Testing Our Own Thinking

People-powered, AI-Generated AI-assisted draft, verified, and corrected: Anthropic Claude Sonnet 5 Edited and curated by: Guy Wolf *Disclaimer: AI has been known to malfunction “hallucinate” therefore, a human-in-the-middle approach with a separate draft and verification AI has been used to provide the best quality, most accurate apolitical research on this topic Subscribe now Written Wednesday, October 28, 2026. Parts 1 to 5 built an argument. This part tries to break it. Why test it It is easy to collect evidence for what we already believe. Analysts use structured techniques to fight that habit. I name four plainly and use each one. Technique 1: Analysis of Competing Hypotheses This method lists several possible explanations and then looks for the evidence that rules some out, instead of piling up evidence for a favorite. It guards against falling in love with one story. The question: what actually separates the movements that won from the ones that did not? I compared the five wins from Part 2 with the 1988 Burma and 1989 Beijing uprisings, which Sharp himself discusses as cases where huge nonviolent demonstrations ended in many deaths and did not end the regimes.1 The ratings are my judgment from the sources in Part 2. Six cases are a thinking exercise, not statistics. Comparing the Six Cases 1. Huge crowds or turnout Philippines, 1986: Present Poland, 1989: Present Chile, 1988: Present Serbia, 2000: Present Hungary, 2026: Present Burma, 1988, and Beijing, 1989: Present 2. Security forces or insiders shifted or declined to enforce the regime’s will Philippines, 1986: Present Poland, 1989: Partly; no Soviet rescue Chile, 1988: Present Serbia, 2000: Present Hungary, 2026: Accepted the result Burma, 1988, and Beijing, 1989: Absent 3. Independent counting exposed fraud Philippines, 1986: Present Poland, 1989: Not applicable Chile, 1988: Present Serbia, 2000: Present Hungary, 2026: Not applicable Burma, 1988, and Beijing, 1989: Not applicable 4. Strikes or workplace noncooperation Philippines, 1986: Partly Poland, 1989: Present Chile, 1988: Partly Serbia, 2000: Present Hungary, 2026: Absent Burma, 1988, and Beijing, 1989: Unclear 5. Outside pressure or support Philippines, 1986: Partly Poland, 1989: Partly Chile, 1988: Partly Serbia, 2000: Partly Hungary, 2026: Unclear Burma, 1988, and Beijing, 1989: Unclear What the Comparison Suggests The first finding is important: huge crowds appear in every case, including the failures. Crowd size alone cannot explain why some movements win and others lose. The second factor is more revealing. In each successful case, the people expected to enforce the regime’s will stopped doing so, shifted their allegiance, or accepted the outcome. In the failed uprisings, the coercive apparatus did not make a comparable shift.2 That fits Sharp’s theory and the Clausewitz reading in Part 3. Crowds matter because of what they do to the enforcers. Technique 2: Key Assumptions Check This method lists what must be true for an argument to stand, then asks how confident we should be in each assumption. Four Assumptions Underlying This Series 1. Assumption: Past wins are a fair guide to the present. What if it is wrong? We tend to study winners, and the success rate fell from about half to about a third after 2010. Confidence: Medium. What would change my mind? A 2026 review showing new campaigns beating the historical averages. 2. Assumption: A model built for dictatorships helps in strained democracies. What if it is wrong? Democracies have courts, free media, and elections that dictatorships lack. The most effective tools may therefore differ. Confidence: Medium to low. What would change my mind? Clear cases of mass noncooperation changing outcomes inside democracies. 3. Assumption: Enforcers and officials can be moved. What if it is wrong? Loyal or ideologically committed forces may refuse to change their allegiance. Confidence: Medium. What would change my mind? Evidence of refusals to enforce orders in comparable settings. 4. Assumption: Counts can be made visible. What if it is wrong? When counting is controlled by the regime, independent monitors may be blocked. Confidence: Medium to high where monitors operate. What would change my mind? Reports of monitors being barred. The Weakest Assumption The second assumption is the weakest, and I say so plainly. Sharp wrote for dictatorships. Applying his ideas to Brazil, Israel, or the United States is my inference, not his claim. Technique 3: Devil’s Advocacy Here is the strongest case against everything so far, argued as hard as I can. Nonviolent campaigns won about a third of the time in the 2010s, down from half, even though more people than ever used them.3 Governments learned. Researchers report that the large majority of nonviolent campaigns since 2007 have met lethal violence, and that easier communication also means easier surveillance.4 The wins I picked are famous because they succeeded, and failures are forgotten. Hungary was an election, not a resistance, and Serbia’s win on October 25 (Part 2) was also decided at the ballot box, though after two years of street protest. And telling people in places with loyal, ruthless enforcers to stay peaceful may be asking them to absorb the beatings. I think the first three points are largely right, and they are why this series keeps saying “conditions” and “odds.” The fourth I address in the objections below. Technique 4: Indicators and Warning This method lists signals to watch for in advance, so you are not arguing about what counts as a sign after the fact. In the days before the U.S. vote on November 3, and in any coalition talks or disputes that follow elsewhere, these are worth watching. Signs that institutions are holding: Results published in detail and matching independent counts Certification on the legal schedule Courts hearing complaints and ruling on them The losing side conceding or going to court Signs of trouble: Counting or certification rules changing mid-count Official and independent tallies diverging without explanation, which was the signal in Manila5 Ballots, machines or records being seized without lawful process, one of the scenarios election-administration experts list6 Officials refusing to certify with no legal basis Coordinated calls to stop counting If warning signs appear, Sharp’s advice on blocking coups applies: deny the people taking over any legitimacy, and withdraw the cooperation they need, immediately, using the same methods as before.7 An outside lens: how courage spreads Here is a tool from a different field. In 2007 the sociologists Damon Centola and Michael Macy, who study how behavior spreads through social networks, separated “simple” contagions from “complex” ones. A simple one, like news or the flu, needs only one contact. A complex one, which is costly, risky or unfamiliar, needs reinforcement from several people before someone acts.8 Joining a strike is a complex contagion. Their finding is surprising. Long-distance ties, the “weak ties” that make news go viral, can fail to spread risky behavior, because a signal that crosses a long tie arrives alone, with no one nearby to confirm it. What works is dense clusters of people who see each other act, connected by “wide bridges” of overlapping ties.9 Centola’s 2010 experiment found that people joining an online health forum signed up more in clustered networks than in randomized ones.10 This helps explain a pattern in the cases. Solidarity grew from workplace committees. Otpor built local cells. Chile registered voters neighbor by neighbor. It also fits Chenoweth’s worry that a digital age of quick mobilization has produced big crowds without the dense networks that sustain risky action.11 The practical lesson is to build many small groups where people know each other, and then connect them in several places, not just through a hashtag. Where it strains. The theory comes from models and experiments on behaviors like joining a website, not from revolutions. Real movements involve emotion, leaders, repression and state counter-moves, and nothing in network science says why an army would defect. Treat it as a useful question, not a result. The belief holding the whole series up Nearly every argument here rests on one popular claim: if enough ordinary people peacefully defy a government, it gives way. In its most famous form it is the “3.5 percent rule.” It deserves its own stress test. Where it came from. Erica Chenoweth and Maria Stephan, comparing 323 campaigns, found that nonviolent ones achieved their goals about 53 percent of the time versus 26 percent for violent ones. They defined success strictly: the goals had to be fully met within about a year of the campaign’s peak and as a direct result of it, so a government falling to a foreign army did not count, and India’s independence movement was excluded because Britain’s weakened military was judged a deciding factor.12 The 3.5 percent figure was an observation that no campaign with that level of active participation at a peak event went on to fail.13 It became a “rule” through a 2013 TEDx talk and a 2019 BBC article.14 How it got flattened. Chenoweth herself has said hitting 3.5 percent does not guarantee success, particularly where change is harder, and that the idea fits regime-change campaigns better than ordinary reform.15 Researchers have found at least two cases that broke the rule, and she has called it a description of the past rather than a prediction.16 Where it holds. The BBC reports that Estonia’s Singing Revolution and Georgia’s Rose Revolution reached the threshold.17 Serbia’s Belgrade crowd, Manila’s EDSA and Poland’s 10 million union members fit the same pattern of very broad participation followed by success. Where it fails. Beijing in 1989 drew huge crowds and ended in massacre, and the 2010s saw record-sized campaigns with falling success. Chenoweth’s own explanation is that winning campaigns do more than march: they use disruptive methods that get around repression, and they hold discipline even as the state turns violent.18 The honest, conditional answer. Large crowds raise the odds of success because they make it likelier that insiders will defect and because they are harder to stop, but crowds are a tool and a sign, not the cause. The cause is a shift in who obeys. Nonviolent resistance is more likely to win than armed resistance on average, and the odds are much better when it combines broad participation, varied and disruptive tactics, discipline under pressure and a plan for dividing the regime’s supporters. It is not a guarantee, and it is not safe. Two fair objections “This is a regime-change playbook for foreign powers.” This is a serious charge. Russia and China describe “color revolutions” as foreign-engineered, and use that framing to justify restricting foreign-funded groups and tightening controls.19 Defenders note that Sharp’s ideas have been used against governments friendly to Washington, and that most users were fighting their own rulers. What I can say is that the argument for the tools does not depend on who first used them, and that every win in Part 2 was carried out by citizens of the country involved, who took the risks. “Asking people to stay peaceful asks the powerless to absorb violence.” This one I take most seriously. Sharp does not pretend otherwise. He writes that fighting dictators will bring casualties and hopes strategy can reduce them.20 The data are consistent with that: most nonviolent campaigns meet some lethal force, yet nonviolent campaigns still outperform armed ones on average.21 But a statistic is cold comfort to someone standing in front of a tank, and whether to take such risks is a personal and moral decision that no article should make for anyone. In brief Tested against the cases, crowd size does not separate winners from losers. A split among the people who enforce power does. The case that this no longer works as well is largely right: success rates have dropped to about a third. The “3.5 percent rule” is a pattern in past data, not a law. Network science suggests risky action spreads through small, dense groups that know each other, which matches how Solidarity, Otpor and Chile’s registration drive were built. Something you can do today Ask three people you trust whether they would join a specific, lawful, nonviolent action if you organized it. Their answers show whether you have a crowd or a network. Something you can share Crowds don’t topple regimes. Defections do. Every uprising that won had a point where the people enforcing power stopped. Every one that failed didn’t. #CivilResistance About this article Written October 28, 2026, with Claude Sonnet 5.5 (knowledge cutoff end of June 2026). The comparison table is my own judgment, not a dataset. Notes Gene Sharp, From Dictatorship to Democracy, 4th U.S. ed. (Boston: Albert Einstein Institution, 2010), 60–61. Roy Rosenzweig Center, “Introductory Essay: Poland”: 1989.rrchnm.org. Erica Chenoweth, Harvard Kennedy School Wiener Conference Call: hks.harvard.edu. Summary of Chenoweth, “Trends in Nonviolent Resistance and State Response,” Waging Nonviolence, July 2017: wagingnonviolence.org; Jonas Holmberg, “The Decline in Success of Nonviolent Conflicts,” Uppsala University: journals.uu.se. University of the Philippines, “Remembering the 1986 Snap Election”: oica.upd.edu.ph; National Democratic Institute: ndi.org. Outlook India, “Trump and the US Midterms 2026: Nine Concerns Over Election Administration”: outlookindia.com. Sharp, Dictatorship to Democracy, 74–75. Damon Centola and Michael Macy, “Complex Contagions and the Weakness of Long Ties,” American Journal of Sociology 113, no. 3 (2007): 702–34: pdodds.w3.uvm.edu. Review of Damon Centola, How Behavior Spreads, American Journal of Sociology: ndg.asc.upenn.edu; Centola and Macy, “Complex Contagions.” Damon Centola, “The Spread of Behavior in an Online Social Network Experiment,” Science 329, no. 5996 (2010): 1194–97: faculty.cs.byu.edu. Chenoweth, Harvard Kennedy School Wiener Conference Call. David Robson, “The ‘3.5% Rule,’” BBC Future, May 13, 2019, republished at transcend.org. Robson, “The ‘3.5% Rule.’” Erica Chenoweth, “My Talk at TEDxBoulder,” Nov. 4, 2013: rationalinsurgent.com; Robson, “The ‘3.5% Rule.’” “Were the No Kings Protests the Largest Single-Day Demonstration in American History?,” The Guardian via AOL: aol.com. Social Change Lab, “What Determines the Success of Movements Today?,” Resilience.org, Jan. 31, 2023: resilience.org. Robson, “The ‘3.5% Rule.’” Chenoweth, Harvard Kennedy School Wiener Conference Call. “Color Revolution Doctrine,” Model Diplomat glossary: modeldiplomat.com. Sharp, Dictatorship to Democracy, x. Waging Nonviolence summary of Chenoweth, “Trends in Nonviolent Resistance and State Response”; Robson, “The ‘3.5% Rule.’” Part 6 of the series “Power Is Borrowed.”