A gift, not one of Steffen’s: the beta line goes in the footer, no signature, no counter. No recipient named, so the author-note is addressed to the paper’s authors. No ideas given: every choice below is the engine’s, and nothing in the Given overrides the precedent rule.
Christian Kreuder-Sonnen, Alexander Schmotz and Michael Zürn, “The complex effects of regime complexity: a conflict model”. European Journal of International Relations 32(2) (2026): 611–638, online 2025. DOI: 10.1177/13540661251368630. Open access (CC BY), at journals.sagepub.com/home/ejt. Available: the published PDF, 28 pages, read in full (text, Table 1, Figures 1–6, notes, references). Not available: the online appendix (cluster membership, replication code) and the Interface Conflicts 2.0 Dataset itself (Zürn, Kreuder-Sonnen and Fuß 2024, GESIS, doi 10.7802/2716). Every number below is read from Table 1 and from Figures 4 and 6, whose cells print their counts; nothing is read off the dendrograms.
Primary: an argument. A conflict model of regime complexity (Figures 1 and 2). Where international institutions overlap, their rules can collide; when an actor turns such a collision into a manifest interface conflict (“incompatible positional differences between actors about the prevalence of different international norms or rules”), the conflict is managed somehow, and the management has an effect on the order: integration or fragmentation. The hinge is the conflict’s origin. A non-instrumental interface conflict (NIIC) emerges as a side-effect when a new event or issue meets two institutions with different functional logics (the World Bank’s no-politics charter against the UN human-rights office); its parties also value the public goods the institutions provide, so it is “likely to be handled in a cooperative way”. An instrumental interface conflict (IIC) is created on purpose by an actor who wants to challenge an institution, by pitting another institution’s norm against it (Israel and the US against UNRWA, citing UNHCR’s refugee definition; the African Union against the UN Security Council over Libya); at least one party “intentionally takes a position incompatible” with the other’s, so management is likely non-cooperative, “at least initially”. NIICs resemble coordination games; IICs collaboration games, which “can shift toward cooperation over time”. Cooperative management tends to integrate, non-cooperative to fragment. Formalised here as a toy bargaining model (section 12): the player is one party; the other party’s behaviour is fixed by the clash’s origin; the order effect follows the management with the paper’s own rates.
Secondary: an estimate. Cluster analysis (Gower distance, Ward linkage) of 77 conflicts, 1970–2018, from the ICD 2.0. Table 1: 42 intentional (54.5%) and 35 unintended; 59 managed cooperatively (76.6%); 46 of 76 integrative (60.5%). Simple three-variable solution (Figure 4): cluster 1, “benign”, 35 NIICs, 94% cooperative (33), 74% integrative (25); cluster 2, “hybrid”, 26 IICs, 100% cooperative, 77% integrative (20); cluster 3, “malign”, 16 IICs, 0% cooperative, 6% integrative (1). Note 10: a sub-cluster of 8 NIICs that were cooperative and still fragmenting, and of 6 such IICs. The extended solution (Figure 6) adds object (interests vs values/means), spheres (within vs across spheres of authority) and outcome (symmetric vs asymmetric gains): the three clusters survive; object and outcome do not separate them; spheres does, partly (benign 0% across, hybrid 96%, malign 50%), which the paper leaves as an open puzzle. Reconstructed: the deck of nine clashes reproduces the three clusters in proportion, and the order effect is drawn with the clusters’ rates.
Also handed over: three cases. Malign: UNRWA against UNHCR norms (2017 on: funding cut, mandate renewed by the General Assembly anyway, no agreement). Hybrid: IRENA (2009) against the IEA: initial mistrust, then the IEA upgraded its renewables unit, a partnership (2012), a memorandum (2019); the paper’s explanation is that IRENA’s cause “resonated increasingly well with trans- and international audiences”, so the IEA faced “a legitimacy crisis” if it kept resisting, and the US joined IRENA in 2011. Benign, and messy: investment law against indigenous rights in Zimbabwe (the von Pezold arbitration), mainly non-instrumental with an instrumental streak, mainly cooperative with non-cooperative traits.
Two inconsistencies in the paper, noted for the author-note. (1) The conclusion says that roughly two-thirds of IICs are malign and one-third hybrid; the figures say 16 malign and 26 hybrid of 42, the other way round (p. 621 also says non-cooperative management is “dominant” among IICs, which holds only for the initial phase). The game follows the figures. (2) Table 1 gives 30 conflicts about interests and 52 across spheres; Figure 6’s cells give 52 about interests (23+15+14) and 30 across spheres (0+22+8): the table’s codes look reversed. The game uses neither variable.
A mechanism made intelligible, which settles a quarrel. One camp said overlapping institutions let states play them against each other, so complexity fragments global governance (Drezner; Benvenisti and Downs). The other said overlap brings flexibility and, in the long run, a division of labour (Keohane and Victor; Gehring and Faude). The paper: both are right, about different conflicts. What decides is why a clash began. Clashes that arise by accident are almost always settled together and mostly knit the institutions closer. Clashes started on purpose split: some are fought to the end and tear the order apart; more of them, surprisingly to the paper’s own baseline, start as fights and end in cooperation, and the paper’s best guess from its energy case is that this happens when the challenger’s cause wins over a wider public.
The world has lots of clubs that make rules: one for money, one for refugees, one for energy. Sometimes two clubs’ rules say opposite things about the same thing. Sometimes that happens by accident: something new turns up, and each club answers it its own way. Then the two clubs usually give a little each and end up working together better than before. But sometimes somebody makes two clubs clash on purpose, to knock one of them down. Those fights often end with everyone digging in, and the clubs drift apart. Unless the side that picked the fight has a cause that lots of people cheer for. Then the club that was attacked starts to bend, and the two end up partners.
| Game | Position | Format | Feedback device | Chassis |
|---|---|---|---|---|
| Curie’s Quartz (curie) | experimenter with the paper’s instrument held as a skill | discovery by survey under a budget of panels | the balance (a spot of light held still) and an album that draws itself | ligne claire album, Paris 1898 (Drawn): album white, ink, uranium yellow, clear blue, signal red; Lilita One + Patrick Hand; ink-framed caption boxes |
| Lovelace’s Cards (lovelace) | builder (hangs cards on the Engine) | construction puzzle against par | the loom weaving the pattern | the Engine on pasteboard (Instrument): buff, ink, plum, verdigris, thread red and blue; Abril Fatface + Old Standard TT; punched-card buttons |
| Noether’s Ball (noether, the deck rebuild) | judge of changes (“does it matter?”) | a deck of nine cards, one answer each, then the reveal | two pictures of the trip, the formula plate, a shelf of badges | night-shift machine hall (Techno): gunmetal, steel, neon cyan, neon magenta, hazard yellow; Chakra Petch + Saira + Share Tech Mono; steel lever bars |
What this game does differently.
The aim (start screen; top of the play screen; first line of How to play): Settle as many of the nine clashes as you can. You have 14 pages to give, in all.
The yardstick during play is in the aim’s units: a row of nine marks (a mark turns olive the moment a clash is settled, white with a tear when it ends in a stand-off) and a column of fourteen pages that empties as the player gives them away or the crowd takes them. A settled clash is visibly a success at once; the tear is visibly a failure at once.
The verdict (end screen), three lines composed from what the player did, then the poster of the nine:
No score number: deals are counted on the row of marks during play; the verdict says what happened in words.
A. Who Started It? (agent: one of the two parties; strategic with hidden types and a budget).
B. Seventy-Seven (experimenter: the paper’s analyst; discovery).
C. Push (subject: the player is a rule; real time, no words on the play screen; the reckless one).
A (the pick). 0:00. A poster: two big circles, orange and green, overlapping; the lens between them striped orange and green. Lowercase, very large: who started it? Under it: “The world’s rulebooks overlap. Where two say opposite things, they clash.” The aim on its own line. I tap BEGIN.
0:15. Clash 1 of 9. My circle (orange) is the Development Bank; my rule: no politics in our loans. Theirs (green) is the Rights Office; their rule: people driven off their land by a project have rights. A black dot falls into the lens and lands there: “a dam that will flood villages”. One line: nobody planned this clash; a new problem landed where both rulebooks apply. They say: “Let’s find a way through this together.” Under the picture: Rewrite how many pages of your rulebook to meet them? 0 to 4, and a column of fourteen pages at the right. I touch 1: the border in the lens slides a notch into my circle and one page lifts from my column. I press MAKE THE OFFER. “You rewrote 1 page. They rewrote one of theirs: a deal.” The lens overprints olive and a row of stitches runs across it. Years later: knit, the Bank checks rights before it lends. Under it: “They would have settled for 1 page.” Easy.
1:00. Clash 2. The Council; theirs, the Regional Bloc. No dot: a fat black arrow shoves the green circle into mine. “Started on purpose: a bloc that wants your Council out of its region dug up an old rule of non-interference.” They say: “Your rule stops at our border.” I think: last time one page did it; this one is angrier, so two. “You rewrote 2 pages. They took them and pushed on: no deal.” Then dots gather, under my circle. “The crowd: other governments, the press, people far away. This one is with you: the bloc’s cause wins few friends outside its camp. Holding out costs you nothing.” Last word: I give nothing. Stand-off: the lens tears white. “They would not have settled for any number of pages: they wanted your Council out.” I frown: two pages for nothing. Twelve left.
2:00. Clash 3. The Oil Agency; theirs, the Sun and Wind Agency, and the arrow again: founded on purpose by countries that want the world off oil, to challenge me. “Your rules must give way to ours.” I have learnt: 0. “No deal.” Dots gather, under them, a lot of dots. “This crowd is with them: their cause is popular. Holding out against it has cost you a page already (your own members are drifting to them), and will cost you two more if you end without a deal.” I weigh it: twelve pages, eleven now; bend or hold? I give 2. “They take them, and sign: a deal, after a fight.” The lens overprints. The ghost: “They’d have settled for 2, at the very start too.” So a fight picked on purpose can end in a deal: when the world is with the one who picked it. I grin.
3:00 to 5:00. Clash 4, a dot: one page, a deal. Clash 5: a dot and an arrow, a government glad to use a new problem; I give one, no deal, no crowd; one more, a deal. Clash 6, an arrow: now I hold first; the crowd goes to them; I bend for two. Clash 7, an arrow: I hold; the crowd comes to me; I hold again; it tears, and I don’t mind. Clash 8, a dot: one page. Clash 9, an arrow, and I have three pages: hold, the crowd goes to them and takes one, I give my last two: a deal, with nothing left. 5:30. The poster of my nine, the three lines, and the paper’s 77 as three heaps of dots beside my nine.
C (the runner-up, and the reckless one). 0:00. Two shapes touching, an orange and a green, and the words you are a rule. hold. 0:10. My thumb on my edge. I press; the green eases; I ease; it eases further; the overlap stitches. Pleasant, unexplained. 0:40. A green shape slides in behind a black arrow. I press; it shoves; I ease; it shoves harder; the overlap grinds and tears. I do not know whether I did something wrong. 1:20. Another arrow, and dots flood in behind the green. I hold; my shape fades; I hold harder; it fades faster; I let go and the overlap stitches. Did I lose? Did I win? 2:30. I have felt three kinds of neighbour and cannot name them. Without words the player cannot know that the arrow means on purpose or the dots a public, and the paper’s sentence is out of reach: “and I don’t know what I’m supposed to feel”.
A, Who Started It? The player is inside the paper’s model at the place it puts agency: a party managing a clash. The origin is the hidden truth the player must read, as the paper’s coders had to; the management is what the player does; the order effect is what follows. The discovery lands inside the first two minutes (pages given to a clash picked to pull you down are lost; the crowd sits with you, and holding out is free) and its twist inside three (the crowd behind a challenger makes holding out cost, and the clash picked on purpose ends in a deal). The deliberation is real every time: pay now or wait, how much, and what to keep for later. The sentence afterwards has no “guess”: I settled the accidents for a page each, held against the bloc that wanted my Council out, and bent to the sun-and-wind agency once the crowd was behind it. Originality: no game in the table has put the player across a table from a counterpart whose motive is hidden, or drawn regime complexity as overprinted inks.
Author test. The authors would recognise their Figure 2 run forwards (origin, management, effect), their coordination games (a page given is a page returned) and collaboration games (a page given is a page taken), the “at least initially” of IIC management, their IRENA explanation turned into a mechanism, their three clusters in exact proportion in every deck, the Zimbabwe case as the messy one, and their 77 on the last screen. A colleague who knows the answer still has to spend fourteen pages well against a deck that does not say which challenger is which until the crowd forms, and may enjoy discovering that their own hybrid cluster is the one that eats the budget.
Runner-up: C, Push. Switch if the authors want a ninety-second toy for a keynote, in which the three paths are felt rather than read, and are willing to put the three origins in words on the start screen.
What the reckless concept would have given. The other side’s motive felt through the thumb, without a word. Kept from it: the other side’s answer is drawn as motion in the picture (they move to meet you, or they push on), and the result line in words comes after the motion. Given up: the silence and the real time.
One game of nine clashes, each one or two offers: about five to six minutes. Screens: the poster (start); the clash (play; one per clash, the same layout); the end poster (verdict); How to play; About. No presets: the paper has no regimes to switch; its three clusters are in every deck, in the paper’s proportions (35 : 26 : 16 is 4.1 : 3.0 : 1.9 in nine): four that began by accident (one of them messy, after the Zimbabwe case), three started on purpose with a popular cause (hybrid), two started on purpose to pull a body down (malign). Ordered for the player: clash 1 is an accident (the basic deal); clashes 2 and 3 are one of each kind started on purpose, in random order (the discovery and its twist, both inside three minutes); clashes 4 to 9 are the rest, shuffled. Each deck is drawn from a pool of fourteen clashes. Score: none shown; the row of marks and the column of pages are the yardstick; the verdict is in words. The game ends once; PLAY AGAIN deals a new deck.
Register: a Swiss International Style poster (Austere family: a Swiss grid poster). It comes from the setting of the data and from the paper’s object at once. The institutions in the ICD live largely in Geneva and its neighbours (the UN’s human-rights office, UNHCR, WHO, the WTO, ILO), and the style that was Switzerland’s house style from the 1950s to the 1970s, when the dataset begins, is the one that called itself international. Its signature image, flat transparent inks in overlapping circles (Müller-Brockmann’s concert posters), is the paper’s object: overlapping institutions. The drawing is the mechanism: circles are bodies, the lens is the overlap, the overprint is integration.
#EAE6DC (the ground; a tear shows it); ink #141414 (type, the arrow, the falling dot, the crowd, the primary button); signal orange #F2581B (your rule: your circle, your pages, your border); grass green #1C9A55 (their rule: their circle, their border); olive #55601A (the overprint of the two: a deal, and nothing else).Kind: a toy, calibrated to the paper’s three clusters. The paper hands over an argument and a cluster analysis, not a model; the game formalises the argument as a bargaining toy in which the player is one party and the other party’s behaviour is fixed by the clash’s origin, and calibrates the toy’s deck and its “years later” draws to Figure 4.
Where each staged discovery comes from.
Robustness (calibration.js, section D). The central discovery (reading how a clash began settles at least as many clashes as any strategy blind to it, and loses fewer pages to clashes nobody can settle; with pages no object, every accident and every popular challenge can be settled and no hostile one) survives: popular challengers asking 3 pages; accidents asking 2 (messy 3); a crowd that takes nothing; a crowd that takes 2, then 3; purses of 11 and 18; and a crowd that sides the wrong way one time in five. One honest limit: with a purse of 18, blind generosity (“meet halfway”) also settles all seven; the reader still loses no pages to the hostile challengers where the blind player loses two. The purse is what makes reading matter for the count; who can be settled at all does not depend on it.
Plan. (A) The draw, over 2000 decks: composition, the floor (clash 1 an accident; 2 and 3 one of each kind started on purpose), reproducibility by deck number, uniformity of the years-later draws. (B) The paper reproduced by a “historical” player who behaves as the paper’s actors do (meets accidents at once; holds against a clash started on purpose, then gives way if the crowd backs the challenger): origin share against Table 1 (54.5 %), cooperative share (76.6 %), integrative share (60.5 %), each cluster’s cooperation and integration against Figure 4, the cooperative-but-fragmenting sub-clusters against note 10; tolerance 2 percentage points (3 for the hybrid share of IICs, 4 for the benign integrative rate overall, which is higher by the declared drop of the two idiosyncratic stand-offs). (C) The obvious strategies. (D) Robustness. (E) The rules: pages never negative, offers capped, hostile never settles, settlement exactly at the price, crowd only after a first offer that fails, dig-in.
Result: 48 of 48 pass (calibration-result.txt). Origin 55.6 % on purpose (paper 54.5); cooperative 77.8 % (76.6); integrative 61.2 % (60.5); benign cooperative 100 % (94.3, declared), integrative when settled 76.5 % (75.8); hybrid share of IICs 60.0 % (61.9), integrative 77.4 % (76.9); malign cooperative 0 (0), integrative 6.2 % (6.3); cooperative but fragmenting 23.4 % benign (24.2) and 22.6 % hybrid (23.1). The historical player settles all seven settleable clashes in every one of 2000 decks and spends exactly 14 pages.
2000 decks; nine clashes, of which seven can be settled; 14 pages.
| Strategy | Settled | Accidents (of 4) | Popular (of 3) | Pages lost to the hostile |
|---|---|---|---|---|
| hold everything | 0.00 | 0.0 | 0.0 | 0.0 |
| one page each, then hold | 2.34 | 2.3 | 0.0 | 1.7 |
| meet halfway (2), then hold | 5.35 | 3.0 | 2.3 | 3.3 |
| generous: 2, then 2 more | 4.18 | 2.3 | 1.9 | 5.6 |
| blind top-up: 1, then 3 more | 3.10 | 1.6 | 1.5 | 4.9 |
| gambler: pay 2 at once to every clash on purpose | 6.16 | 3.5 | 2.7 | 3.8 |
| reader: accidents at once; on purpose, wait for the crowd | 7.00 | 4.0 | 3.0 | 0.0 |
| oracle (knows every kind) | 7.00 | 4.0 | 3.0 | 0.0 (3 pages left) |
The spread is wide (0 to 7), so decisions matter. One simple rule wins everywhere, and it is the sentence the player should leave with: settle the accidents at once; with a clash started on purpose, see whom the world backs. It is also the paper’s historical behaviour, which is why the purse is set to what it costs (14). The gambler comes close (6.2): paying at once is a fair bet, which keeps the first offer on a clash started on purpose a real decision. A first-time learner (one page to the first accident, two to the first clash started on purpose, then the rule) settles six of seven in the scripted sessions.
mechanism.js, the only copy): CONFIG with every number tagged; the pool of fourteen made-up clashes; the deck (seeded, mulberry32); offers, the crowd, the end, the years-later draw; the game’s own voice for every result; the verdict composed from the player’s clashes by kind. Calibration first run: 35 of 39, failing on a variant-patching bug in the check itself, an ill-posed robustness test (a budget-bound oracle used to ask “can every accident be settled?”), the benign integrative rate (overall vs when settled; split into two checks), and the purse-of-18 variant, which showed that blind generosity also settles everything when pages are plentiful (the robustness criterion was restated and the limit recorded in section 16).loop.js, loop-log.txt): read as the stranger. Fixes: the grey paper line under a popular challenger said “settled in the end” even when the player had not settled it (now outcome-neutral); “the one page you gave count” (grammar); “two pages more” (now “two more pages”); an accident refused for too few pages said “they took it and pushed on”, the hostile’s line (now “not enough for them”). The first deal a player sees could fray (about one deck in four): a floor now swaps clash 1’s years-later draw with another accident’s in the same deck, so the deck’s frays are unchanged and the calibration stands.template.html, build.py): the Swiss poster. First screenshots: the outermost stripes vanished into their own circles (stripes reordered green-first, lens outlined); the page preview was invisible (green over green; now a green hatch on stock); the falling dot sat on the stitches (fades when the lens changes); results ran below the fold (the story lines fold away after the first offer, the first result and the crowd stay on screen for the last word, and the next button is scrolled into view); the “pushed on” move shifted the green circle off its lens (now the arrow jabs twice).session-test.js, session-log.txt, 22 screenshots): game A (the learner, deck 1: six of seven), How to play opened mid-clash and returned, About from the end; game B (hold everything: none of seven, the crowd takes nine pages); game C (deck 1 replayed by its number, the gambler: six of seven). Every offer starts at 0; no errors; no sideways scroll at 380 px. Length: an estimate of five to six minutes, not measured with a person.Naming. build.md asks for the author’s name in the title. With three authors, the names are set as part of the title lockup on the poster (who started it? / kreuder-sonnen / schmotz / zürn) and in the page title, instead of a possessive.