Spread of Misinformation & Collective Action

Key Themes

Pentland: Intentions, Collective Action, and Democracy

Pentland reframes misinformation not as a problem in itself but as a symptom of a deeper issue: how misinformation turns into—or prevents—collective action. His central claim is that mutual suspicion, amplified by misinformation about the other side's intentions, undermines democracy.

Core finding: In a nationally representative survey of 5,000 Americans, Pentland found dramatic and asymmetric misperceptions of political intentions:

• The left believes the right wants to cheat (limit voting access, subvert law); the right believes the left wants to subvert democracy under the guise of protecting it.
• Yet when asked directly and anonymously, people on both sides deny these extreme intentions—the misperceptions are nearly one-directional and severe.

Intervention: A five-minute intervention showing people data on the actual intentions of the other side—derived from the same survey—reduced belief in extreme intentions and changed voting behavior within weeks. A replication at Stanford (the "Mega Study," testing 140+ interventions on 32,000 people) found this intention-correction intervention to be among the most effective across multiple categories.

Mechanism: Pentland argues that intentions are hard to perceive and that people rely on observable actions (many of which could be deceptive). His solution is a "clearinghouse" of verified information about intentions, modeled loosely on Cold War confidence-building measures (the START Treaty and data-sharing agreements that allowed each side to inspect strategic intent without trust).

Policy idea: Require politicians to state intentions before election and hold them accountable afterward (citing examples like Brazil's independent commissions evaluating whether elected officials deliver on stated commitments).

Rand: Belief Formation, Interventions, and Scalable Solutions

Rand opens with an important repositioning: misinformation is not just a social-media problem. Mainstream media and political elites are also major sources and amplifiers of false claims.

Factors that make people believe claims (true or false):

1. Repetition — even claims you don't want to believe, even against your politics, even for cognitively sophisticated people, repetition induces belief through low-level cognitive effects.
2. Consistency with prior beliefs — can be motivated (confirmation bias) or rational (Bayesian updating about source reliability and worldview fit).
3. Trustworthy sources — credible sources matter, both for motivated and rational reasons.

Factors that selectively increase belief in false claims:

1. Lack of reasoning — people who engage in more critical thinking are less likely to believe false claims regardless of political alignment.
2. Lack of media/digital literacy — operationalized various ways, but literacy consistently correlates with resistance to misinformation.
3. Inattention — social-media environments encourage rapid scrolling and low engagement, reducing critical evaluation.

Main interventions presented:

1. Accuracy nudges: A Twitter field experiment with ~5,000 users who had shared links from Breitbart or Infowars was sent a simple DM: "How accurate is this headline?" (with a non-political example). Despite no engagement, sending the message primed accuracy as a goal. Result: a causal effect showing these users subsequently increased sharing of high-quality outlets (Times, CNN, Fox News) and decreased sharing of low-quality outlets (Breitbart, Daily Caller), with effect strength correlating with the outlet's fact-checker rating.

2. Crowdsourced fact-checking: In a study with Facebook, 1,100 crowdworkers rated headlines (just reading the lead, no research) on accuracy; three professional fact-checkers did detailed research. With ~15 lay raters per headline, the crowd's aggregate rating correlated with fact-checkers as well as fact-checkers correlated with each other (~0.62), and worked equally well for political and non-political content. Replication on COVID misinformation across 16 countries showed good crowd performance nearly everywhere.

3. Why crowdsourcing scales: Crowds are fast, scalable, don't centralize truth in one authority, but face challenges: mobilizing people to engage (public-goods problem) and avoiding tyranny-of-the-majority outcomes where majorities can certify falsehoods about minorities.

Implication: Combining interventions (platforms show accuracy nudges to users, integrate crowdsourced ratings into algorithms) offers a decentralized approach to misinformation that doesn't require platforms to be the arbiters of truth.

Revel: Democracy, Institutional Design, and Alternatives to Elections

Revel argues that while Pentland and Rand address misinformation and polarization, there is a complementary institutional problem: democratic institutions themselves are failing. She identifies three core components of democratic legitimacy:

1. Shared principles (e.g., Bill of Rights)
2. Institutional design (Constitution, rules of governance)
3. Belief that the democratic process matters more than winning (the "glue" of legitimacy)

This belief rests on two legs: intrinsic (the process is fair) and instrumental (the process delivers good outcomes). Both are threatened.

Why institutions fail:

• Campaign financing filters who can run
• Gerrymandering presets winners
• Unresponsiveness: Gilens & Page study shows predicted probability of policy adoption is almost completely unresponsive to public opinion

Revel argues these are examples of institutions being "hacked" (exploited in ways they weren't designed for), consistent with historical cycles of institutional rise and fall (Ibn Khaldun, Polybius, Sima Qian).

Alternative governance models:

1. Lottocracy — randomly selecting citizens (as in jury duty). Advantages: protects diversity of perspective, resistant to hacking. Disadvantages: low participation, less connection to voting.

2. Liquid democracy — delegating votes transitively on a per-issue basis. Instead of voting directly on every issue, you delegate to someone you trust (who may in turn delegate to others). Uses "second-order knowledge" (who knows what). Revel's mathematical work with collaborators shows liquid democracy can identify local experts in certain conditions; concern: differential willingness to delegate may reproduce existing social hierarchies.

Context dependency: All these models have different failure modes and apply differently depending on the question type (factual vs. value-based), group characteristics (polarization level, epistemic common ground), and institutional context (country-level governance vs. social-media platform governance).

Implementation risk: Any governance model can be hacked if not designed with adversarial resilience in mind. Revel emphasizes the need for theory of change grounded in social context, not just philosophical ideals.

Key Tensions & Questions Raised

1. Crowd wisdom vs. polarization: When crowds are politically polarized, crowdsourced accuracy ratings may fail; alternatively, selecting "experts" via delegation could amplify ideologues if there's no epistemic consensus.

2. Institutional reform as necessary but insufficient: Berinsky notes that liquid democracy might be more feasible in social-media platform governance (where one decision-maker can change things) than in national elections (with institutional baggage). But Revel emphasizes that institutional reforms alone don't address the "social fabric" problem—the underlying culture wars and affective polarization.

3. Local vs. national solutions: Pentland suggests allowing communities to decide their own rules (within rights-protecting guidelines) as a way to bypass national polarization. Berinsky notes that early U.S. history had local variation (women and Black landowners could vote in some states, later disenfranchised) and emphasizes reputation mechanisms in small communities where decisions are coupled and people have vulnerability to each other's choices.

4. Misinformation vs. institutional design: Sally Haslinger (audience) points out that cultural/identity-based distrust (white conservatives fearing demographic displacement) may dwarf any institutional fix. Revel acknowledges she doesn't have an answer and distinguishes her work (institutional design) from addressing the "social fabric" (Pentland and Rand's domain).

Notes

This panel exemplifies IDSS's mission: bringing systematic theory and empirical evidence to bear on critical social problems. Notably, three different disciplinary angles on the same problem—behavioral psychology (Pentland), cognitive science (Rand), and political science/mathematics (Revel)—show complementary rather than competing approaches. Pentland's intention-verification work is novel and draws on Cold War precedent (confidence-building); Rand's accuracy nudges build on a large experimental literature but present the Twitter field experiment as the strongest evidence of scalability. Revel's modeling of liquid democracy is exploratory and grounded in social choice theory but faces real implementation questions in polarized contexts.

The panel also surfaces a recurring tension in democratic-reform work: institutional fixes can't solve culture-war-driven distrust, yet fixing the social fabric alone (polarization, affective dislike) without institutional reform leaves voters' rational grievances (unresponsiveness, gerrymandering) unaddressed. The implication is that comprehensive reform requires work on multiple fronts simultaneously.

Berinsky's emphasis on reputation mechanisms and the importance of community-scale decision-making (where decisions are coupled and people have skin in the game) connects to historical institutions and suggests that solutions may differ at different scales—platform governance vs. national politics vs. local community deliberation.