Ten lessons on why states fight, cooperate and disagree about the rules: theory, bargaining, institutions, trade, climate and the people left out of state-centred accounts.
An introductory undergraduate survey. The title is a real course title, used by MIT. This is an original guide, not an MIT course or a claim that one syllabus is globally the most popular. It introduces major debates, not a complete diplomatic history or legal qualification.
No prior politics course. Basic percentages and willingness to compare rival explanations.
Course outline
The border is not the whole story
Distinguish an outcome from its proposed cause, and locate explanations at different levels.
Three lenses, one stubborn fact
Use realism, liberalism and constructivism to generate different testable expectations.
The fence that scares the neighbour
Explain a security dilemma without assuming either party wants conquest.
Why fight over a divisible pie?
Calculate a bargaining range and distinguish information from commitment problems.
The promise and the handcuffs
Compare signalling, monitoring and costly commitment without confusing them.
A veto is not a vote count
Apply the UN Charter's voting distinctions and separate legal authority from political agreement.
Same rule, different border
Distinguish WTO most-favoured-nation treatment from national treatment and separate gains from distribution.
The clean air nobody can fence in
Explain free riding and connect climate cooperation to the Paris Agreement's actual design.
Permission to fight is not permission to target
Separate the law on resort to force from distinction and proportionality in armed conflict.
Who disappears when the map zooms out?
Build a multi-level explanation and audit whose experience its evidence excludes.
Sources and curriculum note
Reviewed October 4, 2026. English edition; Chinese translation pending. Rules are simplified for teaching; consult the full text before applying them to real disputes.
Read every lesson below. The interactive reader above contains the same explanations, with visual tools and quizzes.
1. The border is not the whole story
Learning goal: Distinguish an outcome from its proposed cause, and locate explanations at different levels.
A newspaper headline might say "two states escalated". That describes an outcome; it does not explain it. An explanation needs a mechanism: a leader misread a signal, a coalition rewarded toughness, or each state feared the other would gain a strategic advantage. International Relations studies conflict and cooperation across borders, including states, organisations and people. A state-centred account is useful, but it is not the whole field. The introductory MIT course asks why conflict and cooperation occur, rather than treating news events as self-explanatory.
Use levels of analysis as filing drawers, not rival religions. An individual-level claim focuses on a decision-maker. A domestic claim focuses on institutions, groups or political incentives inside a state. An international claim concerns interaction, power or rules between actors. A complete argument can connect drawers: a domestic election changes a leader's incentives, which changes bargaining with another government. The level tells you where a proposed cause sits, not whether it is correct.
Consider the invented states Alder and Birch. Both increase naval patrols. One account blames an election; another blames a new military capability. Ask what each account would predict if its proposed cause were absent. If patrols also rise in years without elections, the election story needs revision, not immediate rejection. If the capability arrived after the patrols began, it cannot explain their onset. Ordering evidence in time is often more useful than accumulating dramatic details.
The discipline also asks whose outcomes are counted. A deal that lowers the chance of interstate war might expose coastal communities to new harms. Do not call it a success without naming the criterion. Keep an outcome, a causal explanation and a moral judgement in separate sentences. That habit will stop later theories from turning into labels that merely restate what happened.
Source basis: S1, S2, S11, S17. Cases, mechanisms and exercise values are original teaching examples.
Worked example
Alder adds patrols before its election campaign begins. Does that timing refute the claim that the campaign caused the initial increase?
Identify the outcome: onset of additional patrols.
Identify the proposed cause: the campaign.
The cause follows the outcome, so it cannot explain this onset as stated.
The campaign might still explain a later increase. Keep the claim narrow.
Practice problem and solution
A governing coalition blocks a foreign agreement because one industry would lose. Name the level of this proposed cause.
Domestic. The mechanism runs through a group and its influence inside the state.
Mental model: Name the outcome, the mechanism and the level before judging a story.
Common trap: Calling an outcome an explanation.
2. Three lenses, one stubborn fact
Learning goal: Use realism, liberalism and constructivism to generate different testable expectations.
Realism puts security and power near the centre of analysis. In an international system without a general government above states, an actor must consider what others can do, not only what they promise. "Anarchy" here means absence of overarching authority, not constant chaos. Different realist arguments emphasise different causes; a theory is not a rule that every state attacks whenever it can.
Liberal approaches add domestic institutions, economic ties and international institutions to the explanation. They ask how rules and repeated interaction can support cooperation, and how preferences differ inside states. Cooperation is possible without a world government, but it is not guaranteed by a meeting or a treaty signature. A monitoring rule must supply useful information, and an enforcement arrangement must change incentives to matter.
Constructivism asks how identities, interests and meanings are formed through social interaction. Two equally powerful neighbours may be interpreted differently because one is understood as a partner and the other as a rival. This is not a claim that weapons are imaginary. It is a claim that the meaning of a capability cannot always be read from its size alone. Norms can shape what actors want, not merely limit how they pursue fixed goals.
Take an invented inspection agreement. A realist lens asks whether inspections expose a relative vulnerability. A liberal lens asks whether they reduce uncertainty and make cheating costly. A constructivist lens asks whether reciprocal access changes the relationship itself. To compare them, look for discriminating evidence. Signing alone supports none of these explanations uniquely. Changes in compliance after a monitoring failure might support one mechanism; changes in what officials regard as legitimate might support another. Use the lenses to ask sharper questions, not to colour-code countries.
Source basis: S3, S13, S14. Cases, mechanisms and exercise values are original teaching examples.
Worked example
Birch complies only after inspections begin, but leaders continue calling Alder an enemy. Which mechanism is better supported?
Separate behaviour from stated identity.
The timing supports an information or incentive mechanism.
Unchanged language weakens, but does not conclusively refute, an identity-change explanation.
Ask what inspectors revealed and what consequences followed.
Practice problem and solution
A treaty improves compliance because public reports expose cheating. Name the best-supported mechanism: information or identity.
Information. The stated causal link is exposure, not a change in who the parties think they are.
Mental model: A lens earns its place by explaining an intermediate link.
Common trap: Treating realism as pessimism, liberalism as kindness and constructivism as wishful thinking.
3. The fence that scares the neighbour
Learning goal: Explain a security dilemma without assuming either party wants conquest.
A security dilemma can arise when one actor's attempt to become safer makes another feel less safe. Realist theory links this problem to uncertainty and self-help under anarchy. The important distinction is between intentions and capabilities. A leader can sincerely intend defence while acquiring equipment that could support attack. A neighbour observes the equipment more easily than the sincerity.
Imagine Alder builds a watchtower to prevent surprise attacks. Birch sees improved targeting capacity and builds a second tower. Alder reads that response as evidence of hostile intent and expands again. Neither actor needs to begin with a plan for conquest. Each can rationalise its next move as protection. That is why identifying a defensive motive does not establish that a policy will have defensive effects.
Reassurance aims to interrupt the inference. It might make a capability more observable, limit where it can be used, or pair a restriction with verification. In this original case, moving the tower away from the border reduces offensive reach but preserves local warning. Whether that works depends on what each side can verify and what the equipment can actually do. Cheap words alone may leave the original uncertainty untouched.
Not every rivalry is a security dilemma. If one actor seeks territory regardless of reassurance, the problem may be incompatible goals rather than mistaken inference. Test the explanation: does a credible limit reduce the response, or does escalation continue when the defensive ambiguity is removed? A negative result does not prove aggression, but it narrows the explanation. Keep the model conditional. It helps explain a route to escalation; it does not diagnose every conflict from a photograph of military hardware.
Source basis: S3, S12. Cases, mechanisms and exercise values are original teaching examples.
Worked example
Alder says its missiles are defensive but refuses any range verification. Has it removed Birch's uncertainty?
Intent is asserted, not verified.
The relevant capability remains usable across the border.
A cheap statement may not change Birch's estimate.
Seek an observable restriction rather than another declaration.
Practice problem and solution
A leader wants defence but acquires equipment usable for attack. What remains ambiguous to the neighbour: intentions or the existence of equipment?
Intentions. The equipment is observed; its intended use is not established by observing it.
Mental model: Defensive intent and defensive effect are not the same.
Common trap: Assuming all escalation proves aggression, or all defensive statements prove reassurance.
4. Why fight over a divisible pie?
Learning goal: Calculate a bargaining range and distinguish information from commitment problems.
War is costly. If two actors can divide a disputed benefit, why not agree on a division and avoid those costs? Fearon's 1995 paper develops this rationalist puzzle and examines mechanisms that can prevent mutually preferred deals. Two central mechanisms are private information with incentives to misrepresent, and commitment problems. The argument is a framework for explanation, not a claim that people always calculate correctly or that every war has the same cause.
Use a deliberately simple hypothetical: a benefit of 100 points, Alder's probability of winning 0.6, and war costs of 10 points to each side. Alder's expected war payoff is 0.6 times 100 minus 10, or 50. Birch's is 0.4 times 100 minus 10, or 30. If Alder receives x from a peaceful deal, Birch receives 100 minus x. Both weakly prefer peace when x is at least 50 and at most 70. Strict preference holds inside that interval. These are invented utility points, not lives, money or a forecast.
Private information can obstruct agreement when actors know different things about strength or resolve and have incentives to exaggerate. A demand may be a truthful warning or a bluff. More messages do not automatically solve that problem. A commitment problem is different: parties may understand today's bargain perfectly but expect one side to exploit tomorrow's changed power. Information about that future shift can make the problem clearer without making the promise credible.
Do not read the pie as every dispute. Some goods are hard to divide, actors may value symbols or have non-linear preferences, and domestic constraints matter. The model assumes the parties evaluate the same divisible benefit, know the relevant probabilities and can implement the settlement. Its strength is diagnostic: first show a feasible range under stated assumptions, then identify which assumption or enforcement condition fails. "They fought because they disagreed" has not yet answered the puzzle.
Source basis: S4. Cases, mechanisms and exercise values are original teaching examples.
Worked example
Keep the 100-point benefit and 0.6 win probability, but make each war cost 20. Derive the new range.
Alder war payoff: 60-20=40.
Birch war payoff: 40-20=20.
Birch accepts if 100-x is at least 20, so x is at most 80.
The weakly preferred range is 40≤x≤80; strict preference is 40<x<80.
Practice problem and solution
Alder wins with probability 0.5 over 100 points; both war costs are 15. What is Alder's maximum share in a deal Birch weakly prefers to war? Enter the number.
65. Birch's war payoff is 50-15=35, so 100-x≥35 implies x≤65.
Mental model: A bargaining range is not a guarantee that a bargain can be enforced.
Common trap: Using information to solve an enforcement problem.
5. The promise and the handcuffs
Learning goal: Compare signalling, monitoring and costly commitment without confusing them.
A signal gives an audience something from which to infer intentions, capability or resolve. A statement can be cheap to make. An action can impose a cost or change what the sender can do next. Neither is automatically credible: ask whether actors with different intentions would behave differently. A cost that every type willingly pays may separate nothing. Fearon's bargaining analysis makes the incentive to misrepresent central, so "more communication" is not a complete remedy.
Imagine a fictional government demands port access. It can issue a speech, invite reciprocal inspections or remove a battery from the coast. The speech leaves capabilities unchanged. Inspections may improve information while leaving the battery in place. Removal alters a capability, but may be reversible. The meaning of each action depends on the question being answered. Evidence of restraint today does not establish an ability to bind the next government.
Domestic politics can alter both preferences and room for agreement. In our invented example, a legislature must approve any port treaty. A negotiator may genuinely be unable to promise a term the legislature will reject. That constraint might support credibility, but it can also shrink the set of feasible agreements. Do not assume democracy always supplies credibility or peace. Show which institution imposes which constraint and whether the foreign audience understands it.
The useful distinction is between changing beliefs and changing incentives. Verification reveals conduct. A penalty makes some conduct less attractive. A physical limit makes some conduct harder. An agreement can combine all three, and each can fail independently. If inspectors cannot access the relevant site, the information link fails. If penalties will never be imposed, the incentive link fails. If the limit is easy to reverse, the commitment may last only until the next bargaining round. Trace the weakest link before calling a promise secure.
Source basis: S2, S3, S4, S13. Cases, mechanisms and exercise values are original teaching examples.
Worked example
A government posts a bond forfeited if verified cheating occurs. Which parts of the design supply information and incentives?
A missing inspection breaks detection; a non-collectable bond breaks the penalty.
Evaluate both links rather than calling the bond itself verification.
Practice problem and solution
In a fictional system a legislature must approve a treaty and rejects its central term. Does this widen or narrow the feasible agreement set?
Narrow. The negotiator cannot credibly offer that central term under the stipulated rule.
Mental model: Separate a message, a monitor and a constraint.
Common trap: Calling every costly action a commitment.
6. A veto is not a vote count
Learning goal: Apply the UN Charter's voting distinctions and separate legal authority from political agreement.
The UN Charter sets out sovereign equality in Article 2(1), peaceful dispute settlement in Article 2(3), and a prohibition on threat or use of force against the territorial integrity or political independence of any state in Article 2(4), with its full wording and context. These provisions do not say that all states have equal material power. Article 51 preserves individual or collective self-defence if an armed attack occurs, subject to the Charter's conditions. Read authority to use force separately from rules governing how force is used.
Security Council voting is not a simple majority exercise. Under Article 27, procedural decisions require nine affirmative votes. Other decisions require nine affirmative votes and, under the text, the concurring votes of permanent members. The Council's official explanation of practice says a permanent-member abstention does not prevent adoption when the required nine affirmative votes are obtained. A permanent-member negative vote on a non-procedural matter does. The Charter also specifies abstention by a party to a dispute for certain Chapter VI and Article 52(3) decisions; this short lab assumes that issue is absent.
For a fictional non-procedural proposal, eleven yes votes plus a permanent-member no is not adoption. Nine yes votes with a permanent-member abstention and no permanent-member no can be adopted. For a procedural proposal, a permanent-member no does not create a veto; the affirmative threshold still matters. Always establish the category before counting. These exercises are stylised applications of the rule, not records of actual resolutions.
Institutions can supply a place to negotiate, coordinate and authorise action, but political deadlock is possible. Lack of Council agreement does not make every unilateral action lawful, and Council voting alone does not tell you whether a use of force satisfied all legal requirements. Article 25 concerns members accepting and carrying out Council decisions under the Charter; Chapter VII includes determinations under Article 39 and measures under Articles 41 and 42. Do not treat a political preference, a vote and a legal entitlement as synonyms.
Source basis: S5, S6. Cases, mechanisms and exercise values are original teaching examples.
Worked example
A non-procedural draft receives ten yes votes. One permanent member abstains; none votes no. Is that abstention a veto?
Classify it as non-procedural.
Ten yes votes exceed the nine-vote threshold.
Official Council practice treats permanent-member abstention as not blocking adoption.
Under the lab assumptions, it can be adopted.
Practice problem and solution
A non-procedural draft has twelve yes votes and one permanent-member no. Enter adopted or blocked.
Blocked. The numerical threshold is met, but the permanent-member negative vote vetoes this non-procedural draft.
Mental model: Classify, count, then check the applicable veto rule.
Common trap: Treating abstention as a no, or nine yes votes as always sufficient.
7. Same rule, different border
Learning goal: Distinguish WTO most-favoured-nation treatment from national treatment and separate gains from distribution.
Trade rules often address discrimination rather than banning every border charge. The WTO's explanation distinguishes most-favoured-nation treatment, which generally compares treatment of trading partners, from national treatment, which compares imported and domestic products after entry into the market. The principles operate differently across the agreements on goods, services and intellectual property. This lesson's stylised cases concern goods, not the full law of each sector.
If a state gives one partner's comparable imports a lower tariff, the MFN question is whether other WTO members should receive that treatment too. Exceptions exist, including qualifying free-trade arrangements and specified preferences; you must check the applicable agreement and conditions. National treatment asks a different question. Once an imported good enters the market, is it treated less favourably than the domestic counterpart under the relevant rule? A customs duty at the border is not, by that fact alone, a violation of national treatment.
The economic question is distinct again: who gains and who loses? In an original teaching model, consumers gain 80 points while a domestic producer loses 30, giving a net 50-point gain for those counted. That sum does not tell us whether the producer can bear the loss, whether workers can change jobs, or whether compensation occurs. A policy can expand an aggregate benefit while creating opponents who bear concentrated costs. Do not label every opponent irrational simply because an aggregate total is positive.
A careful policy answer uses separate pages in its mental notebook: legal rule, causal economic mechanism and distributional judgement. A favourable aggregate number does not establish legality. A compliant rule does not establish fairness. A discriminatory measure may fall within an exception, but naming an exception does not prove its conditions are met. For a real case, compare like goods, identify the relevant agreement and inspect the actual measure. The toy numbers help identify questions; they are not estimates of any country's trade policy.
Source basis: S2, S7. Cases, mechanisms and exercise values are original teaching examples.
Worked example
A government taxes imported tea more than domestic like tea after sale, but applies the same tariff to all foreign suppliers. Which distinction matters?
Equal border treatment across partners addresses an MFN comparison.
Different internal treatment raises a national-treatment question.
Do not substitute one comparison for the other.
A full legal conclusion still needs the relevant facts and agreement.
Practice problem and solution
A lower tariff is granted to one foreign partner, not others. Name the principle to investigate first, before checking exceptions.
MFN. This is a comparison across foreign partners, subject to applicable exceptions.
Mental model: Specify the comparison before naming the rule.
Common trap: Confusing an aggregate gain with compensation, legality or fairness.
8. The clean air nobody can fence in
Learning goal: Explain free riding and connect climate cooperation to the Paris Agreement's actual design.
A collective-action problem can exist even when all parties would prefer a jointly improved outcome. If one actor pays for mitigation while others share some benefits, each may prefer that others contribute first. This is a conditional incentive claim, not a statement that no one cares about the climate. The E-IR environment discussion treats environmental politics as crossing borders and requiring coordination among actors with unequal interests and resources.
Here is an invented two-player payoff table expressed in net utility points. If neither contributes, each receives 0. If both contribute, each receives 3. If Alder contributes alone, Alder gets -1 and Birch gets 4; reverse the choices and reverse the payoffs. Holding Birch's choice fixed, Alder does better by not contributing: 0 beats -1, and 4 beats 3. Birch faces the same incentives. Mutual non-contribution can therefore be stable even though both prefer 3 to 0. The table demonstrates the logic; it is not an estimated climate model.
The Paris Agreement is not a global command assigning identical national quotas. UNFCCC explains that each party must prepare, communicate and maintain successive nationally determined contributions it intends to achieve and pursue domestic mitigation measures with the aim of achieving them. NDCs are communicated every five years, and successive contributions represent progression and the highest possible ambition. These procedural and conduct obligations must not be mistaken for a guarantee that every pledged numerical target will be achieved.
The global stocktake assesses collective progress; the first was in 2023, with a five-year cycle. Its outcome informs later contributions. Reporting and review can improve visibility and pressure, but they do not automatically erase differences in cost, responsibility or capacity. In the toy table a domestic co-benefit could change the unilateral payoff and break the free-riding incentive. In policy analysis, specify such a mechanism rather than treating cooperation as either inevitable or impossible. Ask who pays, who benefits, what is observed and what changes when performance falls short.
Source basis: S2, S8, S18. Cases, mechanisms and exercise values are original teaching examples.
Worked example
In the toy table, Birch contributes. Alder gets 3 by contributing and 4 by not contributing. Why does the jointly better outcome remain vulnerable?
Hold Birch's contribution fixed.
Compare Alder's two net payoffs: 4 exceeds 3.
Alder has an incentive to benefit without contributing.
The collective preference for mutual 3 over mutual 0 does not remove this unilateral incentive.
Practice problem and solution
Both prefer mutual contribution to mutual non-contribution, but each gains by letting the other pay. Name the incentive: free riding or guaranteed cooperation.
Free riding. Collective benefit does not remove the temptation to avoid one's own cost.
Mental model: Distinguish shared goals from each actor's best response.
Common trap: Treating a pledge, a process obligation and measured achievement as the same fact.
9. Permission to fight is not permission to target
Learning goal: Separate the law on resort to force from distinction and proportionality in armed conflict.
Two questions must be kept apart. Under what conditions may a state resort to force? How must parties behave during an armed conflict? The UN Charter's force rules address the first. International humanitarian law addresses the second. A party's asserted right of self-defence does not make civilians lawful targets. Conversely, whether a particular attack complied with conduct rules does not settle whether resorting to force was lawful.
The ICRC's statement of customary Rule 1 says parties must distinguish civilians from combatants; attacks may be directed against combatants and must not be directed against civilians. Civilians who take a direct part in hostilities are subject to a specific exception for such time as they do so; civilian status is not lost merely through opinion or association. The broader protection of civilian objects and definition of military objectives require their own rules. Our cases focus on the distinction in Rule 1 and the proportionality principle, not a complete targeting checklist. A political identity or a hostile opinion alone does not turn a civilian into a combatant.
Proportionality in attack concerns expected incidental civilian harm in relation to the concrete and direct military advantage anticipated. The ICRC casebook states the prohibition where that expected harm would be excessive. It is not a demand for equal weapons, equal casualties on both sides, or a rule allowing direct attacks on civilians whenever a benefit seems large. The evaluation is prospective: what was expected based on relevant information, not a numerical ratio reconstructed solely from the outcome.
In a hypothetical planning exercise, a commander first identifies the proposed target. If it is civilians as such, a favourable claimed advantage cannot rescue the attack. If it is a military objective, incidental civilian harm still raises proportionality and other conduct obligations. This lesson does not supply a calculator or legal permission for a real strike. Neither "some civilian harm occurred" nor "we had a military aim" is a complete conclusion. A careful analyst names the applicable rule, the information needed and the uncertainty. Legal categories are safeguards, not rhetorical trophies for whichever side one supports.
Source basis: S5, S9, S10, S19. Cases, mechanisms and exercise values are original teaching examples.
Worked example
An official says a strike was proportionate because both sides used the same size weapon. What is missing?
Weapon equality is not the relevant legal comparison.
Identify the military advantage anticipated.
Identify expected incidental civilian harm.
Assess whether harm would be excessive and consider the other applicable rules.
Practice problem and solution
A proposed strike has a military objective but may cause incidental civilian harm. Name the rule comparing expected harm with anticipated military advantage.
Proportionality in attack. Target status does not settle whether incidental civilian harm would be excessive.
Mental model: Keep resort-to-force authority and conduct rules on separate tracks.
Common trap: Using proportionality to justify directly targeting civilians.
10. Who disappears when the map zooms out?
Learning goal: Build a multi-level explanation and audit whose experience its evidence excludes.
A map of states is one representation of global politics. It can hide relationships inside and across borders: workers, migrants, firms, advocacy groups and households. The E-IR introductory text includes civil society and political economy alongside states and organisations. Theories that centre state security may answer important questions while leaving the distribution of insecurity unexplained. The remedy is to specify the question, not to pretend one scale explains everything.
Feminist IR examines gendered power, including which activities and experiences count as political or as security. It is not simply a tally of women leaders, nor a claim that all women share one preference. Postcolonial approaches examine how colonial histories and hierarchies shape knowledge, representation and global relations. They do not require assuming that every contemporary outcome has a single colonial cause. Both approaches can challenge the categories used by an analysis, not merely add another variable to its existing table.
Consider an invented port agreement. Governments celebrate fewer naval confrontations; shipping firms celebrate stable routes; dockworkers face disrupted employment; nearby households face new pollution. These outcomes can coexist. A claim that the agreement "improved security" needs an object: security for whom, against what, over which period? National averages may conceal differences. Household experience alone may not explain interstate bargaining. Connect the scales with explicit mechanisms rather than ranking one as always authentic and another as always false.
For the capstone, choose a question: why was the agreement signed, or how did it redistribute risk? Build one explanation with interests, interaction and institutions. Add a rival explanation with a genuinely different intermediate link. Then list evidence that could weaken your preferred account. Finally audit the sample: whose records are available, whose accounts are absent and whether official categories hide relevant work or harm. A defensible conclusion can remain conditional. The course's method is to separate facts, assumptions, mechanisms and judgement, then show where additional evidence could change the answer.
Source basis: S1, S2, S3, S15, S16, S17. Cases, mechanisms and exercise values are original teaching examples.
Worked example
You have official press releases and ship traffic records but no worker or household evidence. What can you conclude about distribution?
The available evidence may describe official aims and shipping activity.
It does not establish worker or household outcomes.
Label that part of the conclusion as unverified.
Seek appropriate accounts and measurements rather than converting absence into zero harm.
Practice problem and solution
Official sources show fewer confrontations, but no household data exist in the case. Enter verified or unverified for the claim that household security improved.
Unverified. The supplied evidence supports one outcome, not a different outcome for a different group.
Mental model: Ask whose outcome is explained, whose voice is missing and what would change your conclusion.
Common trap: Turning a missing voice into an invented finding.