16 — Civic Life, Trust & National Service

16 — Civic Life, Trust & National Service

Stage 1: EVIDENCE BRIEF

Referee panel: Prof. Olivia Brandt (#57, statistics, C), Prof. June Takahashi (#17, econometrics, C), with domain specialist Prof. Leila Farouk (#40, political psychology, C)

Why this thread exists. Symposium 1 passed "civics education + voluntary national service" as a consensus remedy for low trust and polarization. Nobody tested it. This brief asks three separate questions and keeps them apart: (a) are civic knowledge, trust and social connection actually falling, and by how much? (b) does civics education change behavior or attitudes, or only test scores? (c) does national service or structured contact change trust across groups, how long does the effect last, and what does it cost? A remedy can be good for (a) and useless for (c). S1 ran these together.

All figures are as the source states them. We ran 22 searches plus direct fetches of primary pages. Anything we could not verify is marked [unverified].

---

A. Civic knowledge

1. NAEP 2022 civics, grade 8: the average score fell 2 points from 2018, the first decline since the assessment began in 1998. 22% scored at or above Proficient and nearly one-third were below Basic. (NAGB, 2023) https://www.nagb.gov/news-and-events/news-releases/2023/eighth-grade-scores-decline-in-civics-and-us-history.html
2. NAEP 2022 U.S. history, grade 8: down 5 points, continuing a decline that began in 2014. 13% at or above Proficient and 40% below Basic. (same source, 2023)
3. Annenberg Constitution Day Civics Survey 2025: 70% of U.S. adults could name all three branches of government, up from 65% in 2024 and 32% in 2018. 13% could name none. The August 2025 sample was n=1,684, with a margin of error of ±3.5 pts. (APPC, Sept 2025) https://www.annenbergpublicpolicycenter.org/americans-knowledge-of-civics-increases-annenberg-survey-finds/
4. Annenberg 2025, First Amendment: 79% named freedom of speech, 34% freedom of the press and 12% the right to petition. 40% could name three or more rights, up from 30% in 2024. (same source)

B. Does civics education change behavior?

1. Weinschenk & Dawes (BJPS, 2021 online / 2022 issue): they used transcript-linked longitudinal data. After controls and sibling (family) fixed effects, the civics-to-turnout relationship is "essentially null", with "no support for the idea that high school courses that focus on service learning, civic skills development or political issues increase turnout." https://www.cambridge.org/core/journals/british-journal-of-political-science/article/abs/civic-education-in-high-school-and-voter-turnout-in-adulthood/3E759C8ABF051C6FDA8BAC5E212C1B26
2. Jung & Gopalan (EEPA, 2024): 15 states adopted the Civics Education Initiative test mandate in 2015–18. The estimated effect on the turnout of 18–22-year-olds was −0.002 (95% CI −0.045 to 0.042), "a fairly precise estimate of a null effect". Caveat: the follow-up covers only one election cycle. https://journals.sagepub.com/doi/10.3102/01623737231195887
3. **Holbein & Hillygus, Making Young Voters (Cambridge UP, 2020):** young people report intending to vote and then do not follow through. The authors locate the gap in noncognitive skills (self-regulation, follow-through) and in the cost of voting, not in knowledge or motivation. https://academic.oup.com/poq/article-abstract/85/2/725/6294224
4. Action civics RCT evidence is not in yet. The I-ACTED cluster RCT (about 1,500 students in about 50 U.S. schools, waitlist control) published its protocol in Oct 2025, with no results. (BMC Public Health, 2025) https://link.springer.com/article/10.1186/s12889-025-24838-y

C. Trust and social capital

1. Pew, Sept 2025: 17% trust the federal government to do what is right "just about always/most of the time". By party: Republicans 26%, Democrats 9%. The figure was 77% in 1964 and has been below 30% continuously since 2007. https://www.pewresearch.org/politics/2025/12/04/public-trust-in-government-1958-2025/
2. Gallup, June 1–15, 2026: average confidence across 14 institutions was 27%. The record low was 26% (2023), and 2026 is the fifth straight year below 30%. Highest: small business 67%, military 61%, police 45%. Lowest: Congress 9%, TV news 14%, newspapers 17%. https://news.gallup.com/poll/712436/confidence-institutions-remains-near-time-low.aspx
3. Interpersonal trust (GSS, via Pew May 2025): the share saying "most people can be trusted" was 46% in 1972, 34% in 2018 and 34% in 2023–24. By age: 18–29 26%, 65+ 44%. By income: $150k+ ≈50%, under $30k ≈20%. https://www.pewresearch.org/2025/05/08/americans-trust-in-one-another/
4. Gallup church, synagogue or mosque membership: 47% in 2020, the first time below a majority. It was 70% in 1999 and 73% in 1937. Millennials were 36% in 2018–20. https://news.gallup.com/poll/341963/church-membership-falls-below-majority-first-time.aspx
5. Surgeon General advisory (2023): about half of adults report loneliness. In-person time with friends fell from about 60 min/day (2003) to about 20 min/day (2020). Social isolation is associated with about 29% higher mortality risk. https://www.hhs.gov/sites/default/files/surgeon-general-social-connection-advisory.pdf
6. AEI Survey Center on American Life (2021): adults with no close friends rose from 3% (1990) to 12% (2021). Adults with three or fewer rose from 27% to 49%. https://www.americansurveycenter.org/research/the-state-of-american-friendship-change-challenges-and-loss/
7. Volunteering (AmeriCorps/Census CEV 2023): the formal volunteering rate was 28.3% (75.7 million people), up 5.1 pts from the 2021 pandemic trough. 54% helped neighbors informally. https://www.globenewswire.com/news-release/2024/11/19/2983816/0/en/More-than-75-7-Million-People-Volunteered-in-America-AmeriCorps-Reports.html

D. National service: U.S. facts, 2025–26

1. April 2025 AmeriCorps disruption: about $400M in grants were terminated (NCAN). More than 40% of grantees were terminated on Apr 25, 2025, and about 85% of federal agency staff were let go in mid-April (NPQ). On June 5, 2025, Judge Boardman (D. Md.) issued a preliminary injunction under the APA covering 24 states + DC. https://ncan.org/Web/Web/News/Judge-Orders-Restoration-of-AmeriCorps-Funding-in-Lawsuit-Challenging-Grant-Cancellations.aspx and https://nonprofitquarterly.org/the-good-the-bad-and-the-ugly-whats-next-for-americorps/
2. Appropriations: FY25 enacted $1.262B. The President's FY26 request was $107.7M. The House committee proposed $642.95M (−49%) in Sept 2025. The final bipartisan FY26 bill provided $1,253,806,000, near-level (Jan 2026). https://voicesforservice.org/legislative-updates/house-proposes-nearly-50-percent-cut-to-americorps-for-fy26/ and https://voicesforservice.org/news/bipartisan-house-senate-appropriations-bill-provides-near-level-funding-for-americorps/
3. Settlement (reported Sept 12, 2026): AmeriCorps must give 30 days' written notice before terminating grants or dismissing members in FY26 and commit "substantially all" FY26 funds by Sept 30. Litigation is paused to Feb 1, 2027. https://wrnjradio.com/new-jersey-among-states-reaching-settlement-over-americorps-funding-cuts/
4. AmeriCorps' own books: for its FY2024 consolidated statements the OIG's auditors issued a disclaimer of opinion, the eighth consecutive year. There were 11 material weaknesses (three dating to FY2017) and 78 open recommendations. https://www.americorpsoig.gov/reports/audit/audit-americorps-fiscal-year-2024-consolidated-financial-statements
5. Return-on-investment claim: Belfield (2013), commissioned by Voices for National Service and Civic Enterprises, estimates $3.95 of social return and $2.20 of taxpayer return per $1. This is a benefit-transfer model, not an impact evaluation. https://voicesforservice.org/wp-content/uploads/2016/03/Summary_Sep18_Econ_Value_National_Service-6.pdf
6. **Mettler, Soldiers to Citizens (2005): WWII veterans who used GI Bill benefits belonged to about 50% more civic organizations and took part in about 30% more political activities** than non-users. The evidence is a retrospective survey, not randomized. https://networks.h-net.org/node/9997/reviews/10655/ortiz-mettler-soldiers-citizens-gi-bill-and-making-greatest-generation
7. Cost per AmeriCorps member: we could not verify a current official cost-per-member figure or the current Segal Education Award amount; the primary pages returned 403. Any per-member figure used in this thread is [unverified].

E. Contact, mixing and service: causal evidence

1. Paluck, Green & Green (Behavioural Public Policy, 2019): a meta-analysis of 27 RCTs finds a pooled effect of d = 0.39. No study tested racial/ethnic prejudice in adults over 25, and larger studies show smaller effects. https://www.cambridge.org/core/journals/behavioural-public-policy/article/contact-hypothesis-reevaluated/142C913E7FA9E121277B29E994124EC5
2. Mousa (Science, 2020), soccer leagues in post-ISIS Iraq: Christians randomly assigned Muslim teammates were +26 pp more likely to vote for a Muslim player for a sportsmanship prize and +13 pp more likely to register for a mixed team. They were +49 pp more likely to train with Muslims six months later. There was no significant change in attending mixed social events, patronizing Muslim-owned restaurants, or attitudes toward Muslims generally. https://www.povertyactionlab.org/updates/new-study-iraq-mixed-religion-soccer-teams-helped-build-social-cohesion-healed-wounds-after
3. Lowe (AER, 2021), cricket leagues in India: collaborative cross-caste contact (same team) increased cross-caste friendships and reduced own-caste favoritism. Adversarial contact (opposing teams) generally reduced cross-caste interaction. https://www.aeaweb.org/articles?id=10.1257%2Faer.20191780
4. Voelkel et al. (Science, 2024), Strengthening Democracy Challenge: n=32,059, 25 interventions. 23/25 reduced partisan animosity (up to about 10.5 pp). Only 6 reduced support for undemocratic practices (up to 5.8 pp) and 5 reduced support for partisan violence (up to 2.8 pp). At the two-week follow-up (n=8,644), about 32% of the animosity effect remained and about 7% of the effect on undemocratic practices. Animosity and antidemocratic attitudes are "clearly distinct." https://bpb-us-w2.wpmucdn.com/web.sas.upenn.edu/dist/9/244/files/2024/10/science.adh4764.pdf
5. Bagues & Roth (AEJ: Policy, 2023), Spanish conscription lottery: random assignment to serve outside one's region raised sympathy and trust toward the region of service. The effects persisted for decades. Men from regions with strong separatist movements came to identify more with Spain. https://www.aeaweb.org/articles?id=10.1257%2Fpol.20210237
6. Okunogbe (AEJ: Applied, 2024), Nigeria NYSC random state assignment: serving in a region with a different ethnic majority increased national pride and knowledge and made participants more likely to live outside their home region later. But immersion also strengthened ethnic pride. https://www.aeaweb.org/articles?id=10.1257%2Fapp.20210266 and https://voxdev.org/topic/institutions-political-economy/promoting-national-integration-nigeria
7. France's Service National Universel (Cour des comptes, Sept 2024): full generalization would cost €3.5–5bn/yr plus about €6bn in infrastructure. Cost was €2,900 per participant against €2,000 budgeted, with 40,000 youth in 2023 against a target of about 800,000/yr. Upper-class and high-achieving youth were over-represented, and 46% of 2023 volunteers had parents in uniformed services, which undercuts the goal of social mixing. https://www.publicsenat.fr/actualites/societe/risques-pour-la-securite-des-jeunes-cout-largement-sous-estime-la-cour-des-comptes-etrille-le-deploiement-du-snu

---

Contested evidence: where the literature genuinely disagrees

(i) Does civics education raise participation?

• Null camp: Weinschenk & Dawes (sibling FE) and Jung & Gopalan (DiD on test mandates) both find nothing on turnout.
• Positive camp: older observational work, and Holbein & Hillygus's argument that a different kind of schooling (skills, follow-through) matters.
• Not yet tested causally: action civics. The first large RCT (I-ACTED) has only a protocol.
• Referee view: content-knowledge civics is supported as a knowledge intervention. As a turnout or trust intervention it has well-identified nulls against it.

(ii) Is civic knowledge falling?

• NAEP grade 8 fell 2 points in 2022, the first decline in the series.
• Annenberg's adult knowledge measure has risen sharply since 2018 (32% → 70% naming all three branches).
• These measure different populations with different instruments. Neither supports "freefall," and neither supports "we're fine."

(iii) Does contact or service change intergroup trust durably?

• Yes, when collaborative, sustained and random-assigned: Bagues & Roth (decades), Lowe (collaborative), Okunogbe (national identity).
• Narrowly: Mousa (behavior toward teammates, not the outgroup at large).
• Decays fast when brief: Voelkel (two-week retention of about 32%).
• Paluck et al. warn that no RCT covers adults over 25 on racial prejudice, and that effects shrink with study size.

(iv) Does national service pay for itself?

• Belfield's $3.95:1 is advocacy-commissioned and model-based.
• No U.S. RCT of AmeriCorps on civic outcomes was located in this review.
• Mettler's GI Bill evidence is retrospective and confounded by selection into use.

(v) Why is trust falling?

• Candidates include secularization and associational decline (the Putnam lineage), inequality (trust rises steeply with income), partisan sorting (trust in government flips with the party in power), and media.
• The data above are descriptive. We found no well-identified study that ranks these causes against each other. Anyone claiming a single cause should expect to be rated Unsupported.

— Brandt, Takahashi, Farouk

Stage 2: OPENING POSITIONS (Part 1 of 2)

---

Prof. Tamsin Kerr (#30, public policy, C-L): national service is civic infrastructure, and we just watched it get dismantled

Causal claim. Trust tracks shared institutional experience. The U.S. has run down the institutions that put strangers from different backgrounds on a common task. National service is the cheapest scalable rebuild. Mettler's GI Bill work shows a program that treats people as citizens produces citizens: users had about 50% more civic memberships and about 30% more political activity (Brief #21).

Evidence.

• Belfield's cost-benefit study finds $3.95 of social return per $1 (#20).
• Bagues & Roth show that mixing young adults across regions builds trust that lasts decades (#27).
• In 2025 the administration terminated about $400M in grants and let about 85% of staff go (#16). The President asked for $107.7M, down from $1.262B (#17), and AmeriCorps was effectively cut in half.
• France shows a democracy can decide to make service universal (#29).

Remedy. Restore and grow AmeriCorps to 250,000 full-time slots over five years, and write statutory protection against impoundment-style terminations.

Tradeoff I concede. Expansion costs real money. Some AmeriCorps programs are mediocre, and a slot count is not an outcome.

Falsifiable prediction. If I'm right, counties with high AmeriCorps presence should show slower declines in volunteering and local trust than matched counties. The April 2025 terminations are a natural experiment: affected localities should show measurable drops in formal volunteering in CEV 2025 relative to unaffected ones.

---

Prof. Rosa Delgado-Finch (#7, political economy, Soc): trust is a material relation, so fix the material relations

Causal claim. People trust others and institutions when institutions deliver for them. The trust gradient runs almost perfectly by income: about 50% of those earning $150k+ say most people can be trusted versus about 20% of those under $30k (Brief #11). That is not a civics-literacy gap. It is the lived experience of precarity.

• The decline in trust since the 1970s (46% → 34%) overlaps the collapse of private-sector unions. Unions were the largest cross-racial, cross-religious membership organizations working people had.
• Civics classes will not replace a shop floor.

Remedy.

• A paid public service option at a living wage, not a stipend. It should be open to any adult, not only the young, and built on civil-service norms.
• Pair it with labor-law reform so that workplaces become civic institutions again.

Tradeoff I concede. A wage-level service program costs several times what AmeriCorps stipends do. It also risks becoming a jobs program judged on jobs, not cohesion.

Falsifiable prediction. Within-person panel data should show trust rising after job-security gains, such as unionization or a transition to stable public employment. It should not rise after civics exposure.

---

Prof. Nathaniel Cross (#64, religious studies, R): the civic crisis is a congregational crisis

Causal claim. America's civic life was never primarily run from Washington. It ran through congregations, fraternal lodges and families.

• Church membership fell from 70% in 1999 to 47% in 2020 (#12).
• Adults with no close friends quadrupled from 3% to 12% (#14).
• In-person time with friends fell by two-thirds (#13).
• The collapse of congregational membership is the main driver of the loneliness epidemic. You cannot replace that with a federal grant program.
• Federal service has its own record: the agency cannot produce auditable financial statements (#19).

Remedy.

• Remove federal obstacles to civil society.
• Protect religious organizations' eligibility for partnerships on neutral terms.
• Expand the charitable deduction to non-itemizers.
• Let localities, not agencies, set the civic agenda.

Tradeoff I concede. Religious congregations can bond more than they bridge. Congregational decline is driven largely by belief change, which policy cannot and should not engineer.

Falsifiable prediction. Across counties, congregational density should predict volunteering and interpersonal trust more strongly than AmeriCorps presence, after controlling for income and education.

---

Prof. Victor Lindqvist (#37, libertarian legal theory, Lib): the state has no business manufacturing citizens

Causal claim. Trust in government is low largely because government has earned it. Pew's measure has been under 30% since 2007 (#9) through administrations of both parties. The solution is a smaller government people can trust, not a bigger one teaching them to.

Evidence.

• AmeriCorps has received eight consecutive audit disclaimers and carries 11 material weaknesses (#19). An agency that cannot account for its own money does not merit expansion.
• Compulsory service is conscription by another name. The Vietnam-era draft lottery literature shows draftees suffered roughly 15% lower earnings for years afterward (Angrist 1990).
• France's SNU could not even achieve mixing among volunteers (#29).

Remedy.

• No compulsory service, federal or state.
• Let AmeriCorps sunset or become a block grant to states.
• Remove regulatory barriers to voluntary association, such as nonprofit compliance costs and occupational licensing that keeps volunteers out.

Tradeoff I concede. Private civil society has not arrested the decline either. Volunteering was 28.3% in 2023 (#15), and I cannot show that shrinking the state raises it.

Falsifiable prediction. States that expand state-run service corps (e.g., the California and Maryland programs) will show no differential gain in CEV volunteering or GSS-type trust over five years relative to comparison states.

Stage 2: OPENING POSITIONS (Part 2 of 2)

---

Prof. Kevin Marsh (#48, education policy, R): teach the content, and stop pretending protest projects are civics

Causal claim. A republic needs citizens who know how it works. The NAEP data are an alarm. In 2022 grade 8 civics saw its first decline ever, and only 22% were Proficient (#1). U.S. history lost 5 points, with 40% below Basic (#2). Civic knowledge is in freefall. "Action civics" crowds out content with activism assignments that teachers pick.

Remedy.

• Knowledge-rich civics and U.S. history sequences in K-12.
• A NAEP civics assessment at grades 4, 8 and 12 on a regular cycle.
• State adoption of the citizenship-test requirement for graduation.

Tradeoff I concede. Knowledge may not move turnout. The Jung & Gopalan null on test mandates (#6) is a real problem for anyone who sells tests as a participation lever. I defend knowledge on its own terms.

Falsifiable prediction. States with strong content standards and a required civics course should show higher NAEP civics scores (when state-level results exist) and higher Annenberg-style adult knowledge a decade later.

---

Prof. Ruth Abernathy (#51, teacher education, C-L): knowledge is necessary; practice is what makes citizens

Causal claim. Kevin is half right. The evidence says knowledge alone does not move behavior: test mandates show a precise null on youth turnout (#6), and ordinary civics coursework shows near-null effects with sibling fixed effects (#5). Holbein & Hillygus locate the gap in follow-through skills (#7). Action civics closes that gap and has been shown to raise youth turnout.

Remedy.

• Required civics with a project-based component in which students research a local problem and engage officials.
• Pair it with pre-registration and 16-year-old pre-registration drives.

Tradeoff I concede. Project choice can be politicized, so I would accept viewpoint-neutral guardrails such as student-chosen, non-partisan and procedural topics.

Falsifiable prediction. The I-ACTED RCT (#8) should show positive effects on civic engagement at 12 months. If it shows nulls, my case weakens substantially.

---

Prof. Leila Farouk (#40, political psychology, C): we know how to reduce animosity for about two weeks

Causal claim. The strongest experimental evidence we have on depolarization is Voelkel et al. (#26). 23 of 25 interventions reduced partisan animosity, by up to about 10.5 pp. The most effective ones used sympathetic cross-partisans and shared identities. Scale these through media and schools, and antidemocratic attitudes will fall with them. Contact literature points the same way (#23–25).

Remedy. A federally funded "what works" portfolio for depolarization and contact interventions. Durability testing at 6 and 12 months would be a condition of funding.

Tradeoff I concede. Effects decay: about 32% remains at two weeks. Online interventions are not civic life.

Falsifiable prediction. Repeated-dose interventions should retain more than 50% of the effect at 3 months. If they do not, the online-intervention approach is a dead end.

---

Prof. Dmitri Volkov (#33, national security, C-R): mixing is the mechanism, and service is only the container

Causal claim. The best causal evidence in this domain says what matters is random out-of-region assignment to a collaborative team for months. Nothing else in the literature reliably does this.

• In Spain, out-of-region conscripts gained trust that lasted decades, and men from separatist regions identified more with Spain (#27).
• In Nigeria, NYSC assignment raised national pride (#28).
• In India, collaborative teams built friendships while adversarial matches did not (#25).
• Israel's IDF is the proven example of a service institution that forges a common civic identity from a fractured immigrant society.

Remedy. A voluntary but well-incentivized national service track (civilian and military) that assigns participants across regions by lottery into mixed teams. Education benefits would be large enough to draw from every class. Evaluation would be built in.

Tradeoff I concede. Voluntary programs self-select, as France's did (#29). Without strong incentives we get the already-civic children of uniformed parents.

Falsifiable prediction. In a lottery-assigned U.S. pilot, out-of-region placement should raise trust toward the host region and cross-partisan warmth at 2 years, measurably above in-region placement.

Stage 3: CROSS-EXAMINATION (Part 1): Lindqvist v. Kerr on AmeriCorps, four rounds

Lindqvist (Lib) → Kerr, Round 1.
Tamsin, three problems.

• "Effectively cut in half" is wrong. The President asked for $107.7M. The House committee proposed $642.95M. What Congress enacted for FY26 was $1,253,806,000, against $1.262B in FY25 (Brief #17). That is a 0.7% cut. The April 2025 terminations were real, but a federal court enjoined them in 24 states + DC within six weeks (#16), and the Sept 2026 settlement imposes 30-day notice and a full-obligation commitment (#18). The story of 2025–26 is not that AmeriCorps was halved. It is that Congress and the courts refused to halve it.
• Your $3.95. It comes from a report commissioned by the program's own advocacy coalition (#20). It is a benefit-transfer model: it assumes earnings gains and values output at wage rates. It is not an impact estimate.
• You want to hand a quarter-million slots to an agency that has not produced auditable financial statements for eight straight years (#19). Would you give a private charity with that record a 5x budget increase?

Kerr (C-L), Response 1.

• On "cut in half": I'll take the hit on wording. I should have said proposed cut. The enacted level is near-flat. The operational damage was real, though. Over 40% of grantees were terminated on a Friday evening (#16). Programs at the end of grant cycles such as PowerCorps were not renewed, and 85% of agency staff were gone. Appropriations did not measure what happened on the ground in 2025.
• On Belfield: fair that it's commissioned. But the relevant causal literature is not zero. Bagues & Roth and Okunogbe (#27, #28) are service programs with random assignment, and they show durable trust effects.
• On audits: that is an argument for fixing the agency's financial management, which is overhead, not for starving the grantees who deliver the service.

Lindqvist, Round 2.
Your causal evidence is Spanish conscription and Nigerian mandatory service. Both involve random assignment across regions, and neither resembles AmeriCorps. AmeriCorps members mostly serve locally, often in their own communities, in programs they chose. You are borrowing the identification of compulsory programs to justify a voluntary program that lacks the mechanism.

• Name one randomized or quasi-experimental estimate of AmeriCorps' effect on members' trust or on community trust. The brief found none (#22 and Contested iv).
• On audits, 11 material weaknesses, three of them dating to FY2017, is not overhead trivia. It means the agency cannot tell Congress where grant money went.

Kerr, Response 2.

• You're right that AmeriCorps as currently designed is not the Bagues-Roth treatment. That is a design critique, and I accept it. It moves me toward Dmitri's version: cross-regional, team-based, lottery-assigned.
• I cannot name a randomized estimate of AmeriCorps' effect on trust. The long-term member studies I know from the 2000s are quasi-experimental with comparison groups of applicants who didn't enroll, and I could not re-verify their effect sizes for this thread. So I will not rest on them.
• I will not concede the budget. Near-flat funding after an 18-month assault on the program is not an equilibrium. Grantees face three-year cycles, and the chilling effect on applications is real.

Lindqvist, Round 3.
"The chilling effect on applications is real" is a claim without a number. And you've now conceded:

1. The program's design lacks the causal mechanism you cited.
2. There is no causal estimate for the program itself.
3. The headline cut didn't happen.

What's left is "AmeriCorps provides services communities value." That may be true, but it is a service-delivery argument to be judged against other uses of $1.25B. It is not a trust argument.

Kerr, Response 3.

• I accept points 1 and 2 as stated.
• On point 3 I've already amended.
• I reject your conclusion that AmeriCorps is only a service-delivery program. The Mettler evidence (#21) is not randomized, but it is consistent with a policy-feedback mechanism: citizens treated well by public programs participate more. Your own opening concedes private civil society has not arrested the decline either.
• My amended position is to hold AmeriCorps at FY26 levels with statutory guardrails and fixed audits. The expansion money should go to a randomized, cross-regional pilot. If that pilot shows nothing at 2 years, I'll stop arguing for scale.

Lindqvist, final.
That is a position I can vote against honestly rather than dismiss. I'd still sunset the agency, but I'd support a pilot that is honestly evaluated, if it is voluntary and the evaluation is independent of the grantee coalition.

---

Castellano (#1, labor econ, C-L) → Lindqvist on conscription earnings.
Victor, your "roughly 15% lower earnings for draftees" comes from Angrist's use of the Vietnam draft lottery. As I recall the paper, the estimate applies to white veterans who served during a war, and it identifies the cost of lost civilian labor-market experience during wartime service. It says nothing about a 10-month civilian service year for volunteers with an education award. The referees also could not fetch and verify the figure for this thread. You are importing a wartime-conscription penalty into a voluntary-civilian debate.

Lindqvist, response.
I cited it against compulsory service, which remains on the table if S1's framing is taken seriously. But I accept that it is the wrong comparator for voluntary civilian service, and that the magnitude is unverified here. I'll restate the point: forced diversion of a year from education or work has an opportunity cost that proponents never score. France's own auditors put generalized SNU at €3.5–5bn a year for a single cohort (#29).

---

Kerr (C-L) → Cross (R).
Nathaniel, nothing in the brief supports "the collapse of congregational membership is the main driver of the loneliness epidemic." The numbers you cite are co-trends: church membership down, friendships down, time with friends down. Screens, remote work, suburban form and delayed marriage all moved at the same time. The Surgeon General's advisory attributes loneliness to many factors, not mainly to religion. And congregations bond: the GSS/Pew trust gap by race (White 40% vs Black 21%, #11) coexists with Black Americans' historically high church attendance. Congregations have not closed that gap.

Cross, response.

• I overstated. "Main driver" is not something the brief can support, and I withdraw it. My claim is narrower: congregations are the largest remaining associational infrastructure, and membership fell 23 points in two decades (#12). On any Putnam-style reading, that is a first-order loss of social capital.
• On the race point: it is a non sequitur. It shows congregations do not erase inequality-driven distrust, which I never claimed.
• My policy claim stands without the causal overreach. Government should not crowd out, disqualify or tax-disadvantage the associations that still function.

Kerr, reply.
Accepted as narrowed. But "don't crowd out" needs evidence of crowding out. Can you name a program where federal service displaced congregational volunteering?

Cross, reply.
Not from this brief's evidence, no. I'll put it as a hypothesis, not a finding. My county-level prediction stands as the test.

Stage 3: CROSS-EXAMINATION (Part 2)

---

Abigail Stroud (#63, family sociology, R) → Delgado-Finch (Soc).
Rosa, the income gradient in interpersonal trust (#11) is cross-sectional. Education shows an equally steep gradient: master's-degree holders are about 52% trusting versus about 25% for high-school graduates. You cannot tell from that table whether income causes trust, trust causes income, or both come from family and neighborhood.

• Trust in government is not a material-security story either. It flips with the party in power: Republicans 26%, Democrats 9% under the current administration (#9). That is partisan identity, not paychecks.
• On unions, the brief contains no identified estimate linking union decline to trust decline. That's a co-trend like Nathaniel's.

Delgado-Finch, response.

• On government trust: conceded. Partisan flipping dominates short-run movement.
• On interpersonal trust, the direction question is fair, and the brief doesn't settle it. I'll mark union decline as "Unsupported in this review", not "Wrong". There's a literature on union membership and civic participation that the referees didn't pull, and I won't cite effect sizes I can't show.
• I keep the structural point. The age gradient (18–29 at 26%, #11) and the friendship collapse (#14) hit hardest the groups with the least job security. That is consistent with my mechanism, though I grant it is not proof.

Stroud, Round 2.
"Consistent with" also describes Nathaniel's story, my family-structure story and the smartphone story. The brief's Contested (v) section says no identified study ranks them. We should all be putting that in our remedies instead of our certainties.

Delgado-Finch.
Agreed on the epistemics. I'll move the paid-service proposal into "pilot and measure" too, but with wages paid at a level that doesn't recruit only the already-comfortable. France's 46% uniformed-parent skew (#29) is what low compensation plus volunteering gets you.

---

Abernathy (C-L) → Marsh (R) on "freefall".
Kevin:

• A 2-point decline in grade-8 NAEP civics after 24 years (#1), during a pandemic that took several points off every NAEP subject, is not "freefall."
• Adult knowledge on Annenberg's measure has more than doubled on the three-branches item since 2018, from 32% to 70% (#3).
• Your own citation (#6) shows the citizenship-test mandate you propose has a precise null on turnout.

Marsh, response.

• "Freefall" is too strong for civics. I'll amend to "declining from a low base." 22% Proficient is still the headline. U.S. history's 5-point drop and 40% below Basic (#2) is the steeper problem, and it predates COVID (decline since 2014).
• On Annenberg: an adult phone/web panel on a handful of items is not NAEP, and the 2024–25 jump may reflect salience in an election year. I'll call it contested, not dispositive.
• On the test mandate: I said in my opening that I defend it as a knowledge floor, not a turnout lever. I'll drop the claim that the mandate improves civic participation, since I didn't make it and the evidence says it doesn't.

Marsh (R) → Abernathy (C-L), counter-challenge.
Ruth, you wrote that action civics "has been shown to raise youth turnout." Where?

• The only U.S. cluster RCT (#8) is a protocol with no results.
• Weinschenk & Dawes specifically tested service-learning and civic-skills coursework and found "no support" for turnout effects (#5).
• You are asking states to mandate a pedagogy whose first real test hasn't reported.

Abernathy, response.

• You're right. I overstated. The action-civics evidence I had in mind is pre/post and quasi-experimental work on outcomes like civic self-efficacy, not turnout. I cannot cite a causal turnout effect. I retract "shown to raise youth turnout."
• My amended claim: action civics is promising and untested. Mandating it before I-ACTED reports would be premature. Holbein & Hillygus (#7) still point to follow-through skills as the missing piece, and knowledge-only instruction has been tested and failed on behavior (#5, #6).

Marsh.
Then we agree on the most important thing: nobody in this room has causal evidence that any school civics intervention raises turnout. That is precisely what Symposium 1 asserted without testing.

---

Miles Hendricks (#99, game theory, Lib) → Farouk (C).
Leila, your opening says antidemocratic attitudes "will fall" as animosity falls. Your own source says the opposite. Voelkel et al. report 23/25 treatments cut animosity, but only 6 moved support for undemocratic practices and 5 moved support for partisan violence. They call the constructs "clearly distinct" (#26). Two-week retention on undemocratic practices is about 7%.

• You would be funding a federal portfolio built on a relationship the flagship study rejects.
• A government-funded "depolarization" program also invites the obvious question: whose definition of "antidemocratic"?

Farouk, response.

• Conceded on the linkage. I misstated the paper. The finding is that animosity reduction does not automatically carry to antidemocratic attitudes. I retract "will fall with them."
• My remedy survives in narrower form. Fund evaluation, not deployment. Require pre-registered 6- and 12-month durability outcomes. Outcomes should be behavioral where possible, such as Mousa-style prize votes and team registration (#24), not only survey thermometers.
• On "whose definition": the Voelkel outcome measures are items like support for ignoring court rulings or for partisan violence, and both parties' respondents are measured symmetrically. I would require that neutrality in any funded work.

Hendricks, Round 2.
Behavioral outcomes are better. But Mousa's behavioral effects were toward teammates, and nothing moved toward the outgroup at large (#24). Paluck et al. found no RCT on adult racial prejudice past 25 and shrinking effects in larger studies (#23). The honest summary is that contact reliably changes how you treat the specific people you worked with. Whether it generalizes is unknown.

Farouk.
That's a fair summary of Mousa and Paluck. Bagues & Roth (#27) is the counterexample. Trust generalized to the region, not just to bunkmates, and lasted decades. The difference is dose: months of full-time collaborative living versus a 10-week league or a 10-minute video. Dose is the research question.

---

Simone Achebe (#28, comparative politics, C-L) → Volkov (C-R).
Dmitri, the brief contains no evidence at all on Israel, yet you called the IDF "the proven example" of forging common civic identity.

• Israel's service model has long exempted or excluded large groups, most prominently most Arab citizens and, historically, many ultra-Orthodox men. That exemption is itself a major source of civic conflict there. [This is widely reported; it was not verified in this review.]
• "Proven" by what design?
• And your own Nigerian source cuts both ways. Okunogbe finds that NYSC immersion strengthened ethnic pride alongside national pride (#28).

Volkov, response.

• On Israel: withdrawn as "proven." I have no identified estimate to offer, and the exemption point is exactly why a selective national institution can divide as well as unite. It's an illustration, not evidence.
• On Nigeria: I read it as complementary identities, not a contradiction. National pride and knowledge rose. The fact that ethnic pride also rose is a warning against expecting service to erase subgroup identity. It is not a finding that service divides.
• My core claim rests on Spain and India (#25, #27), where the causal designs are strongest. I stand on those two.

Achebe, Round 2.
Spain is a compulsory, male, 1980s–90s program in a country with regional rather than racial or partisan cleavages. What is the transport assumption to a voluntary U.S. program where the salient divide is partisan and self-sorted?

Volkov.
The transport assumption is that months of interdependent teamwork with randomly assigned strangers builds trust toward their group, and that collaborative rather than adversarial structure is required (#25). Whether it transports to partisan cleavages is unknown. That is why I propose a lottery-assigned pilot with evaluation, not a national rollout. France shows what rolling out without evaluation gets you (#29).

---

Takahashi (referee) note on Kerr's France citation. In her opening, Prof. Kerr cited France as showing a democracy "can decide to make service universal." The Cour des comptes report (#29) shows the opposite for France's execution: 40,000 participants against an 800,000 target, a 45% cost overrun per participant and failed social mixing. The referees will rate it in Stage 4.

Kerr, response. I'll amend: France shows a democracy can decide on universal service. It also shows the decision is the easy part. The SNU is evidence against rushing universality.

Stage 4: REFEREE FACT-CHECK

Brandt (#57), Takahashi (#17), Farouk (#40, recused on her own claims; Brandt and Takahashi rated F1–F4 alone)

Scale: Supported means the source confirms the claim. Contested means credible sources disagree. Unsupported means no adequate source was found, or the source does not say it; this includes claims we could not verify. Wrong means a credible source contradicts the claim. Brief item numbers (#) point to the Stage 1 sources.

# · Speaker · Claim · Rating · Basis
K1 · Kerr · GI Bill users had ~50% more civic memberships and ~30% more political activity · Supported (as a descriptive finding) · Mettler 2005 (#21). The source confirms the numbers. The design is a retrospective survey.
K2 · Kerr · National service returns $3.95 per $1 · Contested · Belfield 2013 (#20) says this. It is a commissioned benefit-transfer model with no impact evaluation behind it, and no independent estimate was found.
K3 · Kerr · Cross-region mixing builds trust lasting decades · Supported · Bagues & Roth 2023 (#27)
K4 · Kerr · ~$400M in grants terminated; ~85% of staff gone · Supported · NCAN; NPQ (#16)
K5 · Kerr · "AmeriCorps was effectively cut in half" · Wrong · FY26 enacted $1.2538B vs FY25 $1.262B (#17). The −49% was a House committee proposal. Retracted and amended in Stage 3.
K6 · Kerr · France shows universal service is achievable · Wrong (as characterized) · Cour des comptes 2024 (#29): 40k participants vs an 800k target, a 45% cost overrun per participant, failed mixing. Amended in Stage 3.
K7 · Kerr · >40% of grantees terminated Apr 25; some end-of-cycle programs not renewed · Supported · NPQ (#16)
K8 · Kerr · "Chilling effect on applications is real" · Unsupported · No application data found
K9 · Kerr · GI Bill shows a causal policy-feedback effect · Contested · Mettler's interpretation. Self-selection into benefit use is unaddressed by the design (H-Net review).
D1 · Delgado-Finch · Trust ~50% at $150k+ vs ~20% under $30k · Supported · Pew 2025 (#11)
D2 · Delgado-Finch · Interpersonal trust fell 46% → 34% · Supported · GSS via Pew (#11)
D3 · Delgado-Finch · Union decline explains trust decline · Unsupported · No identified estimate in this review. Downgraded by speaker in Stage 3.
D4 · Delgado-Finch · 46% of 2023 SNU volunteers had uniformed-service parents · Supported · Cour des comptes (#29)
C1 · Cross · Church membership 70% (1999) → 47% (2020) · Supported · Gallup (#12)
C2 · Cross · No-close-friends share quadrupled, 3% → 12% · Supported · AEI 2021 (#14)
C3 · Cross · In-person time with friends fell by about two-thirds · Supported · Surgeon General 2023 (#13): 60 → 20 min/day
C4 · Cross · Congregational decline is "the main driver" of loneliness · Unsupported · Co-trend only. The Surgeon General's advisory is multi-causal. Retracted.
C5 · Cross · AmeriCorps cannot produce auditable statements · Supported · OIG FY2024 disclaimer (#19)
L1 · Lindqvist · Trust in government <30% since 2007 · Supported · Pew 2025 (#9)
L2 · Lindqvist · 8 consecutive audit disclaimers; 11 material weaknesses · Supported · OIG (#19)
L3 · Lindqvist · Vietnam draftees had ~15% lower earnings (Angrist 1990) · Unsupported · Referees could not retrieve and verify the figure. Castellano also noted it is a veteran/wartime estimate, the wrong comparator. Amended.
L4 · Lindqvist · SNU failed to achieve mixing · Supported · Cour des comptes (#29)
L5 · Lindqvist · Formal volunteering 28.3% (2023) · Supported · CEV 2023 (#15)
L6 · Lindqvist · FY26 enacted ≈ 0.7% below FY25 · Supported · $1.2538B / $1.262B = −0.65% (#17)
L7 · Lindqvist · Terminations enjoined within ~six weeks · Supported · Apr 25 → June 5, 2025 (#16)
L8 · Lindqvist · Belfield is advocacy-commissioned, benefit-transfer · Supported · Report front matter (#20)
L9 · Lindqvist · Generalized SNU costs €3.5–5bn/yr · Supported · Cour des comptes (#29)
M1 · Marsh · NAEP 2022 civics: first decline; 22% Proficient · Supported · NAGB (#1)
M2 · Marsh · "Civic knowledge is in freefall" · Wrong · A 2-pt NAEP drop plus a rising adult Annenberg measure (#1, #3) contradict "freefall." Retracted/amended.
M3 · Marsh · U.S. history −5; 40% below Basic; decline since 2014 · Supported · NAGB (#2)
M4 · Marsh · Annenberg rise may reflect election-year salience · Unsupported · Speculative. The 70% reading was taken in Aug 2025, a non-election year, and it rose further than 2024's. This cuts against the explanation.
M5 · Abernathy · Pandemic "took several points off every NAEP subject" · Unsupported · Not verified in this review. Plausible, but other-subject NAEP results were not fetched.
A1 · Abernathy · Knowledge-only civics fails to move turnout · Supported · Weinschenk & Dawes (#5); Jung & Gopalan (#6)
A2 · Abernathy · Action civics "has been shown to raise youth turnout" · Unsupported · The only RCT is a protocol (#8), and W&D found no service-learning effect. Retracted.
A3 · Abernathy · Adult three-branches knowledge 32% → 70% · Supported · Annenberg (#3)
F1 · Farouk · 23/25 interventions cut animosity, up to ~10.5 pp · Supported · Voelkel 2024 (#26)
F2 · Farouk · Antidemocratic attitudes "will fall with" animosity · Wrong · Voelkel: only 6 and 5 treatments moved antidemocratic outcomes, and the constructs are "clearly distinct" (#26). Retracted.
F3 · Farouk · ~32% of the effect retained at two weeks · Supported · Voelkel (#26)
F4 · Farouk · Voelkel outcome items are measured symmetrically across parties · Unsupported · Plausible from design, but not verified against the paper's instrument in this review
H1 · Hendricks · 6 and 5 treatments; ~7% retention on undemocratic practices · Supported · Voelkel (#26)
H2 · Hendricks · Mousa: effects toward teammates, not the outgroup at large · Supported · J-PAL summary (#24)
H3 · Hendricks · No contact RCT on racial prejudice in adults >25; shrinking effects in larger studies · Supported · Paluck et al. 2019 (#23)
V1 · Volkov · Spain: separatist-region conscripts identified more with Spain · Supported · Bagues & Roth (#27)
V2 · Volkov · NYSC raised national pride · Supported · Okunogbe 2024 (#28)
V3 · Volkov · Collaborative vs adversarial contact differ · Supported · Lowe 2021 (#25)
V4 · Volkov · IDF is "the proven example" of forging civic identity · Unsupported · No evidence in this review. Retracted.
AC1 · Achebe · Israel exempts or excludes most Arab citizens and many ultra-Orthodox · Unsupported (in this review) · Widely reported but not verified here. Speaker flagged it herself.
AC2 · Achebe · NYSC also strengthened ethnic pride · Supported · Okunogbe (#28)
S1 · Stroud · Trust ~52% (master's) vs ~25% (HS) · Supported · Pew 2025 (#11)
S2 · Stroud · Government trust is partisan: R 26% vs D 9% now · Supported (snapshot) · Pew 2025 (#9). The longer "flip" pattern is consistent with Pew's series but was not separately verified.
T1 · Takahashi · SNU per-participant cost 45% over budget · Supported · €2,900 vs €2,000 (#29)

Tally (51 claims): Supported 35 · Contested 2 · Unsupported 10 · Wrong 4.

Referee comments.

1. Every "Wrong" rating fell on a claim that overstated the case for the speaker's own remedy: "cut in half," "freefall," "will fall with them," "France shows it's achievable." Each bloc overstated at least once, and each speaker retracted on the record.
2. The four causal retractions (C4, D3, A2, V4) share one pattern: a co-trend or a vivid example was presented as a mechanism. Contested (v) in the brief stands. No one has identified the cause of the U.S. trust decline.
3. Farouk recused on F1–F4, and Brandt and Takahashi rated them without her. F2 is rated Wrong on the plain text of the paper she herself put in the brief.

On-record amendments requested and received: Kerr (K5, K6), Cross (C4), Lindqvist (L3), Marsh (M2), Abernathy (A2), Farouk (F2), Volkov (V4), Delgado-Finch (D3). Marsh is asked to withdraw M4 or supply evidence.

Marsh: Withdrawn. I didn't check the survey date. Point to the referees.

Stage 5: STEELMAN

---

Left (Kerr) steelmans the libertarian and communitarian Right

"The strongest case against a federal national-service push:

• The one program we have has not earned expansion. AmeriCorps has had eight consecutive audit disclaimers and 11 material weaknesses (#19). No causal estimate exists of its effect on anyone's trust (#22).
• The best causal evidence comes from compulsory programs (Spain, Nigeria). Their mechanism is random cross-regional assignment. Adopting it in the U.S. would mean either compulsion or a lottery the volunteers didn't choose, and both are real impositions on liberty.
• The one democracy that tried universal service recently, France, blew its budget and failed to mix classes (#29).
• Civil society is where trust lives: congregations, clubs, family. Those institutions have lost 23 points of membership in two decades (#12), for reasons Washington can't reverse and shouldn't try to. A state that tries to produce civic sentiment risks producing either nothing (Voelkel's two-week decay, #26) or indoctrination.
• The humble remedy is to stop damaging what works, measure honestly and let people associate freely."

Lindqvist (Lib): Accepted as fair. I'd add one line: trust in government is low partly because government's performance is low, and a program that can't pass an audit is an odd vehicle for restoring it.
Cross (R): Accepted, with one correction. The communitarian right does not oppose all public roles. We favor local and state vehicles and neutral treatment of faith-based partners. "Don't try" is Victor's position, not mine.

---

Right (Cross, with Volkov) steelmans the progressive Left

"The strongest case for public investment in national service:

• The civic decline is real and multi-dimensional. Interpersonal trust is down 12 points (#11), friendships are collapsing (#14), and trust in government is at 17% (#9).
• Private civil society has not reversed it despite decades of freedom to try. Volunteering was only 28.3% even in its best recorded year (#15).
• The strongest causal evidence says a specific public design works: months of collaborative, cross-group service with random assignment produces trust that lasts decades (#27) and national identity that coexists with subgroup identity (#28). Markets and congregations do not randomly assign strangers to shared tasks. Only public institutions have done that at scale.
• The 2025 terminations show the risk of leaving these institutions to executive discretion. Courts found APA violations (#16), which argues for statutory durability.
• Service at a real wage also addresses the income-trust gradient (#11), since otherwise only the comfortable can afford to serve, which is exactly the SNU failure (#29)."

Kerr (C-L): Accepted as fair, and better stated than my opening.
Delgado-Finch (Soc): Accepted, with one correction. The left's case isn't only service. It's that material security and workplace institutions are civic infrastructure. The steelman frames it as a service-design argument, and ours is broader.

---

Center (Farouk) steelmans both, and names the shared blind spot

"Both blocs are right about the other side's weakest evidence:

• The left overstated what AmeriCorps is known to do.
• The right overstated what congregational decline is known to cause.

What the center adds is a warning. **Symposium 1's 'civics education plus voluntary national service' combined the two interventions with the weakest behavioral evidence:**

• knowledge-civics, which is null on turnout (#5, #6);
• voluntary, local service, which has no causal estimate and self-selects (#22, #29).

It then omitted the one design with strong evidence: random cross-group assignment to collaborative work (#25, #27). The strong-evidence version is also the most liberty-constraining, and that is the real tradeoff."

Lindqvist: Accepted as fair, and it is the most uncomfortable paragraph in the thread for me.
Kerr: Accepted.
Marsh: Accepted with one correction. Knowledge-civics is justified as knowledge. It failed only as a turnout lever, which is not a test it should have been sold on.

Stage 6: REVISED PROPOSALS

No proposal in this domain has a CBO or JCT score. Where a cost is illustrative it is labeled that way, and referees flag the assumption.

---

16-P1 · AmeriCorps stabilization with guardrails and an audit condition

Mechanism.

• Authorize AmeriCorps at roughly the FY26 enacted level ($1.254B) for three years.
• Codify the settlement's terms (#18): 30-day notice before grant terminations, and obligation of appropriated funds by fiscal year-end.
• Condition any funding above that level on the OIG issuing an unmodified audit opinion.
• Require grantees to report standardized outcome measures.

Cost. Roughly flat against the FY26 baseline. There is no official score.
Precedent. The 2025 terminations (#16) were enjoined under the APA, and the 2026 settlement (#18) imposes similar notice terms by agreement rather than statute.
Key risk.

• It locks in a program with no causal evidence on civic outcomes (#22).
• The audit condition could be unattainable in the near term, given 11 recurring material weaknesses (#19).

---

16-P2 · Cross-Regional Service Lottery Pilot (randomized, voluntary)

Mechanism.

• Offer about 20,000 voluntary full-time service slots a year for three cohorts, civilian and optionally military-affiliated.
• Oversubscribed applicants are lottery-assigned between in-region and out-of-region placements, on mixed collaborative teams.
• Pay a stipend plus an education award large enough to recruit across income levels.
• An independent evaluator with pre-registered outcomes measures trust toward the host region, cross-partisan warmth, and behavioral measures (Mousa-style) at 1, 2 and 5 years.

Cost. No official score. The referees' illustrative range is $0.5–1B over three years, assuming per-slot costs comparable to a full-time AmeriCorps position [unverified: no verified per-member figure, #22].
Precedent.

• Spain's conscription lottery produced decades-long trust and identity effects (#27).
• Nigeria's NYSC raised national pride but also ethnic pride (#28).
• In India, collaborative team contact worked and adversarial contact backfired (#25).

Key risk.

• Transport: those were compulsory programs across regional or ethnic cleavages, not voluntary programs across U.S. partisan cleavages.
• Self-selection among volunteers (#29).

---

16-P3 · Compulsory universal national service (testing Symposium 1's implied endpoint)

Mechanism. Every U.S. resident aged 18–25 would do 12 months of civilian or military service, with exemptions for disability and hardship.
Cost. No official score. France's auditors estimated €3.5–5bn a year for about 800,000 youth in a short program, plus about €6bn in infrastructure (#29). A U.S. cohort is several times larger, and a 12-month program is far longer, so the cost would plausibly run to tens of billions per year [illustrative extrapolation, unverified].
Precedent.

• France's SNU missed its targets, overran costs 45% per participant and failed to mix classes (#29).
• Spain's compulsory service had positive identity effects (#27) but was abolished.
• Israel was offered in debate, but the claim was retracted (V4).

Key risk.

• Liberty and conscription objections.
• Large opportunity costs.
• Implementation failure at scale (#29).

---

16-P4 · Knowledge-rich civics and U.S. history, with assessment continuity (no turnout claims)

Mechanism.

• Federal support, through existing ESSA and American History & Civics grants, for states adopting content-rich civics and U.S. history sequences.
• Maintain the NAEP civics and U.S. history assessments on a regular cycle, and seek state-level reporting.
• Grants may not be justified on turnout, and citizenship-test graduation mandates get no federal incentive, given the null in #6.

Cost. Small, within existing appropriations. No official score.
Precedent. The 15 states that adopted CEI test mandates in 2015–18 saw a precise null on turnout (#6). NAEP has tracked civics since 1998 (#1).
Key risk. It moves knowledge, which is itself contested (#1 vs #3), but there is no evidence it moves trust or participation (#5). Content choices invite culture-war fights over curriculum.

---

16-P5 · Civic Evidence Fund (pre-registered, durable, behavioral)

Mechanism.

• An IES and NSF competitive fund of about $50M/yr [illustrative] for RCTs of civic education (including action civics), contact and depolarization interventions.
• Required elements: pre-registration; outcomes at 12 months or more; behavioral measures where feasible; outcome measures symmetric across parties; public data.

Cost. About $50M/yr [illustrative]. No official score.
Precedent.

• Voelkel's megastudy (#26) found large short-run effects and heavy decay.
• The I-ACTED trial (#8) has not yet reported.
• Paluck et al. (#23) document the gap: no RCTs on adults over 25.

Key risk. Research outputs arrive in 3–5 years. Funded researchers' political skew could delegitimize findings unless measures are audited for symmetry.

---

16-P6 · Action civics requirement in federal civics grants

Mechanism. Federal civics grant recipients must include a project-based component in which students research and act on a local issue, with viewpoint-neutral guardrails.
Cost. Within existing grants. No official score.
Precedent. Weinschenk & Dawes found no turnout effect from service-learning and civic-skills coursework (#5). There is no RCT result yet (#8).
Key risk.

• It mandates a pedagogy before its first rigorous test reports.
• Politicization of project choice.

---

16-P7 · Neutral civil-society match

Mechanism.

• A federal 1:1 match, capped per organization, for local organizations that mobilize volunteers for secular public services (tutoring, food, eldercare).
• Religious organizations are eligible on neutral terms, with no religious activity funded.
• Administration is block-granted through state service commissions.

Cost. Scalable. The referees illustrate $250M/yr [illustrative]. No official score.
Precedent. Charitable-choice and faith-based partnership rules since the late 1990s [not evaluated in this review]. Congregational membership fell from 70% to 47% (#12).
Key risk.

• No evidence that matching grants raise volunteering or trust.
• Establishment Clause litigation.
• It could subsidize the already-civic (bonding, not bridging).

Stage 7: RECORDED VOTE

Provisional vote. It is superseded by the blind independent ballot in the thread "Independent Ballot".

Blocs: Left 34 · Center 33 · Right 32 (moderator #100 abstains). Consensus requires at least 60% in each bloc: Left ≥21, Center ≥20, Right ≥20.

Proposal · Left yes · Center yes · Right yes · Consensus?
16-P1 AmeriCorps stabilization + guardrails + audit condition · 32/34 (94.1%) · 27/33 (81.8%) · 14/32 (43.8%) · No (Right short)
16-P2 Cross-Regional Service Lottery Pilot · 29/34 (85.3%) · 28/33 (84.8%) · 21/32 (65.6%) · YES
16-P3 Compulsory universal national service · 9/34 (26.5%) · 5/33 (15.2%) · 6/32 (18.8%) · No
16-P4 Knowledge-rich civics, assessment continuity, no turnout claims · 23/34 (67.6%) · 29/33 (87.9%) · 27/32 (84.4%) · YES
16-P5 Civic Evidence Fund · 30/34 (88.2%) · 31/33 (93.9%) · 22/32 (68.8%) · YES
16-P6 Action civics requirement in federal grants · 25/34 (73.5%) · 13/33 (39.4%) · 4/32 (12.5%) · No
16-P7 Neutral civil-society match · 18/34 (52.9%) · 22/33 (66.7%) · 26/32 (81.3%) · No (Left short)

Result: 3 of 7 proposals reach consensus (P2, P4, P5). Compulsory service (P3) fails in every bloc.

---

Vote explanations

Prof. Victor Lindqvist (#37, Lib).

• P1 No, P2 Yes, P3 No, P4 Yes, P5 Yes, P6 No, P7 No.
• P2: "Voluntary, lottery only among those who opt in, and independently evaluated. That is the most honest test anyone has proposed, and I'd rather know."
• P7: "Federal money follows congregations into compliance regimes."
• What would change my mind on P1: two consecutive unmodified audit opinions, plus any credible causal estimate that AmeriCorps members or communities gain trust.

Prof. Tamsin Kerr (#30, C-L).

• P1 Yes, P2 Yes, P3 No, P4 Yes, P5 Yes, P6 Yes, P7 No.
• P3: "France's audit (#29) ended my appetite for skipping to universality."
• P7: "No: I'd support it if bridging, meaning cross-group composition, were a condition."
• What would change my mind on P1: evidence that the funding floor protects low-performing grantees from competition. I'd then want performance-based recompetition added.

Prof. Nathaniel Cross (#64, R).

• P1 No, P2 Yes, P3 No, P4 Yes, P5 Yes, P6 No, P7 Yes.
• P2: "The Spanish evidence (#27) is the best in the domain. A pilot with no compulsion is how conservatives should test big ideas."
• P6: "Mandating a pedagogy that hasn't reported its first trial is not conservative, and it isn't evidence-based either."
• What would change my mind on P6: I-ACTED showing durable, viewpoint-neutral civic effects at 12 months.

Prof. Ruth Abernathy (#51, C-L).

• P1 Yes, P2 Yes, P3 No, P4 Yes, P5 Yes, P6 Yes, P7 Yes.
• P6: "Yes, reluctantly. I retracted the turnout claim, but the knowledge-only approach has been tested and failed on behavior."
• What would change my mind on P6: an I-ACTED null.

Prof. Leila Farouk (#40, C).

• P1 Yes, P2 Yes, P3 No, P4 Yes, P5 Yes, P6 No, P7 Yes.
• P5: "My own retraction (F2) is the argument for it. We don't know what lasts."
• What would change my mind on P2: a power analysis showing 20,000 slots can't detect a 0.1 SD trust effect at five years with realistic attrition.

Prof. Rosa Delgado-Finch (#7, Soc).

• P1 Yes, P2 Yes, P3 Yes, P4 No, P5 Yes, P6 Yes, P7 No.
• P3: "Yes, as a minority view. Universal and paid is the only way to avoid SNU-style class skew."
• P4: "No: content fights will swamp it."
• What would change my mind on P3: evidence that wage-level voluntary service achieves class mixing without compulsion.

Plausibility note (referees).

• P1 splits Right because of the audit record (#19).
• P6 and P7 are mirror images: each bloc rejects the other's untested institutional preference.
• P2 passes Right narrowly (21/32) on the strength of the voluntary design. Two center-right members said they would defect if participation were made a condition of any federal benefit.

Stage 8: VERDICT

Moderator Prof. Adelaide Wainwright (#100, political science & institutional design, C)

Established

• Trust and social connection have fallen a lot and stayed down.
• 17% trust the federal government, and the figure has been under 30% since 2007 (#9).
• Average institutional confidence is 27% (2026), near the 26% record low (#10).
• Interpersonal trust fell from 46% to 34% since 1972 and is lowest among the young (#11).
• Church membership is below a majority (#12). People with no close friends quadrupled (#14). Time with friends fell by about two-thirds (#13).
• Knowledge-only civics does not measurably raise turnout. A sibling-fixed-effects design (#5) and a 15-state test-mandate design (#6) both find nulls.
• Brief, online depolarization interventions reduce partisan animosity but decay fast. About a third of the effect remains at two weeks. They largely fail to move antidemocratic attitudes, which are a distinct construct (#26).
• Sustained, collaborative, randomly assigned cross-group service can durably change trust and identity, at least under compulsory designs across regional or ethnic lines (#27, #28). Adversarial contact can backfire (#25).
• 2025–26 facts:
• AmeriCorps suffered mass grant terminations and staff losses in April 2025 (#16).
• Courts enjoined them (#16), and Congress funded FY26 at near-level, $1.254B (#17).
• A September 2026 settlement imposes notice and obligation terms through Feb 2027 (#18).
• The agency's financial statements have been unauditable for eight straight years (#19).

Contested

• Whether civic knowledge is falling. Grade-8 NAEP dipped 2 points (#1) while the adult Annenberg measure rose sharply (#3).
• Whether AmeriCorps as designed produces civic effects, and the $3.95:1 ROI (#20).
• Whether the GI Bill's civic effects were causal (#21).
• Whether contact effects generalize beyond the people one worked with. Mousa's generalization is null (#24). Bagues & Roth found regional generalization (#27).

Unknown

• The cause of the U.S. trust decline. Secularization, inequality, union decline, partisan sorting, media and technology are all co-trends. No identified study ranks them. Four professors' causal claims were rated Unsupported or Wrong and retracted.
• Whether action civics works. The first rigorous trial (#8) has not reported.
• Whether the cross-regional service mechanism transfers to voluntary programs and to partisan (not regional or ethnic) cleavages in the U.S.

Research agenda

1. Report I-ACTED, and fund replications with at least 12-month behavioral outcomes.
2. Run a lottery-assigned U.S. cross-regional service pilot with 1-, 2- and 5-year follow-ups (P2).
3. Use the April 2025 AmeriCorps terminations as a natural experiment on local volunteering (Kerr's prediction).
4. Run county-level tests of congregational density versus public-service presence as predictors of trust (Cross's prediction).
5. Test repeated-dose depolarization interventions for 3-month retention above 50% (Farouk's prediction).

Consensus reforms (provisional)

• P2: the Cross-Regional Service Lottery Pilot.
• P4: knowledge-rich civics, justified as knowledge and not as a turnout lever.
• P5: the Civic Evidence Fund.

Each crossed 60% in all three blocs. Every one of them is modest, experimental or evidentiary. None is a claim that we already know how to rebuild trust.

Where Symposium 1 was wrong or too glib

Symposium 1 declared "civics education plus voluntary national service" a consensus remedy. This round shows it was wrong in three specific ways:

1. It picked the two components with the weakest behavioral evidence. Knowledge-civics is null on turnout (#5, #6). Voluntary local service has no causal estimate and self-selects toward the already-civic, as France's 46% uniformed-parent share shows (#29).
2. It left out the mechanism that actually has evidence: sustained, collaborative, randomly assigned mixing (#25, #27). That mechanism is also the most liberty-constraining. S1 hid that tradeoff instead of debating it.
3. **It confused consensus on values with consensus on efficacy. Nearly everyone likes civics and service. Almost nobody here could produce an identified effect on trust. When the proposals were stated precisely, compulsory service failed in every bloc, and the only national-service proposal to pass was a pilot designed to find out whether S1's premise is true.**

The honest summary for Patrick: we know civic life has thinned. We do not know why. We have one promising, testable mechanism for rebuilding cross-group trust, and we have not yet tested it in the U.S.

Independent Audit

Audit Panel — Prof. Laura Pennington (#97, operations research, C) & Prof. Olivia Brandt (#57, statistics, C). Pennington leads this audit. Brandt refereed this thread, so she recuses from findings on her own rulings. Corrections below are appended to the record; nothing above is deleted.

1. Source check (re-fetched 2026-09-24)

# · Claim (thread) · Cited · Verified · Status · URL
1 · NAEP grade 8 civics, 2022 (#1) · −2 points, first decline since 1998; 22% Proficient; nearly one-third below Basic · Same · Confirmed · https://www.nagb.gov/news-and-events/news-releases/2023/eighth-grade-scores-decline-in-civics-and-us-history.html
2 · NAEP U.S. history, 2022 (#2) · −5 points; decline since 2014; 13% Proficient; 40% below Basic · Same · Confirmed · same
3 · Annenberg 2025 (#3–4) · 70% / 65% / 32% name all three branches; n = 1,684; speech 79, press 34, petition 12; 40% name 3+ rights vs 30% · Figures confirmed. The survey used three panels (n = 1,334–1,684), not one sample. APPC attributes the gain to political salience ("a torrent of executive actions and the resulting judicial cases") · Minor discrepancy · https://www.annenbergpublicpolicycenter.org/americans-knowledge-of-civics-increases-annenberg-survey-finds/
4 · Pew trust in government (#9) · 17%; R 26%, D 9%; 77% in 1964; below 30% since 2007 · Same (Sept 2025) · Confirmed · https://www.pewresearch.org/politics/2025/12/04/public-trust-in-government-1958-2025/
5 · Gallup institutional confidence, 2026 (#10) · 27% across 14 institutions; record low 26% (2023); 5th year below 30%; small business 67, military 61, police 45, Congress 9, TV news 14, newspapers 17 · Same (June 1–15, 2026) · Confirmed · https://news.gallup.com/poll/712436/confidence-institutions-remains-near-time-low.aspx
6 · GSS interpersonal trust via Pew (#11, S1, row #11 White/Black) · 46 → 34 → 34; ages 18–29: 26, 65+: 44; income ~50 vs ~20; master's 52 vs HS 25; White 40, Black 21 · Same · Confirmed · https://www.pewresearch.org/2025/05/08/americans-trust-in-one-another/
7 · AmeriCorps terminations and injunction (#16) · ~$400M; ~85% of staff; June 5, 2025; 24 states + DC; APA · Same · Confirmed · NCAN (URL in #16)
8 · FY26 AmeriCorps request and House mark (#17) · Request $107.7M; House $642.95M (−49%); FY25 $1.262B · Same (House committee, Sept 9, 2025) · Confirmed · https://voicesforservice.org/legislative-updates/house-proposes-nearly-50-percent-cut-to-americorps-for-fy26/
9 · FY26 enacted (#17; basis for Wrong rating K5) · $1,253,806,000 · Same (bill text Jan 20, 2026) · Confirmed · https://voicesforservice.org/news/bipartisan-house-senate-appropriations-bill-provides-near-level-funding-for-americorps/
10 · AmeriCorps OIG audit, FY2024 (#19) · Disclaimer of opinion, 8th consecutive year; 11 material weaknesses; 3 since FY2017; 78 open recommendations · Same ("declined to issue an opinion… for the eighth year") · Confirmed · https://www.americorpsoig.gov/reports/audit/audit-americorps-fiscal-year-2024-consolidated-financial-statements
11 · Jung & Gopalan (#6) · "15 states adopted 2015–18"; −0.002 (CI −0.045 to 0.042); one election cycle · Estimate, CI and "fairly precise… null" confirmed. The source says "as of 2022, 18 states have implemented a version of CEI" and all treatment states passed the law in 2015–2017. Follow-up is one presidential and one midterm election · Minor discrepancy · https://journals.sagepub.com/doi/10.3102/01623737231195887
12 · Voelkel et al. (#26; ratings F1–F3, H1) · n = 32,059; 23/25 treatments, up to 10.5 pp; 6 treatments up to 5.8 pp; 5 treatments up to 2.8 pp; 2-week retention (n = 8,644) ~32% and ~7%; constructs "distinct" · Same. The authors say animosity "is not a unifying construct" (r = 0.25 and −0.02) · Confirmed · https://bpb-us-w2.wpmucdn.com/web.sas.upenn.edu/dist/9/244/files/2024/10/science.adh4764.pdf
13 · France SNU (#29) · €3.5–5bn/yr plus €6bn; €2,900 vs €2,000 per participant; 40,000 vs ~800,000; 46% had uniformed-service parents · Same · Confirmed · https://www.publicsenat.fr/actualites/societe/risques-pour-la-securite-des-jeunes-cout-largement-sous-estime-la-cour-des-comptes-etrille-le-deploiement-du-snu
14 · Bagues & Roth (#27) · Random conscript assignment; trust toward region of service lasting decades; more identification with Spain · Same · Confirmed · https://www.aeaweb.org/articles?id=10.1257%2Fpol.20210237

Counts: 12 Confirmed · 2 Minor discrepancy · 0 Not supported · 0 Could not access.

2. Internal consistency

• Tally. 51 rows. Wrong 4 (K5, K6, M2, F2); Contested 2 (K2, K9); Unsupported 10 (K8, D3, C4, L3, M4, M5, A2, F4, V4, AC1); Supported 35. The posted tally 35 · 2 · 10 · 4 is correct.
• Vote math. All 21 percentages are correct to one decimal. Consensus labels are correct: P2 (29 / 28 / 21), P4 (23 / 29 / 27) and P5 (30 / 31 / 22) pass. "Right short" on P1, "Left short" on P7, and "P3 fails in every bloc" are all right.
• Row M4 (Marsh) needs revising. The referees rated Marsh's salience explanation Unsupported, reasoning that August 2025 was not an election year, and Marsh withdrew it. But the Annenberg release itself attributes the gain to political salience (executive actions and court cases). "Election-year" was the wrong label; the mechanism is the source's own explanation. Correction: M4 should read Contested (partly supported; APPC attributes the rise to political salience, not an election). The Contested (ii) finding in the brief and the verdict ("Annenberg rose sharply") should carry the same caveat.
• Row K6 (Kerr) rates a paraphrase. What Kerr actually said was that France "shows a democracy can decide to make service universal." The row rates "France shows universal service is achievable," which is stronger than her words. A better rating is Unsupported: SNU generalization was announced but never delivered, per the Cour des comptes. It is not Wrong against a claim she did not make.
• Verdict "Unknown" item overstates. It says "Four professors' causal claims were rated Unsupported or Wrong and retracted" under the heading of the trust-decline cause. Only C4 (Cross) and D3 (Delgado-Finch) are about that cause. A2 (action civics) and V4 (IDF) concern other questions. Correction: "two".
• Small label error. Brief #6 and 16-P4 say "15 states adopted CEI in 2015–18." The source says 18 states as of 2022, with treatment states passing the law in 2015–17 (row 11).
• Roster. All names, numbers, fields and leans match. Marsh is #48, education policy, R. Farouk's dual role as domain referee and debater was handled correctly: she recused on F1–F4.

3. Balance

The Wrong ratings are spread across blocs: Left 2 (Kerr ×2), Right 1 (Marsh), Center 1 (Farouk). So are the Unsupported ratings: roughly Left 5, Right 4, Center 1. The verdict does not favor any ideology. It rejects the Left's "halved" narrative and the Right's "freefall" and congregation-as-main-driver claims alike, and it grounds its conclusions in randomized designs.

The two slips we found run in opposite directions:

• M4 over-penalized a Right speaker. Marsh was pushed to withdraw a mechanism the source itself offers.
• K6 rated a Left speaker against a stronger paraphrase than what she said.

These roughly offset. The referee intervention during the cross-examination (Takahashi on Kerr's France claim) was unusual but accurate.

4. Grade: A−

The load-bearing figures are confirmed almost exactly, and the tally and vote math are clean. Deductions: M4 is mis-rated against its own source, K6 rates a paraphrase rather than the words spoken, the verdict counts four trust-decline retractions where there were two, and there is a small CEI state-count error.