The settlement spirals
Every experienced conflict practitioner knows this moment. The agreement is signed. The terms are reasonable, perhaps even generous. The parties shake hands, the minutes are circulated, and the implementation plan is on schedule. Six months later the conflict is back. It is rarely back over the same issue. It returns over something else: a seating arrangement, a procedural slight, an appointment, a word used in a meeting or on social media. And it returns with an intensity that the new issue cannot possibly explain, often with additional points of friction.
We have a vocabulary for this moment, and most of it is a vocabulary of blame. We speak of bad faith, of spoilers, of agreement fatigue, of parties who were never serious. Sometimes those explanations are true. But they are offered so often, across so many different settings, that they have begun to look less like diagnoses and more like alibis. When the same failure appears in a wage negotiation in Gauteng, a national coalition government, a family business succession and a post-war peace accord, the problem is unlikely to be the character of the parties. It is more likely to be the design of the intervention.
This article, the first in the series Running on Identity, makes a simple and uncomfortable claim: we keep fixing the wrong layer. Our methods are built to resolve disputes. But disputes are only the visible surface of conflict. Beneath them runs a system that determines what the dispute means, what a settlement can achieve, and whether peace will hold. That system is identity. It is not one type of conflict among others. It is the operating system on which modern conflict runs.
This is the essence of modern conflict management. It is not identity politics, it is the very heart of the solutions we need, and the causes for our ongoing conflicts. South Africa is a unique conflict laboratory for these problems and solutions – if we are prepared to do the work.
What the conflict field already knows, and the error it keeps making
Modern conflict practice was built on an architecture of issues. Its most influential expression, Fisher and Ury's Getting to Yes (1981), taught a generation to separate the people from the problem and to look behind positions for interests. As a method for resolving disputes, it remains brilliant. But it rests on a quiet assumption: that the parties are stable, that their identities form the background against which the problem is solved, and that the problem itself is the real object of the negotiation. We speak of “rational actors”, and we build our strategies on that assumption.
The field has never been naive about identity. John Burton argued that conflicts driven by non-negotiable human needs, such as identity, security and recognition, cannot be settled by bargaining over interests. Edward Azar showed that protracted social conflicts are rooted in the denial of identity needs to communal groups, usually by the state. Herbert Kelman spent decades on the way national identities make mutual acknowledgement so costly. One of my favourite inspirations, John Paul Lederach, went furthest of all, telling his students that all conflicts are identity conflicts.
So why another argument about identity? Because the field has acknowledged identity without operationalising it. In practice, identity has been absorbed as a category. Analysts ask whether a conflict is about resources, interests, values or identity, as if these were alternatives on a menu. An "identity conflict" becomes a special case: an ethnic war, a religious dispute, something intractable and exotic, to be handled by specialists. Everything else, the labour dispute, the boardroom fight, the coalition breakdown, is treated as an ordinary dispute about ordinary things. Seating arrangements and votes become the tools we use to solve the problem. That is a category error. Identity is not a type of conflict. It is the layer on which every type of conflict runs.
The shift this series proposes is from typology to architecture: from asking what kind of conflict we face to asking at which layer it is really being fought. And we keep on getting the answers wrong because we keep on getting the questions wrong.
Identity conflict is not identity politics
A distinction needs to be drawn before going further, because public debate has made the word "identity" almost unusable.
Identity politics is a strategy. It is the deliberate mobilisation of group membership for political ends: to win votes, build movements, claim resources or defend privileges. It can be practised by the left or the right, by majorities or minorities. It is something people do. It is easy to spot, relatively easy to manage.
Identity conflict is a process. It is what happens when a dispute puts a person's or a group's standing, meaning, continuity or worth at stake. It is something that happens to people, often without their noticing. It needs no politics at all. Two finance directors fighting over a reporting line in a merged company may be in a deep identity conflict while neither belongs to any movement, and while both would insist the argument is purely about efficiency.
The two are related. Identity politics often feeds on identity conflict, harvesting the injury and amplifying it for other purposes. That relationship is the subject of the fourth article in this series. But the distinction matters, because a practitioner who sees identity only when it is politicised will miss it in the places where it does most of its work: the boardroom, the shop floor, the council chamber and the family. Some of my earlier work sets out this distinction in greater detail.
Why "operating system" is more than a metaphor
Metaphors in conflict studies are cheap, and most of them decorate rather than explain. This one earns its place only if it does analytical work. It does, because an operating system has three properties that map precisely onto the way identity behaves in conflict. From my days in litigation I do not take kindly to metaphors that do not earn their keep in the trenches.
First, it runs beneath the applications and makes them possible: No one uses an operating system for its own sake. People use it to run programs. In the same way, no one negotiates "about identity". They negotiate about wages, land, posts and policy, mayoral chains and water delivery. But every one of those negotiations is running on something else.
Second, it is invisible until it fails: Users notice the operating system only when it crashes. Parties notice identity only when a settlement collapses, a meeting erupts, or a reasonable offer is rejected with fury. Until then, it is simply the way the world appears.
Third, it determines what the applications can do: A program cannot perform an operation that its operating system does not permit. A settlement cannot deliver a peace that the parties' identity system does not allow. This is the property that matters most for practice, and the one our methods ignore.
The model has five layers, each the subject of an article in this series. These disputes run on layers that the parties are rarely aware of. So for example we will see:
Article 2: The kernel, collective mattering – this includes standing, collective worth, existential security, whether the group believes that it matters, and is respected.
Article 3: Protocols, narratives and sacred values – face saving, stories, red-lines, collective trade-offs, what can be exchanged and what not, who can say what.
Article 4: Exploitation – conflict entrepreneurs, algorithms and outrage.
Article 5: System state - stability, lock-in, effective persuasion, what the parties say the disputes are about.
The applications are the disputes we can see. The protocols are the narratives and sacred values that decide what an issue means and what may be exchanged for it. The kernel is the deepest need: whether a group believes it matters. The exploits are the actors and technologies that deliberately attack the lower layers for gain. And the system state is the overall pattern of stability: why some conflict systems lock in and resist every attempt at change.
Four signatures of a kernel-level conflict
If identity runs beneath every conflict, the practical question is not whether it is present but how much of the conflict is being fought at the lower layers. Four signatures indicate that the real conflict is no longer where the parties say it is. Keep an eye on this as you follow these arguments on social media:
None of these signatures proves that a conflict is "about identity" in the old typological sense. Together, they show how far down the stack a conflict has travelled, and therefore at which layer an intervention must operate.
South Africa, November 2026: a live test
South Africa is an unusually clear laboratory for this argument. It lives with visible, deeply unresolved identity fault lines and with a deep culture of negotiated settlement. At the time of writing, it is also running the experiment in real time. The 4 November local government elections left the ANC on just under 31 per cent, without outright control of any of the eight largest metros. The DA rose to almost 25 per cent and topped the vote in several of them, including Johannesburg, Tshwane and Ekurhuleni.
The Government of National Unity, formed in 2024, continues nationally. The coalitions beneath it are being renegotiated council by council. On 13 November, the ANC's extended National Executive Committee reportedly resolved that the DA should be its partner of last resort in municipal coalition talks. At the application layer, this is puzzling. The two parties hold the great majority of the GNU's seats in Parliament, the DA leads in the three Gauteng metros, and the ANC has reportedly acknowledged the instability of governing with many small parties. Yet the arithmetic did not decide the outcome.
The four signatures explain why, and theory becomes practice -
Beneath all four runs the race layer, now discussed openly in the language of "the Black vote". The phrase means different things to different parties, and each position deserves to be stated in its own terms. This could do with some unpacking as we progress through the series.
The DA's position:
After the results, the DA announced that more than a million Black South Africans had voted for it. The party does not present this as a racial headcount in the service of racial politics. It presents it as evidence against racial politics: proof that the assumption that skin colour determines how South Africans vote is breaking down, and that a party for all South Africans can draw support from every community. Its non-racialism is often caricatured as "not seeing colour". The party's own account is more careful. The DA's Zakhele Mbhele explicitly rejects an "ahistorical, naively denialist 'Pollyanna' colour-blindness". On this account, race remains relevant in South Africa, but as a social construct rather than a basis for law. Hence the party's rejection of race-based policy in favour of "merit, not colour", and its argument that redress should reach all disadvantaged South Africans directly, whatever their colour, rather than through racial categories that can entrench patronage.
Its critics and the other positions:
The DA's critics argue that where race still tracks disadvantage so closely, removing racial categories from policy risks leaving historical patterns intact. That is a serious argument, and so is the DA's reply. The ANC's position also deserves its own terms. For a movement that led the struggle against apartheid and has governed in the name of the majority for three decades, redress legislation is part of its historical mission, not a policy preference. To many in its base, partnering with a party that challenges those laws in court looks like a departure from that mission, which is what NEC members reportedly argued.
And the voices calling for "Black parties" to unite are responding to a real fear that the majority's political power is fragmenting. Their own critics answer that a party led by Black people does not automatically serve Black interests. The analytical point is not which of these positions is right. Each is a serious answer to the same question, and that question lives at the kernel layer: which constituency counts, who may speak for it, and whether a group's standing is best secured by naming race or by moving beyond it as a basis for policy.
An operating system does not stop running because one party proposes to retire a category, or because another claims to represent it. What matters for the practitioner is that every one of these positions is held as a matter of principle and identity, not tactics, and must be engaged on those terms. It is not. That is why seniority is the real negotiation. Reports suggest the DA opened by proposing to lead most of the hung metros. For a former liberation movement whose identity rests on representing the majority, serving as junior partner to a party its base still associates with a minority, despite the DA's growing Black support, is not a portfolio question. It is a kernel question. For the DA, being ranked a last resort after topping several metros is a denial of standing of the same kind.
This article therefore makes a specific prediction: metro agreements that allocate posts by arithmetic, without settling seniority and standing in a way both bases can accept, will fail faster than those that do. The councils being formed this month will test it.
The same mechanism, far from politics
In a composite pattern familiar to us South African labour practitioners, a wage dispute is settled close to the union's demand. Within months the workforce is in dispute again, over a supervisor's tone or a disciplinary process. Management sees an insatiable union. The workers experience an offer presented as generosity rather than recognition, and daily treatment that tells them they do not count. The wage was an application. The conflict was in the kernel.
The currency mismatch
Both cases point to the central mechanism of my article. Settlements are denominated in the currency of the application layer: money, posts, clauses, timelines. But the debt that drives a kernel-level conflict is denominated in another currency entirely: standing, significance, the assurance that one counts. Paying a kernel debt in application currency is like repaying a loan in a currency the creditor cannot spend. The amount may be correct. The payment does not clear the account, if we can use that analogy. This explains why settlement failures so often look irrational to the party that paid.
From its ledger, the debt has been settled, sometimes generously. From the other party's ledger, nothing has changed, and the insistence that it has is itself a fresh injury. This creates a secondary storm of polarization and other complicating trust conflicts. It also exposes a more troubling possibility that we can track if we know where to look: a settlement can deepen the conflict it resolves. When the stronger party concedes generously, it may be signaling, without intending to, that the weaker party's gains depend on the stronger party's grace. A gift can confirm a hierarchy as surely as a refusal.
Since Marcel Mauss's work on the gift, anthropologists have recognized that a gift which cannot be returned creates obligation and subordination. In conflict, the generous concession offered from above may leave the recipient materially better off and existentially worse off at the same time. That is not ingratitude. It is the operating system responding exactly as it is designed to. The point is not that material settlements are unimportant. Wages, posts and land matter enormously, and kernel needs are often expressed through them. The point is that the way a material outcome is reached and framed carries a second payment, in the kernel currency. Practitioners who design only the first payment should not be surprised when the second account remains in arrears.
What changes for practice?
The full practical framework belongs to the final article of my series, but four shifts follow directly from the argument so far.
The claim on the table
An argument of this kind is only worth making if it can be wrong, if it is falsifiable. So here is the claim this series will defend, stated in a form that can be tested:
(1) Settlements fail in proportion to the collective mattering they leave unresolved
If the claim is right, we should find that agreements which address standing, through process, framing and symbolic recognition, survive longer than agreements of equal material value that do not. We should find that the four signatures predict recurrence. And we should find that the conflicts we call intractable are not uniquely complex, but simply ones in which the kernel has never been reached. If the claim is wrong, the evidence will show it, and the series will have earned that discovery honestly. The next article descends to the kernel itself. The Need to Matter will argue that what groups ultimately fight for is not recognition, which can be granted while nothing changes, but mattering: the assurance that their presence makes a difference and their absence would be noticed. It will ask what it means, for a party or a constituency, to be named a last resort.
(2) The diagnostic question
Each article in this series ends with one question a practitioner or senior political negotiator can carry into the arena the next morning. This is the first: If this dispute were settled on exactly the terms being demanded, what would still be unresolved? Whatever answers that question is where the conflict really lives. Join us for next week’s article.
References and suggested reading
Sources for the South African section
Some of my own related work
Andre Vlok can be contacted at andre@conflict1.co.za for any further information. (c) Andre Vlok November 2026