What we collect, what we never see, and how we share what we learn.
A published methodology, accountable to the educators, students, funders, and researchers who care whether this practice actually does what it claims. Meant to be read, challenged, and improved.
Reweave's promise is one sentence: curiosity, before judgment. We make wordless films and real human stories so children meet a person before anyone tells them what to think about that person. We build AI tools so educators can thread those stories into the math, literacy, and life lessons children are already learning. We hold a reflective practice space so educators can notice what shifted in their classroom and in themselves.
The deeper claim is that curiosity, practiced early, becomes a way of seeing. That habit, at scale, is what we mean by impact.
This page is our answer.
This framework is for the educator first. The journal is theirs. The growth surfaces are theirs. The encrypted notebook is theirs to read and to delete. What we gather as aggregate is what we can learn from with their consent, and what we then return to them as reflection. Reweave is not an extraction engine. It is a practice space that happens to also produce evidence.
Reflective practice is recursive. Each moment of attention feeds the next. The framework is built to catch that loop, not to deliver an outcome.
The arrows are not causal proofs. They are contributions. Reweave is one of many things shaping a classroom. The cycle starts with the educator and ends in the room with the learners they teach. We measure rigorously, and we describe what we are: a contribution to the practice, not a guarantee of outcomes.
Reweave catches reflection at the moment it happens, not in a survey afterward. Eight surfaces, woven into the actual experience of reading stories, watching films, weaving lessons, and teaching.
Maya asked if Norma really fed her sick stepfather one spoonful at a time, sharing every bowl. I did not have an answer. We sat with it together.
Watched Mike alongside my third graders. Three of them noticed his music before his hearing. The order they noticed in matters.
we were not what they thought a school could be. I want to bring this to the math team meeting.
Curiosity and listening showed up clearest. About 22 learners. The quieter ones spoke first this time.
Patience. Three different conversations, all about what happens when I let silence sit a little longer than feels comfortable.
I notice I am quicker to ask what a learner is curious about and slower to tell them what they should be curious about.
They are asking better questions. Of me, of each other, of the stories. I am not the source anymore. They are.
A classroom where the quietest learner feels seen, and the loudest learner does not feel like the center.
Here is the canonical people demo. Three surfaces of noticing show up in the right place at the right time.
Every reflection that lands in the journal carries one or more of these. Tagged by the educator at the moment of capture. Aggregated only across counts, never across content.
Every entry also carries a frame, the mode of attention. The sidebar tabs (i notice, i wonder, i assume), the post-watch self-orientation, the saved line, the after-teaching reflection on the room: each is its own frame. We can see the shape of how educators are paying attention, not only what they paid attention to. The frame field is queryable in aggregate.
Every entry carries a permission value. Defaults are the most protective level appropriate for that kind. Opt-up is gentle, and never the default.
Delete-all on the Yours page removes every row owned by the user. Aggregate queries run at query time, so deletions propagate instantly. There is no derivative aggregate table that would need cascading deletes.
Personal reflection is encrypted in the user's browser before it ever leaves the device. The key is theirs. We hold only the ciphertext, and the math says we cannot reverse it.
When you write a noticing, a moment, or a private journal entry, your browser encrypts it with a key derived from your password. Reweave receives only the ciphertext. We do not have the key, and we cannot recover it.
If you forget your password, your encrypted entries are unrecoverable. That is the cost of an end-to-end-encrypted notebook, and it is the same trade-off a paper notebook in a drawer makes.
Private entries are never read by reweave, never seen by any model, never aggregated, never quoted. Not now, not later, not for any reason.
We do not gather what we cannot return. Everything an educator pours into their notebook becomes visible to them, in shapes that help them notice themselves over time.
3 lessons taught this month. Curiosity showing up most. 14 noticings across 4 people pages.
Real month-by-month counts. Dominant dimension surfaced per month. The fullest month highlighted, gently.
A real constellation built from your noticing entries. Dim trends in plain language: your empathy noticings doubled this season.
The baseline answer you wrote at first journal open, surfaced alongside the annual most-significant-change prompt. A pre-post you can read yourself.
Counts of your entries by kind and permission level. Batch upgrade if you want to share more. Delete-all if you want to share none.
Each card shows where it came from: while reading Norma, after watching Mike, a line from Reginah, monthly pulse in may. The trail is yours to walk back.
"We do not gather what we cannot return."
Free text is hard to learn from. Reweave handles it three ways, building up from what the educator already tells us toward the kind of formal qualitative methodology serious evaluators use.
The educator does the interpretation at the moment of capture. Every taught moment, monthly pulse, quarterly reflection, and annual entry carries explicit dimension chips. Selecting dimensions converts a qualitative experience into structured categorical data without anyone else interpreting the educator's words. This is the workhorse. It powers most aggregate analysis in real time.
For entries an educator has opted up to aggregate or higher, a background worker runs Claude Haiku, the same model already used for PII scrub, to extract themes, surfaced topics, sentiment, and any dimensions the educator did not tag explicitly. The model never reads private entries because they are encrypted at rest and unreadable to anyone but the writer. The tag schema is published and auditable.
A formal qualitative impact methodology developed by Rick Davies and Jess Dart and used by Oxfam, World Vision, and major evaluators around the world. As reweave's community grows and consented entries accumulate, we will gather panels to read through them, identify cross-cutting significant-change themes, and publish thematic reports. The MSC tradition is what would let qualitative claims hold up to a serious evaluator: a documented methodology, applied transparently, with inter-rater reliability where the data supports it.
We are not inventing impact measurement. We are adapting decades of careful work by researchers and evaluators who have wrestled with what it means to measure something hard to measure. A short list of the methodologies and organizations this framework learns from, with more to study as we go.
The gold standard for randomized evaluation of social programs. Their published methodology shapes what we mean by "rigorous evidence" and what we honestly cannot yet provide.
Their "right-fit evidence" approach (methods proportional to the decision the evidence is supposed to inform) is the most direct philosophical fit for this framework. Their embedded learning partnerships are structurally similar to how we treat educators: participants in the work, not subjects of it.
A working playbook for translating research into program design. Their humility about contribution versus causation is the tone we aim for.
Their work on theory of change and pathways to impact shapes how we articulate the loop from a story to a shifted classroom. Their writing on "big bets" philanthropy, patient capital for long-arc change, affirms why reweave measures on a multi-year arc rather than quarterly KPIs.
A rigorous evidence aggregator and systematic-review curator. We watch 3ie's evidence-gap maps to understand where rigorous answers about education exist and where they do not.
A formal qualitative methodology for participatory evaluation. Used by Oxfam, World Vision, and many international NGOs. The MSC panel layer of our methodology is directly inspired by their published user guide.
More to add, as we read more. If we are missing a framework, an organization, or a researcher we should be learning from, please write.
A framework that names its limits is more useful than one that does not.
Reweave is a platform, not a controlled program. Causal claims require partnership with a research organization and a counterfactual. We do not have one yet.
This framework gathers evidence rigorously. Rigorous evidence is necessary, but not sufficient.
Signups, sessions, lessons woven, films watched, all of those are reported. But they are activity, not outcomes. The framework foregrounds outcome and contribution evidence.
This document is versioned and updated as the framework matures. Educators, researchers, and funders are invited to challenge it.
Reweave is early. The collection layers are built. The reflection-back loops are working. The aggregate side, the formal panels, the published reports, those grow as the practice grows. We commit to publishing the methodology, the tag schema, and the permission policy along the way. Educators using reweave are participants, not subjects. The data is theirs first.
The full versioned document lives in the reweave monorepo at IMPACT_FRAMEWORK.md. If you are a researcher, funder, district admin, or educator with a sharp question, a critique, or a methodology we should be learning from, we want to hear it.