The contribution that changes how a school acts once it admits it cannot plan the result. It stops specifying outcomes and the path to them, and shapes the conditions from which good outcomes become likelier. Then it learns by running small, safe-to-fail probes, each with a rule for stopping written down before it starts.
What conditions make good outcomes likelier, and what can we safely try?
Conditions before outcomes moves from blueprint to enabling constraints and probes. Juarrero showed that constraints do not only limit. Some create the coherence from which new order forms. The Cynefin tradition separates governing constraints, which hold behaviour within limits, from enabling constraints, which let patterns form. In practice this means working forwards from present conditions under a declared direction, instead of backwards from a fixed destination. Action becomes a portfolio of small, parallel probes, each with criteria for amplifying or dampening it, agreed before it starts. It is strongest where cause and effect only make sense afterwards. It is weakest when “probe” is a new name for a pilot that nobody intends to stop.
Outside school
Buurtzorg, the Dutch home-care organisation, runs on small self-managing teams of nurses. A team of about a dozen looks after the people in one neighbourhood. It plans its own rotas, takes on its own patients and solves its own problems, with a coach to call on and a small back office behind it.
What makes it interesting is where the rules sit. There are rules for how a team works: how it meets, how it decides, what it does when it disagrees. There are very few about what care should look like for a given patient. That is left to the nurses in the room. In Juarrero’s terms, the organisation has swapped governing constraints on the care for enabling constraints on the team, and the pattern of care forms from those.
A gardener works the same way. Nobody makes a tomato. You set the conditions: soil, water, light, a stake, some spacing. You plant three varieties in different beds, because you do not know which suits this garden. You decide in advance what would make you plant more of one next year, and what would make you pull it up. Then you watch.
The evidence for self-managing organisations is thin and mostly case-based. Lee and Edmondson found real benefits and real limits: such forms are hard to sustain and can create anxiety when roles are fluid. Freeman warned that removing a formal hierarchy can produce an informal one that is less accountable. The point is fit, and what carries over from Buurtzorg is narrower than the slogans about it: shape the conditions, decide in advance what you will watch, and let the result grow.
The term
Every page follows the same Singapore secondary school through the same semester, Term 1 and Term 2. The school is changing how it assesses, both of learning and for it: fewer weighted assessments, more feedback, and ratings that are meant to say something. The five people and the events of the semester stay fixed. Only the contribution reading them changes.
All five, interconnected
The five threads interleaved across the twelve weeks, so the hand-offs between them are visible. Pick a person to follow their term on its own. The people are illustrative, and the same five appear on every contribution’s page.
Before Term 1
Mrs Lim, rereading the plan
Mrs Lim reads her plan again the week before Term 1 begins. Every line names an outcome and a route to it: two checkpoints, a next step on every piece, a five-point scale, a 95 per cent target. None says what the school would do if the route turned out to be wrong. She also knows staff will hear “every piece of work” and remember the triple-marking policy that drove two teachers out of English.
Term 1, Week 5
Mr Ravi, Secondary 2 science
Green-pen responses have become ritual, “I will use more evidence”, and a few students have a chatbot write them. Mr Ravi stops writing comments in Secondary 2 books and starts each lesson from the mistakes most of the class made. It is outside the policy, with no criteria and no date to review it. His marking time halves.
Term 1, Week 6
Mdm Rohani, at 3am on a break
Mdm Rohani reads the class parents’ chat group on a break at 3am. It has decided that without weighted assessments “nobody knows where our children stand”, and some families have signed up for tuition centres that set practice papers. Aisyah has no tuition. The school removed one condition, the term-by-term weighted assessment, and families rebuilt it at home for whoever could pay.
Term 1, Week 8
Mrs Lim, after the first book check
The first book check rates humanities “inconsistent”. Heads of department read it as blame, and teachers start writing feedback for the book check rather than the student. Comments get longer, and one department loses a weekend to them. Mrs Lim sees that the check she added to support the policy has become the strongest condition in the building, stronger than either training session.
Term 1, Week 10
Mrs Lim and Mr Ravi, at the staff meeting
Someone calls the three proposals pilots. Mrs Lim corrects it: a pilot is run to show that a thing works, a probe is run to find out, so the rule for stopping goes in before it starts. Mr Ravi’s whole-class feedback becomes the first probe, in science and geography. It will be amplified if the common mistakes fall from one lesson to the next, and dampened if students can no longer say what their own next step is.
Term 1, Week 10
Mrs Lim, writing up the meeting
The fortnightly ten-minute quiz will be amplified if it finds something the checkpoint would not, and dampened if students start revising for it as a test. The annotation probe will be amplified if students name a specific change and a reason, and dampened if their annotations start to read like green pen. English agrees only to the quiz. Mrs Lim notices that none of the three mentions the 95 per cent target.
Term 2, Week 2
Aisyah, in the first quiz
Aisyah likes the quizzes because they do not count, and she answers more freely than she ever has in a test. Most of her class gets the same energy question wrong, and the next lesson starts from it. The checkpoint would not have shown that misconception until the end of the semester, so the quiz meets its amplify criterion in a fortnight. Then one class asks for the quizzes to count, which would turn them back into the thing they replaced.
Term 2, Week 2
Mr Ravi and Mrs Lim, with the allied educator for learning support
The allied educator brings a list. Students still building their English are rated Emerging in nearly everything, because the ratings judge written explanation. Mr Ravi says several of the same students do well in the quizzes, which ask for a choice or a number. The scale was meant to report on learning, and it has been quietly deciding what counts as learning.
Term 2, Week 4
Aisyah, at the first checkpoint
Aisyah freezes in the timed English paper and is rated Emerging, after a semester of strong work in class. Mdm Rohani complains through the app. A teacher quietly rates borderline students up to avoid more complaints, so the scale bends towards whoever reads it. Aisyah’s quiz scores sit only in Mr Ravi’s own notebook, because the probe said they would never be recorded.
Term 2, Week 5
Ms Tay, reading the cluster newsletter
The dashboard shows 96 per cent of books compliant, and the cluster newsletter cites the school as an exemplar of “feedback culture”. Ms Tay has read the school’s notes and knows nobody there trusts the number. The cluster asks every school for the compliance figure, so the cluster is one of the conditions the school works inside. She starts to ask what changing that condition would do.
Term 2, Week 6
Mrs Lim, reviewing the probes
The annotations have started to read like each other, because students learned which sentences teachers liked. That is the dampen criterion, so she and Mr Ravi redesign the probe, with students annotating a partner’s work, instead of dropping it. The quizzes have met every amplify criterion, and school leaders ask for the scores to go on the tracking system. Mrs Lim sees that recording them would remove the one condition that made them work.
Term 2, Week 8
Mdm Rohani and Mr Ravi, on a Saturday morning
After the shift-working parents’ complaints, the Meet-the-Parents session has a Saturday morning slot, and Mdm Rohani comes for the first time in two years. Mr Ravi shows her Aisyah’s quiz record from his notebook. She sees learning that the Emerging rating did not show. The conversation came through a change nobody had filed under the assessment plan.
Term 2, Week 10
Mr Ravi, at the end of the semester
Whole-class feedback spreads to two more departments, changed on the way: maths works from a sample of ten books, and art does it aloud with no written sheet at all. The quizzes continue, still unrecorded, though leaders keep asking for the data. Mr Ravi worries that the probe everyone praises is the one most likely to be formalised. It spread fastest while nobody was managing it.
Term 2, Week 10
Ms Tay, drafting a proposal
Ms Tay proposes that the cluster stop collecting twice-a-term ratings. She writes it the way the school wrote its probes: one year, a date to look again, and a note of what would make her reverse it. The cluster superintendent’s office still wants the compliance figure, and guidance on reporting for lower secondary, promised for this year, has still not arrived.
This contribution let the school see its plan as a set of conditions, and its experiments as ways to learn instead of ways to prove. Each probe met a criterion and still surprised: whole-class feedback changed shape in every department that took it up, the annotations were performed for and then redesigned, and the quizzes worked because they did not count, which is exactly what leaders now want to change. The book check showed that a check added to support a policy can become the strongest condition in the building. Whether the school can amplify anything without recording it is still open.
Implications
The same five dimensions for all seven contributions. Read them across on the compare page.
A blueprint names the outcome and the route to it, then measures the distance. Conditions before outcomes starts from the other end. In a complex situation the result emerges from interaction, so it cannot be specified, and the work moves to shaping what the result grows from. Juarrero’s point is that constraints do not only limit. Some create the coherence from which new order forms. The Cynefin tradition separates governing constraints, which hold behaviour within limits, from enabling constraints, which let patterns form.
Action changes shape too. One plan gives way to a portfolio of small probes run in parallel, each with criteria for amplifying or dampening it, written before it starts. That is the line between a probe and a pilot. A pilot is designed to confirm a solution already chosen, so stopping it counts as failure. A probe is designed to find out, so stopping it on its own criteria counts as the method working.
Strategic and design thinking usually work backwards from a desired outcome. They fix the destination, then plan or prototype the route. Systems thinking finds the leverage in a structure, but assumes the structure holds still long enough to be pushed. Abduction generates the hypothesis and leaves open how to test it. None of them asks what to do when the outcome cannot be specified and the structure shifts in response to being pushed.
Complexity thinking works forwards from present conditions under a declared direction. The phrase to hold on to is direction, not destination. It also reverses the usual reading of uncertainty. Elsewhere uncertainty is a reason to wait for more analysis. Here it is the reason to structure action as learning, because the only reliable way to know what a complex system will do is to let it run.
It lowers the cost of being wrong, shortens the time to learning, and replaces the false precision of detailed plans with real responsiveness. That maps onto a plain robustness test: if the analysis is wrong, is the action reversible, affordable or recoverable, and will we find out quickly? A blueprint usually fails that test, because it learns only at the end.
It also adds one question to any plan under review: are we specifying outcomes we cannot control? Run against the school’s plan, the answer was yes for most lines. A 95 per cent compliance target, a rating twice a term for every student and a next step on every piece of work all name a result. None names the conditions that would make a good result likelier, or what the school would watch to know it was getting one.
Discipline before a probe starts. Each probe needs a direction, a size small enough to fail without harm, and criteria for amplifying and dampening it, agreed in advance and written down. Several run at once, so no single one carries anyone’s reputation. Evaluation becomes developmental in Patton’s sense: a learning function that runs alongside the work, which is also how the school stays accountable for it.
It asks leaders to write fewer and better rules. Plsek and Greenhalgh argued for minimum specifications in health care, the closest analogue to education: specify the few things that must hold, and leave the rest to the people doing the work. The hard part comes later. A school has to honour a dampen criterion when a probe someone cares about meets it, and resist formalising a probe the moment it succeeds.
The relabelled pilot. A project already decided on is called a probe, the criteria are written after the results, and nothing is ever dampened. The second hollow form is the ritual constraint. A rule meant to enable, such as a green-pen response to every comment, is never read back and nothing follows from it. It stops shaping behaviour and becomes a box to tick.
The third is the alibi. “We are only experimenting” becomes a way to avoid committing to anything. The guardrail is the one that holds for complexity thinking at large: commit to a direction, set learning milestones and explicit triggers, and record what changed as a result. Design Unbound, the fullest account of designing for emergence, admits that its tools are illustrated by cases more than tested. A school’s probe portfolio deserves the same candour about its own evidence.
Methods
The repertoire for this contribution, with an evidence rating for each method. Strong means tested in more than one field. Emerging means promising, and mostly case studies.
Safe-to-fail probe portfolios
Moderate to strongSeveral small experiments run in parallel, each with criteria for amplifying and dampening set before it starts. A single probe tends to become a pilot that somebody needs to succeed.
Enabling constraints and simple rules
Moderate to strongA few rules about how people work together, which let a pattern form, in place of detailed rules about what the result must be. They only enable while someone responds to them.
Minimum specifications
Moderate to strongPlsek and Greenhalgh’s alternative to exhaustive protocols in health care. Specify the few things that must hold and leave the rest to the people doing the work.
Missions
Moderate to strongA declared direction, ambitious and specific enough to organise many different projects without dictating which ones. Mazzucato’s version of direction without destination.
Positive deviance
Moderate to strongFind the people already getting good results under the same constraints as everyone else, and spread what they do. Mr Ravi’s whole-class feedback, working outside the policy before anyone asked for it, was a small case of it.
Developmental evaluation
Moderate to strongEvaluation as a learning function that runs alongside the work, not a verdict delivered at the end. It is the accountability frame for a portfolio of probes.
Reading
Start with the first on the list.
Juarrero, A. (2023). Context changes everything: How constraints create coherence. MIT Press.
The argument that constraints create coherence as well as limiting it, which underpins the idea of an enabling constraint.
Snowden, D. J., & Boone, M. E. (2007). A leader’s framework for decision making. Harvard Business Review, 85(11), 68–76.
Probe, sense and respond: parallel safe-to-fail experiments, with what works amplified and what does not dampened.
Plsek, P. E., & Greenhalgh, T. (2001). The challenge of complexity in health care. BMJ, 323(7313), 625–628.
Minimum specifications in place of exhaustive protocols, from the sector closest to education.
Patton, M. Q. (2011). Developmental evaluation: Applying complexity concepts to enhance innovation and use. Guilford Press.
Evaluation as a learning function, and the accountability frame for running probes.
Pascale, R., Sternin, J., & Sternin, M. (2010). The power of positive deviance: How unlikely innovators solve the world’s toughest problems. Harvard Business Press.
How to find and spread what already works under the same constraints as everyone else.
Pendleton-Jullian, A. M., & Brown, J. S. (2018a). Design unbound: Designing for emergence in a white water world: Vol. 1. Tool sets. MIT Press.
The fullest account of what a designer does differently once emergence is taken seriously, candid that it is a framework more than a tested theory.