uniteTOlove.ca

The Learning City — Explained Simply

v0.1, 2026-07-19 Date: 2026-07-19 ·

Status: v0.1, 2026-07-19 · RENDER — derived from this library's internal records (charter, 2026-07-17), this library's internal records, and this library's internal records (both 2026-07-18), pilot per this library's internal records DRAFT, pre-ratification. The charter behind this program is still proposed. The operator has not signed off on it yet. Its flagship design, the Learning City, is still just an idea. It has not passed its own internal for-and-against review. No circle has met. No book has been picked. No child has been involved in anything described below. This page changes none of that. It only explains it.

This is written for anyone who wants to know what a Toronto research program is actually proposing about education. You don't need to read a charter, a concept brief, and an evidence review to find out — this page does that instead. It explains the plan, why it matters, how anyone would know if it's working, and where it could go wrong. Everything here comes from research already done for this program. Nothing below is a promise.

What it is

Picture a Toronto adult trying to keep up with a fast-changing world. About half of working-age Canadian adults find long, complicated reading genuinely hard — a lease, a tax form, a benefits application. Almost all of them can read at some level. But dense reading is still a real struggle for many. Only 38% feel confident actually using the AI tools already showing up at work and in their kids' homework. Nobody has built the thing that would help close that gap yet.

A Toronto research program is trying to design that thing. Its mission, in plain words: figure out how to raise, educate, and organize a generation for a world where AI does more of the routine work. The real work of running a decent life and a decent city still needs doing. The program's own motto is "Life is 99% practice, 1% theory." It covers a lot of ground: early childhood, K-12 schools, universities, gaming and attention, family and trauma, testing and certification, and more. Its single most fully worked-out idea is called the Learning City. That is the focus of this page.

The Learning City rests on one core bet. Literacy — reading, digital skills, AI skills, civic know-how, any kind — comes less from being taught and more from doing. Nobody can be forced to learn. People have to choose it. And if people are choosing to learn, the plan argues, they should do it the best way available: discovery guided by both AI and real human guides, not permission granted by an institution. That bet turns into four connected pieces. Each one is weak without the others:

something together, led by a trained human guide plus an AI helper. No one applies or gets admitted. The guide opens doors. They don't hold the keys.

strangers across Toronto have common ground to talk from. Who chooses that list is the whole trust question, and the design treats it that way.

what it already does — pointing people to free books, courses, and tools. No one's data gets collected or sold.

fixing something real in their own neighbourhood. The finished project becomes the lesson, and the proof that something was learned.

None of this exists yet. Every piece, and every model considered for building it, still has to survive an internal debate — arguments for and against — before anyone tries it for real. And the charter that funds and governs the whole program is itself still waiting on a final yes.

Why it matters

Toronto and Canada aren't starting from a crisis. The evidence behind this program says so plainly. Canadian adults score above the average of the world's wealthy countries on reading, math, and problem-solving. Canada also has one of the smallest shares of low-scoring adult readers among those countries. Ontario specifically sits a little below the Canadian average on all three skills. Researchers call that a real concern, even while the country as a whole does comparatively well.

One caveat matters here for fairness. A good part of the score gap between Canadian-born adults and immigrants comes from the test only being offered in English or French, not from lower ability. When established immigrants take a similar test in their own first language, that gap mostly disappears. Toronto has the country's highest share of newcomers of any city. A lower average score here likely reflects which language the test was given in, not what people can actually do.

Younger students show a real, if modest, slide. National math scores for 15-year-olds dropped 15 points between 2018 and 2022. That's roughly three-quarters of a school year of learning, by one rough conversion — and that conversion itself is approximate, not exact. The slide actually started back in 2003, long before the pandemic. About one in five Canadian students now scores below a basic level in math. The share below a basic level in reading also rose, from 14% to 18% nationally. Ontario's own share of low-performing readers rose too, even though Ontario's average reading score stayed roughly flat — a steady average can still hide a growing group falling behind.

In Ontario's own school test results, reading scores have been climbing over the past three years. But Grade 6 math has been stuck at only about half of students meeting the provincial standard. The gap for students with special-education needs hasn't closed either. Small year-to-year moves like these are normal noise, not necessarily a trend.

Underneath all of this sits a harder floor. About 58,000 Torontonians have no home internet connection at all. 38% of households have internet slower than the target regulators recommend, rising to 52% of low-income households. Lower-income households own under one computer per person on average. Cost, not lack of interest, is the reason people give most often. Any plan built on AI-guided learning has to solve that gap first, or it will only reach people who already have what they need.

AI itself is moving fast in the other direction. Well over half of Canadians have now tried an AI tool — 57%, up from 47% just five months earlier. Younger adults are moving fastest: 83% of 18-to-34-year-olds have tried AI, compared with 34% of adults over 55. But only 38% feel confident actually using one well. Canada ranks near the bottom of wealthy countries on AI training and literacy. That gap is new enough that no one has a solid way to measure it yet. These numbers come from people describing their own confidence, not from a real skills test.

AI tutoring itself has some real, if young, evidence behind it. One closely watched study found students learned roughly twice as much with a carefully built AI tutor as with a strong in-person class. But it ran just two short lessons, never checked whether the gains lasted, and used students who already knew a lot going in. A six-week pilot in Nigeria, using AI-guided lessons after school, raised test scores measurably. But it was small, teacher-supervised rather than solo AI, and helped already-stronger students the most. Neither result proves AI tutoring works at city scale yet; both are real, promising, and still young.

It's worth being honest about the core idea driving all of this. The program's working belief is that literacy comes less from being taught and more from experience. Nobody can really be taught; people have to choose to learn. AI- and human-guided discovery is supposed to be the fastest way to do it. The real research on this is genuinely split, not a clean yes.

For someone learning something totally new — a child sounding out first words, an adult starting from zero — carefully guided, explicit teaching reliably wins. One well-known research review found unguided discovery performing clearly worse than direct instruction. A separate, large study from the 1960s and 70s compared teaching methods for at-risk young children. It found the same thing: scripted, explicit teaching beat open, child-centred methods on nearly every measure. That study has its own critics, though. The schools involved weren't randomly assigned, and most people citing it loudly today are themselves advocates for that same method.

For people who already have some grounding, especially adults, the picture flips. Discovery that is actively guided — not left alone, but not spoon-fed either — was the best-performing approach of all the types tested in that same research review. It worked even better for adults than for children. The honest reconciliation: how much guidance a person needs depends on how much they already know. Start with clear, direct teaching. Ease into guided discovery as a person's own competence grows.

The idea that "nobody can really teach us, we have to choose to learn" turns out to be well supported — but only as a claim about motivation. As a claim about which teaching method works best for a total beginner, it isn't supported; direct teaching wins there. The plan's actual proposal — guided, not unguided, discovery, aimed mostly at adults — already sits in the part of this research that holds up best.

The shared reading list has its own honest evidence behind it. City-wide reading programs like it are well documented elsewhere, but not for raising test scores — researchers are upfront that they don't have proof of that. What they do reliably produce is shared reference points and conversation across a community, which is the actual goal this piece is going after.

How would we know it's working

Like the rest of the program, the Learning City doesn't launch all at once. Right now, it hasn't launched at all. Before any of the four pieces could even be tried, the whole concept needs to survive its internal for-and-against review. The charter that governs the program needs the operator's own sign-off too.

If and when it moves forward, the plan is built in small, careful steps. One library branch, working with one neighbourhood — and only if Toronto Public Library actually agrees to partner. The plan refuses to go around them if they say no. Three small circles of adult volunteers, meeting six times each. One book, chosen by an open vote from a short list built in public. One person on hand to help people find free library resources, keeping only combined, non-identifying notes on what people asked for. One small neighbourhood project per circle, lasting six weeks or less and costing nothing, ending with a public show-and-tell. And, deliberately: zero children involved, zero accounts or personal data collected from anyone who joins, zero branding connecting it to the movement behind the program.

The plan also states, in advance, exactly what "not working" would look like for each piece. It doesn't wait to decide that after the fact:

doesn't work. Stop and redesign it before scaling up.

two separate tries, the circles aren't holding people's interest. Stop.

people's choice. The same is true if anyone — including someone from the movement itself — is caught trying to rig the vote. Either way, the whole process gets redesigned before trying again.

stops, and rethinks using neighbourhood venues instead.

idea that doing the project teaches the skill. Stop and redesign.

evidence the whole four-part idea is wrong, not just one piece of it. The concept goes back for a full re-review.

improve by the second round, the plan pauses expansion. It reworks outreach instead of scaling anyway.

What could go wrong

The sharpest critique of the whole idea comes from the plan's own authors. They wrote it out rather than dodge it: this could just be school reinvented with fewer safeguards. The research really does show that people with little background in something learn worse through open-ended discovery than through direct teaching. Volunteer-led book clubs, courses, and lecture series have historically ended up mostly reaching people who were already motivated and educated — not the people furthest behind. A reading list picked by the wrong process, paired with civic "projects" that quietly double as free labour, is one bad decision away from becoming a recruiting tool with a library card taped to it.

The plan's own response does not deny any of that. It agrees the research is real. That's exactly why the design insists on guidance rather than leaving people to figure it out alone. That's why it plans to measure who actually shows up, rather than assume. That's why drifting toward recruitment is itself listed as a reason to stop. But the honest caveat holds either way: none of these safeguards are proven yet, only designed on paper. Picture a real pilot that shows the pattern critics predict: some people get nothing out of it, and participation skews toward people who didn't need the help in the first place. That is exactly when the plan's own kill criteria are supposed to fire, not get explained away.

Children are part of the program's wider goals but not the Learning City pilot itself. So the plan draws a hard adults-only line for now. No circle, project, or reading list involving a child happens until two separate approvals are in place: a named safeguarding plan, and the operator's own sign-off. The operator alone cannot waive that second approval. Even once that gate opens, any group involving kids would still need two unrelated, background-checked adults present. No meetings would happen in private homes.

The AI-assisted library help desk is designed to keep no record tied to any one person — no profile of what someone asked about, combined counts only. If that promise were ever broken, even by accident, the plan calls for stopping that whole piece immediately for a full review. The design also promises librarians and teachers a bigger, better role throughout all of this — coach and mentor, not competition. It states plainly that any version of this plan that replaces their jobs, instead of adding to what they do, breaks that promise.

There's also a risk that "learn by doing a real project" quietly turns into unpaid work performed for the program's own benefit. The design tries to block this by requiring participants to choose their own projects. It also treats it as a stop sign if people later say they felt used rather than taught.

One more caution comes straight from the evidence review, not from critics of the plan. Giving people open, unguided access to an AI tool can make them worse at a task once the AI isn't there to help. One study found exactly that backfire: practice scores rose 48%, but unaided exam performance actually fell about 17%. A more carefully designed version of the same tool offered hints instead of answers. It avoided the harm entirely, and boosted practice scores even more. That is the specific evidence behind the plan's insistence on "guided," not just "AI-assisted."

A popular claim also deserves a flag here. A famous 1980s finding claimed one-on-one tutoring produces a giant jump in learning, and it still shows up in vendor pitches today. Later, more careful research was never able to reproduce anything close to that original number. The real, replicated effect is still large, but well under half that size. Any pitch for an AI tutor that leans on the bigger, original number is repeating a figure that didn't survive follow-up testing. The same caution applies to AI-tutoring products already marketed to schools. No independently published study has actually tested most of them yet. Claims about how well they work are still unproven.

One more honest caution is about timing. Early ambitions for this program talk about becoming a world-leading, "most future-literate" population quickly. The evidence review is blunt: for foundational reading and math skills, change like that on a timescale of months has never been shown to happen anywhere. National test averages move only a few points per decade. Adult learners typically need about 100 hours of instruction just to move up one skill level.

Two historical literacy campaigns did report fast, large gains in months: Cuba in the early 1960s and Nicaragua in 1980. But both used a far lower bar for "literate" than modern tests do. Both relied on numbers reported by the same governments running the campaigns. Both mixed the teaching with political messaging. A later study found many Nicaraguan graduates had slid back into illiteracy within a decade. Fast change looks more realistic in newer skills like AI use. Most people start from near zero there, and the floor for basic competence is genuinely low. That is not true for the deep, foundational literacy that national tests measure. None of this is a story of crisis: Canada already starts from a strong position. The honest case for this program is opportunity, not rescue.

Underneath everything above: nothing here has been approved. The Learning City is still an idea inside a program whose own charter hasn't been signed off yet. No library has partnered, no circle has met, and no book has been picked.

Receipts

Everything above comes from three documents, all still in draft form:

nine governing rules, and the research questions it's organized around. Still marked PROPOSED, awaiting the operator's own ratification.

the Learning City, including other approaches considered and set aside for each piece, and the plan's own honest argument against the whole idea.

findings on literacy, learning methods, AI tutoring, and libraries. Each one is flagged for how solid or contested the underlying evidence actually is.

Every number and claim in this explainer traces back to one of those three files. If a figure above doesn't have a citable source in those files, it isn't in this explainer either.

There's no circle to join and no book to vote on yet; none of this has launched. The one honest action available right now is to watch for two things: whether this design survives its own internal for-and-against review, and whether Toronto Public Library actually agrees to partner. By the plan's own rule, nothing above moves forward without that yes.