The first contribution, and the one the others build on. It moves the unit of analysis from the parts and their attributes to the ties between them: who trusts whom, who talks to whom, and how hard one part’s move lands on another. It explains why the same plan works in one school and fails in the next.
What relationships are we not seeing, and which could change the result?
Most analysis is attribute-based. It profiles actors, scores departments and ranks risks one at a time. Relations before things treats the pattern of interaction as primary. Trust, reciprocity, information flow and the tightness of coupling become variables in their own right, not soft context. The lineage runs from Anderson’s “more is different” through Emirbayer’s relational sociology, Fei Xiaotong’s ripples of association and ubuntu’s relational personhood to Bateson’s warm data. Whyte’s relational tipping points show the stakes. Once trust and accountability degrade past a threshold, coordinated action fails however capable the parties are. The contribution is strongest where an intervention works in one place and not another. It is weakest when it turns into the claim that everything is connected.
Outside school
Two primary schools in neighbouring towns adopt the same reading programme in the same year. Same materials, same training days, same visiting consultant. In one, reading improves within two years. In the other, the programme is quietly dropped by the spring. An evaluation compares them line by line and finds nothing to choose between them on budget, staffing or how faithfully the programme was run on paper.
The difference was in the staffroom. In the first school the pilot went to the two teachers everyone already asked for advice, and it spread the way their advice always had. In the second it went to a new appointee the principal trusted and the staff did not yet know. Same ingredients. Different arrangement. Opposite result.
That is the shift this contribution asks for. Anderson put the general point in three words: more is different. Schelling showed it in a social setting, where mild individual preferences for similar neighbours add up to stark segregation that no one intends. You cannot read the outcome back from what each person wanted. Anyone who has drawn up a seating plan for a wedding knows the domestic version. The guest list fixes who is coming. The table plan decides what kind of evening it will be.
Whyte adds the warning. Writing about Indigenous peoples and climate change, he argued that trust, consent, accountability and reciprocity have tipping points of their own. Past them, coordinated response is no longer possible in time, however capable everyone is. Colonialism had eroded those qualities long before the climate crisis arrived. The lesson travels. A relationship is a stock that can run down, and a plan that spends it does not show the cost until the next plan needs it.
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, reading the plan again
Two years ago a triple-marking policy drove two teachers out of English, and the department that remains has not forgotten it. Mrs Lim knows “every piece of work” will be heard as the same request in new words, sent from the same corridor. None of that appears in the plan, because the plan has nowhere to put it.
Term 1, Week 4
Aisyah, in the corridor
By Wednesday the whole of Secondary 2 has compared ratings. Aisyah is Developing in science and Secure in English, and her friends have already decided science is “harsh”. The verdict reached her through three friendship groups in two days. The school’s letter explaining the scale took a week and was read by almost nobody.
Term 1, Week 5
Mr Ravi, at lunch
He tries whole-class feedback in Secondary 2 science, outside the policy, and his marking time halves. He does not tell Mrs Lim. He tells the two colleagues he has eaten lunch with for nine years, and by Friday the geography teacher is trying it too. The idea travels along the oldest ties in the building and nowhere near the plan.
Term 1, Week 6
Mdm Rohani, at 3am on a break
The Secondary 2 parents’ chat group has decided that without weighted assessments nobody knows where their children stand, and three families have signed up for tuition. Most of the two hundred messages come from eight or nine people, and no teacher is in the chat. Mdm Rohani notices the same few names online at that hour, all shift workers like her, and messages one. By Sunday four of them have written to the school together, the first time any of them has written to it at all.
Term 1, Week 8
Mrs Lim, after the book check
The first book check rates humanities “inconsistent”, and the heads of department read it as blame from her office. Within a week teachers are writing comments for whoever checks the books, not for the student. Mrs Lim meant the check as support. It travelled along a tie that has carried suspicion since the triple marking, and it was read the way that tie reads everything.
Term 1, Week 10
Mrs Lim and Mr Ravi, the staff meeting
Staff agree three probes, and English agrees only to the quiz. Annotation fills the books that get looked at, and whole-class feedback leaves them emptier than the format check allows. The quiz is never recorded, so nothing from it can travel up to Mrs Lim’s desk. She sees, late, that the vote was about where the evidence would go.
Term 2, Week 2
Ms Tay, at her spreadsheet
Two of Ms Tay’s schools, two kilometres apart, brought in the same ratings framework in the same year with the same training. One reports rising compliance and a staffroom close to revolt. The other reports less and seems calm. Every column in her spreadsheet says they should match, and none of the columns describes how either school gets on with itself.
Term 2, Week 4
Aisyah and Mdm Rohani, the evening after the checkpoint
Aisyah, strong in class all semester, freezes in the timed English paper and is rated Emerging. Mdm Rohani complains through the app, because she knows nobody in the English department to ring. The complaint lands in a queue, not with a person. Somewhere in the building a teacher quietly rates borderline students up, to keep the next one from arriving.
Term 2, Week 5
Ms Tay and Mrs Lim, on the phone
The dashboard shows 96 per cent of books compliant, and the cluster newsletter cites the school as an exemplar of “feedback culture”. The same week Mrs Lim rings Ms Tay, as she has done for years, and tells her nobody in the school trusts the number. The official channel carries the figure. The old phone line carries what the figure is worth, and only one of them reaches the newsletter.
Term 2, Week 8
Mdm Rohani and Mr Ravi, Saturday morning
The Meet-the-Parents session has a Saturday morning slot because of the shift workers’ letter. Mdm Rohani comes for the first time in two years, and Mr Ravi shows her Aisyah’s quiz record, fortnight by fortnight. She sees learning that the Emerging did not show. It reaches her through a tie that did not exist at the start of the semester, from a teacher Aisyah had already told her to trust.
Term 2, Week 10
Mr Ravi and Mrs Lim, the last staff meeting
Whole-class feedback has spread to two more departments by word of mouth, changed on the way. The quizzes continue, still unrecorded, and Mrs Lim, like every leader, keeps asking for the data. Mr Ravi keeps saying no, on English’s behalf as much as his own. The department agreed to the quiz because nothing from it travels upward, and recording it would end the one agreement of the semester.
Every figure on the dashboard described a department or a student on its own, and what decided the semester sat between them: a grudge from the triple marking, a parents’ chat no teacher was in, a corridor that sorted subjects into harsh and kind, a lunch table that carried a better idea. Reading the ties showed that English’s refusals were about where evidence would go, and that trust in Mrs Lim’s office, more than the policy, set what the school would try. It did not mend that trust, which moved one phone call and one Saturday at a time. Ms Tay’s proposal that the cluster stop collecting twice-a-term ratings was still only a proposal at the end of the semester, and the promised guidance on reporting had still not arrived.
Implications
The same five dimensions for all seven contributions. Read them across on the compare page.
From treating components and their attributes as the unit of analysis to treating patterns of interaction as primary. Attribute-based analysis profiles each actor and adds up the profiles. Relational analysis asks how the same actors behave in a different arrangement, and treats trust, reciprocity, information flow and the tightness of coupling as variables with values of their own.
The lineage is wide. Anderson’s “more is different” made the point in physics. Emirbayer argued that sociology should take relations, not individuals or structures, as its unit. Fei Xiaotong described a society organised as ripples of relationship spreading out from each person, a network topology set down decades before network science. Ubuntu treats a person as an ongoing relational achievement. Bateson’s warm data names the information that lives in the ties. They agree on the unit and on little else, and their resemblance to each other proves none of them.
Most of the neighbours start from the parts. Analytical thinking explains wholes by their parts, and misses the interaction effects. Critical thinking weighs one claim at a time. Strategic thinking profiles rivals and resources. Even systems thinking, the nearest neighbour, usually draws a fixed set of nodes and asks about the arrows between them. It rarely asks whether the nodes would behave differently if the arrows were rearranged, or whether an arrow can wear thin.
The one neighbour that does ask, warm data and relational inquiry, assumes the relations can be reached through conversation, and struggles with scale and with testing. Complexity thinking keeps its question and adds a discipline: name the ties that matter, judge how tightly they couple, and check whether a change in them would change the result.
An explanation for the result that otherwise looks like bad luck. When an intervention works in one setting and fails in another with the same ingredients, attribute-based analysis has nothing left to compare. Relational analysis does: who ran it, who trusted them, who talked to whom, and how tightly the parts were coupled. Schelling’s lesson cuts both ways. Outcomes cannot be read back to intentions, and an intervention aimed at intentions alone may not move the outcome.
It also puts relational assets and liabilities on the balance sheet. Two departments with the same book-check rating can hold opposite stocks of trust, and the dashboard will show them the same. Whyte’s relational tipping points give the stakes. Past a threshold of lost trust, consent and accountability, coordinated response becomes impossible however capable the parties are. That makes trust something to monitor, not something to mention.
To map ties as well as roles. A relational stakeholder map draws the lines between people, not just their names in boxes, and marks which lines carry trust and which carry history. That means asking questions organisations find awkward. Who would this department take a request from? Who do students tell the truth to? Which parents talk to each other, and which talk to no one? The answers are rarely in any system the school owns.
And to treat trust between the ministry, schools, teachers, students and families as the variable that decides whether a reform can take hold at all. It can tip, so it needs watching and investing in as core work, before the plan and not after it has failed. In a school that is less exotic than it sounds. It often means sequencing changes so that the next request does not land on last year’s grievance.
Romantic holism. “Everything is connected” becomes the finding, so nothing can be prioritised and nobody can be held to anything. The guardrail is Simon’s near-decomposability. Most things are weakly connected most of the time, so the work is to find the few strong couplings and act on them.
The second hollow form is warm-data vagueness: rich conversation about relationships with no traceable effect on what anyone does. The fix is to pair warm data with cold data and with probes, and to record what changed as a result. A third is quieter. A relationship, once noticed, is formalised into a role with a title and a target, and the formalising changes the relationship.
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.
Social network analysis
StrongMaps who is tied to whom and how strongly, and finds the hubs, brokers and clusters. Well established. Its weakness in a school is the data, because the ties that matter most are rarely written down.
Relational stakeholder mapping
ModerateMaps the ties among stakeholders instead of listing them with their interests and influence. Marks levels of trust, and who is missing from the map.
Trust and coupling audits
ModerateAsks how much trust each key relationship holds and how hard one part’s move lands on another. Most useful before a change that depends on goodwill.
Warm Data Labs
EmergingBrings people into conversation about a shared issue across contexts such as family, health, work and schooling, and avoids problem-solving so that interdependencies can surface. The evidence is mostly practitioner reports, so pair it with cold data and probes.
Narrative sensemaking
ModerateCollects short accounts from people inside the situation and reads them for the relationships that metrics strip out. Self-signified micro-narratives are the route to doing it at scale.
Reading
Start with the first on the list.
Anderson, P. W. (1972). More is different. Science, 177(4047), 393–396.
The argument that a whole made of the same parts behaves differently once the parts interact, made from inside physics.
Emirbayer, M. (1997). Manifesto for a relational sociology. American Journal of Sociology, 103(2), 281–317.
The case for treating relations, not individuals or structures, as the primary unit of social analysis.
Fei, X. (1992). From the soil: The foundations of Chinese society (G. G. Hamilton & W. Zheng, Trans.). University of California Press. (Original work published 1947)
Society as ripples of relationship around each person, and how that shape governs trust, obligation and the spread of influence.
Schelling, T. C. (1978). Micromotives and macrobehavior. W. W. Norton.
The canonical demonstration that aggregate social outcomes cannot be read back to individual intentions.
Whyte, K. (2020). Too late for indigenous climate justice: Ecological and relational tipping points. WIREs Climate Change, 11(1), e603.
Relational tipping points: trust, consent and accountability as qualities that can degrade past the point of coordinated response.
International Bateson Institute. (n.d.). Warm Data. https://www.batesoninstitute.org/warm-data
The definition of warm data, and the Warm Data Lab method for perceiving relationships across contexts.