Zhijun He  /  Research program

How governing gets negotiated

One question, one family of concepts, and a sequence of cases chosen so that each one puts the framework somewhere it might fail.

the ethnographies it grew from: Ingelaere, Rettig, ThomsonHow governing gets negotiatedthe one questionConvergent impossibilitychapter 2Paradox-breaking mechanismschapter 4Negotiation-based capacitychapter 4Constitutive ambiguityubwengeBidirectionalitywho gets whatSubstitution trapchapter 5Attributive ambiguityHimalaya, North AfricaUnchecked rulesBrazilThe model abroadthe doctoral projectThe gacaca archivewhere the time goes now
The program as a tree. One question is the trunk. The thesis concepts grow up the left limb in the order the chapters build them; the tests away from Rwanda grow on the right; the centre bears this year’s work, the gacaca archive, as the one red bean. The roots are other people’s fieldwork, which is where the fieldwork of my own is meant to take over.
Convergent impossibility no sequential entry point Paradox-breaking mechanisms simultaneous, substitutive, self-reinforcing Negotiation-based capacity governing inside underspecified rules Substitution trap success consumes its successor Constitutive ambiguity the gray zone is the medium, not the residue Bidirectionality condition does the relation deliver to both sides? yes: governance no: predation next: the gacaca archive Beyond Rwanda attributive ambiguity: Himalaya, North Africa unchecked rules: Brazil's firm registry the model exported: Rwanda Cooperation Initiative
Solid outlines: the chain built in the Rwanda thesis. Dashed, shaded: the test that separates governance from predation, and the cases chosen to put the framework where it has no home advantage.

The question

The canonical answer to "where does state capacity come from" runs through prerequisites: a fiscal base, a professional bureaucracy, rules applied consistently to cases defined in advance. In most of the places I study, those prerequisites never existed in deployable form; where they did, they often arrived as instruments of extraction and were inherited as such. Yet several of these places are governed, and governed through relationships the canonical account does not see: between rulers and ruled, but also between states and the interveners, donors, and neighbors entangled in their politics. My program tries to describe how that governing is negotiated, what it costs, and who bears the cost, without defending it or pretending that the canonical route was ever on offer.

1. Convergent impossibility

Four literatures each describe a genesis problem: you need resources to build capacity and capacity to raise resources; legitimacy to extract and extraction to earn legitimacy; institutions to constrain leaders and leaders to build institutions; trust to make peace and peace to grow trust. Read together, these are not four correlated risks but one structure, in which the crises constitute each other and no dimension can be addressed first. I call this convergent impossibility. It explains a specific failure pattern: well-designed sequential interventions that fail in the same places, repeatedly, because the problem is simultaneous.

development theory Resources Capacity state formation Extraction Legitimacy leadership Institutions Leaders post-conflict Trust Peace No entry point each loop waits on another revenue funding restraint protection
Four literatures, four circular dependencies (solid arrows). Read one at a time, each is a chicken-and-egg problem with a known workaround. Read together (dashed arrows), the loops chain into one: revenue builds capacity, resources fund institutions, restrained leaders earn trust, protection makes a state rightful. Every way in passes through a loop that is itself waiting, so there is no crisis to address first.

Tested in: After the Catastrophe (2026), chapter 2, with Somalia as the case where the structure was not broken and the eastern Congo as the case where sequential programs keep meeting it.

2. Paradox-breaking mechanisms

Where convergent impossibility is broken, it is broken by institutions that generate capacity through their own operation rather than requiring it in advance. The Rwandan cases share three features: they act on several paradoxes at once, they put what was at hand in the place of what was not on offer (community knowledge for legal expertise, mandatory labor for a tax base, cultural idiom for bureaucratic audit), and they reinforce themselves, each round of operation producing the capacity that allows the next round to expand.

NOT ON OFFER PUT IN ITS PLACE MECHANISM LOOPS ACTED ON capacity legitimacy leaders peace Legal expertise Community knowledge gacaca Tax base Mandatory labor umuganda Bureaucratic audit Shame-based accountability imihigo Civic community Camp-based incorporation ingando self-reinforcing: each round builds the capacity for the next simultaneous: several loops at once
Each mechanism put what was at hand in the place of what was not on offer, acted on several of Figure 1's loops at once, and fed itself through its own operation; in gacaca, most confessors gave repeated statements, each naming others and opening new cases. The loops each mechanism acts on are my reading of the thesis's chapter 4, not a count the chapter itself reports.

Tested in: After the Catastrophe, chapter 4, on gacaca, umuganda, imihigo, and ingando as one interlocking system rather than four programs.

3. Negotiation-based capacity

What these mechanisms produce I call negotiation-based capacity: a mode of governing different in kind from the rule-based one. The weight of the term falls on negotiation, on a set of power relations more than a stock of administrative ability. Its authority is performative and earned by results. Compliance is bargained for: citizens read ambiguous signals and calibrate how much to give. Its information structure is asymmetric and recursive, and it is assembled by disjointed increments instead of being specified in advance. Susan Thomson's interlocutors in southern Rwanda rose before dawn to work their own fields ahead of any summons to communal labor, pleaded illness, or reached for gendered excuses. One released prisoner described repairing a bridge with other men, an act of self-help that could be read, if seen by the wrong official, as men of the wrong category plotting. In Thomson's reading, the state claimed credit for exactly the kind of communal labor it made suspect. The concept lets me treat as an institutional form something the dominant paradigm can record only as a deficit.

RULE-BASED Rule, specified in advance case case case Same treatment, every case NEGOTIATION-BASED State rules left deliberately open (the gray zone) Citizens ambiguous signal calibrated compliance reads, adjusts thicker arrow: more is learned upward than down DIMENSION RULE-BASED NEGOTIATION-BASED Authorityprocedural: follows from processperformative: earned by results Compliancecommandednegotiated, calibrated case by case Informationpublished, symmetricasymmetric and recursive Constructionspecified in advancebuilt by disjointed increments
Left, the canonical account: a rule fixed in advance, applied the same way to every case. Right, the mode the thesis describes: state and citizens meet inside rules left deliberately open. The state signals ambiguously; citizens read the signal and decide how much to give; the state reads that in turn and adjusts. What passes between them is power as much as information, and it does not pass evenly: the state learns more of its citizens than they learn of its intentions.

Tested in: After the Catastrophe, chapter 4. The comparative claim, that other developmental states exhibit the same mode, is a program for future work, not a finding.

4. Constitutive ambiguity

The single finding that organizes the thesis: in the gacaca studies of Max Rettig and Bert Ingelaere, over seventy percent of gacaca participants said people lied in proceedings, while over ninety percent said gacaca would bring peace. Read as contradiction, this is a failed institution. Read as signature, it is how the institution worked. Lies compelled counter-narratives that produced information; silence revealed what communities knew but would not say; performed compliance kept the framework standing inside which real accountability, however partial, could happen. The ambiguity was not a residue to be cleaned up by better design. It was the medium.

Ingelaere gives the practice its Kinyarwanda name: ubwenge, a word that holds intelligence, prudence, and strategic dissimulation at once, and that judges a statement by what it does rather than by whether it matches the facts. A defendant told him: "The 'truth' between Rwandans is something that is not close. It is far." The two terms name different things. Ubwenge is the skill people bring into the institution; constitutive ambiguity is the space in which that skill does its work, and neither amounts to much without the other.

Tested in: After the Catastrophe, chapter 4, section by section across the four mechanisms; the thesis's original analytical contribution.

5. The bidirectionality condition

If ambiguity is the medium of governance, what separates it from the medium of predation? The thesis locates the difference in the direction of the exchange. Rwanda's gray zones extracted and delivered: labor for infrastructure, confession for release, performed allegiance for civic standing. Under Barre, Somalia's increasingly ran one way. Ambiguity sustains governance when the gray zone delivers to both parties and becomes predation when it delivers to one. What it delivers is rarely whole, as a survivor in Rettig's Sovu study made clear: "There is a difference between peace and security. Today we have security, not peace." This is the framework's hardest test and its most portable question: for any institution operating through strategic underspecification, ask who gets what out of the exchange, and who is left out of it.

Tested in: After the Catastrophe, chapter 4, through the Rwanda–Somalia comparison; the gendered version of the question is stated there as open.

6. The substitution trap

Each mechanism succeeded by substituting, and each substitution consumed the space its replacement would need. Community courts occupied the ground where an independent judiciary would have grown. Mandatory labor occupied the ground of voluntary civic association. Personalized, upward accountability occupied the ground of procedural audit. The claim is about structure and makes no forecast of collapse: the more completely a paradox-breaking mechanism works, the harder it becomes to leave. I take this to be the thesis's most consequential move, because it turns the sustainability question from a matter of opinion about a regime into a matter of institutional sequence.

Tested in: After the Catastrophe, chapter 5, alongside two further tensions: the generational erosion of emergency legitimacy, and the dependence of institutional maturation on the conditions that also obstruct it.

7. Beyond Rwanda: attributive ambiguity

A framework that has only been tried on its home case has not yet been tested. The first test away from home is disaster governance. When a flood or a dam failure can be read either as a climatic extreme or as a governance failure, the reading is decided after the fact, and states know this in advance. Anticipated deniability changes what they build: visible response capacity grows, because it is credited; quiet prevention lags, because it is not. I call this attributive ambiguity, and the distortion it produces "deformed specialization." It belongs to the same family as constitutive ambiguity, but the actor is now anticipating an attribution rather than negotiating a gray zone.

Tested in: "Attributive Ambiguity: How Anticipated Deniability Deforms State Disaster Capacity," with Luoxianer Peng, accepted for the AAG Annual Meeting, New York, February 2027. Cases: Chamoli (2021) and South Lhonak (2023) in the Himalaya; Derna (2023) as a scope-condition probe.

8. Beyond Rwanda: unchecked rules

The second test is a rule that exists on paper and is never verified. Brazil's EIRELI legal form required a minimum capital that no registry checked. A rule-based account predicts it does nothing. Using 69.1 million firm registrations, I ask who nonetheless failed the filter, and why, across four countries, a floor this large survived only where no registry looked. It bound anyway, through the accountants who file the paperwork. The premise is that an unverified rule still allocates; it allocates through the same underspecification that, in Rwanda, was the medium of governance.

Tested in: "Who Fails the Filter? Size, Verification and the Incidence of an Unchecked Minimum Capital Requirement," solo-authored; SSRN working paper, 2026, submitted to NEUDC 2026.

9. Rwanda beyond Rwanda: the model abroad

The third test turns the thesis around. Its final chapter argued that Rwanda's post-genocide mechanisms could not be transferred; at most, analytical lessons could travel. Rwanda now sends abroad both soldiers, as peacekeepers and on bilateral deployments, and systems. On the civilian side, since 2018 the state has run an agency, the Rwanda Cooperation Initiative, whose business is being learned from: it hosts delegations from across the continent and co-designs systems with other governments. The framework is most exposed at the point of transfer, in what travels and in what changes on arrival. Bidirectionality is no longer a relation between a state and its citizens, but among a state, the states that learn from it, and their citizens.

I read this through the colonial relation, used as a lens rather than a verdict. David Chioni Moore proposed setting aside the question of whether a place is postcolonial and asking instead what a postcolonial reading lets us see there. Here it brings into view two moves he traces across postcolonial societies: the search for authentic origins, and the adoption of dominant forms. Home Grown Solutions are presented as a return to precolonial practice; at least one of the systems Rwanda has exported is built to align with World Bank standards. The same agency offers both authenticity and conformity, and the project asks how each is received. Moore's further point, that the direction of prestige need not follow the direction of power, suits a small state that other, larger states now come to learn from.

A study visit is also an act of reading. Moore notes that Kapuściński's account of the fall of Ethiopia's emperor was read in Poland as a story about Poland; delegations to Kigali may likewise be reading Rwanda as a story about themselves. The questions then become relational: who sets the terms of an exchange, what claims to sovereignty it enacts, what standing it confers on the state being learned from, and how that standing repositions it, abroad and at home.

Planned: the doctoral project. Kinyarwanda study under way; preliminary work in Kigali from 2027. Frame drawn from David Chioni Moore, "Is the Post- in Postcolonial the Post- in Post-Soviet? Toward a Global Postcolonial Critique," PMLA 116, no. 1 (2001): 111–28.

What comes next

The thesis closed on a record. The gacaca courts left behind the most granular map of local social relations any African state has produced in the postcolonial period, and the program to digitize and index it reaches its declared completion in 2026. My current project, developed in the MADE graduate cohort of Yale's Genocide Studies Program with David Simon, reads that archive as an object rather than a source: how it has been classified, indexed, counted, and opened to whom, and what that work has asked of the people who carry it out. The framework's question returns in a new form. Gacaca's testimony worked because it could stay ambiguous; indexing assigns every file a name and a hill. What happens to a record of negotiation when it is made legible, and who gets to read it? This is the move from an institutional to an ethnographic reading of the same material, and from other people's fieldwork to my own.

The archive and the model abroad are one line of inquiry followed to its two ends: the record Rwanda keeps of how it negotiated its recovery, and the record it shows to others. They meet at a claim made at the Security Council in June 2026, when Rwanda's justice minister, pressing for custody of the ICTR archives, argued that the question was no longer one of capacity, because the infrastructure and digital expertise had already been built. The capacity the archive project traces at home is the capacity the Rwanda Cooperation Initiative offers abroad. The archive is this year's work and the spine of my MA; the model abroad is the doctoral project it leads to.

One limit of this page needs saying. Its concepts come first and its people second: the scenes above are drawn from other scholars' fieldwork to show what a concept means. That order follows from the page being a synthesis; it is not a method I intend to keep. The fieldwork is meant to reverse it, beginning with what officials, visitors, and citizens in Kigali and in the capitals that learn from it say they are doing, and asking the concepts to answer to them rather than the other way round. I expect some of the concepts not to survive it.

Thesis: After the Catastrophe: Rwanda's Transformation, Macalester College, 2026, advised by Ahmed Samatar, David Chioni Moore, and Kathryn Linn Geurts. MA advisor at Yale: Louisa Lombard. The voices on this page come from the ethnographies the thesis was built on, chiefly those of Bert Ingelaere, Max Rettig, and Susan Thomson. Last updated October 2026.