks by someone busy who isn't a specialist in the subject.
3. It's actually available — in print, and something a library or bookstore can supply in real quantity, not a title that's effectively impossible to get.
4. It's verified — title, author, publication year, and identifying details checked live against a real library or bookstore record, never pulled from memory or copied forward from an earlier list without re-checking.
5. Toronto carries weight — when two books are otherwise equally good, the one that's actually about Toronto wins the slot.

A title that fails one of these five doesn't necessarily get dropped entirely — it may still belong on the wider tier, which doesn't carry the same bar, but it won't make the core list.

Every title that does make the list then goes through a live verification pass: someone actually checks the title, author, and bibliographic facts against a current library or bookstore record, rather than trusting whatever an earlier pass wrote down. This step exists because we found real errors when we skipped it — see below.

Separately from whether a book's facts are correct, we also check whether a reader can actually get their hands on it: does it have a working library catalogue link, a working e-lending link, a real bookstore link? A book can be completely correctly described and still come up short here if, say, its e-lending link is broken — and when that happens, we mark it as partially sourced rather than pretending the problem doesn't exist.

Every correction, demotion, or drop we make along the way gets logged with a reason and a date, in an internal record kept specifically so the history of what changed and why is never lost — even though that internal log itself isn't what gets published; only the current, correct list is.

How it's checked

Concrete enough, we hope, for a reader to challenge us directly:

• Nothing is verified by memory, and nothing is verified by trusting an earlier list's citation. Every title on the list has been checked live, against a real, current record, at least once — a prior pass having listed a book is never treated as sufficient on its own.
• The five-test gate is the actual bar, not a general impression of quality — a book that fails any one of the five tests does not belong on the core list, whatever else is good about it.
• Getting the book and describing the book correctly are two separate checks. A book with perfectly accurate bibliographic details can still be marked as only partially available if the actual links a reader would use don't work — we track that as its own honest status rather than folding it into "verified" and hoping nobody notices the link is dead.
• Corrections are logged, not silently fixed. When we catch an error, we write down what was wrong, what it should be, and when we caught it — so the record of our own mistakes stays visible rather than disappearing the moment they're fixed.
• Nothing on this list is published just because a file exists. Publishing the list is its own deliberate step, separate from doing the work of building and checking it.

What it honestly costs

This is steady, careful compilation-and-verification work rather than the kind of judgment call that needs our most senior review — closer to careful bibliographic fact-checking than to original research. From the one full pass we've actually run: the checked core list runs to roughly a hundred titles, with a further few hundred in the wider tier — a single, contained verification surface, smaller in scope than our full research corpus. The remaining open item is a straightforward one: someone needs to give the final go-ahead before the list is published, and as of this writing that hasn't happened yet.

Known limits

Said plainly rather than glossed over.

• We don't yet have a standing schedule for re-checking links. A library or e-lending link that worked on the day we checked it isn't guaranteed to still work today, and we don't yet have a rule for how often those links get re-verified after the fact.
• How we sort books into the list's smaller thematic groupings isn't written down as a rule. Certain books are grouped by the kind of reader they're aimed at, but the reasoning for which group a specific book landed in exists only implicitly, in the choice itself, not as a separate documented standard.
• What earns or loses a spot on the wider tier isn't a fixed, general rule either. We can point to the specific reason any one book was added or dropped, but we don't yet have a standing, reusable rule that would tell you in advance whether a new candidate belongs there.

What stays internal, and why

The list starts life as an internal working document, and it says so on its own face until a separate, deliberate decision publishes it. The internal correction log — the record of every mistake we caught and fixed along the way — stays internal by design: it's a working QA record for us, not reading material for the public, and if any trace of it ever appears publicly it will be a short, general note that this list is actively checked and corrected, never the log itself, and never any of the earlier mistakes it documents laid out in detail. We think that transparency about our process matters more than looking like we never make mistakes — but the right place for that honesty is a page like this one, not a public airing of every draft error.

Every title on this site's public reading list carries a receipt, including where AI assistance was used in compiling and checking it. Run this in your municipality if you want to build a version of this for your own community, or read How we verify for how this checking discipline runs across the whole site.

— The Unknown Soldier · uniteTOlove · Toronto