Decisions: Democracy and Totalitarianism

“Democracy is not about having elections. It’s about having conversations. Elections are merely the mechanism by which conversations are turned into decisions.” — Nexus, Chapter 5

The Decision Problem

Every society must make collective decisions: Who leads? What laws apply? How are resources distributed? How these decisions are made depends on the structure of the society’s information network.

Harari frames the fundamental choice as: distributed decision-making (democracy) versus centralized decision-making (totalitarianism). Each approach has different relationships with information.

Two Models of Decision-Making

Distributed (Democratic): Many nodes process information independently; decisions emerge from aggregation of diverse inputs; error correction happens through competition and critique

Centralized (Totalitarian): Information flows to a central node that makes all major decisions; uniformity is enforced; error correction is suppressed as disloyalty

The Information Argument for Democracy

Why might distributed decision-making work better? Not because ordinary people are wise, but because:

• Local Knowledge: People on the ground know things central planners don’t
• Diverse Perspectives: Different viewpoints catch different errors
• Competitive Truth-Testing: Ideas must survive public debate
• Adaptive Capacity: Many small experiments beat one big plan

Democracy is essentially a information-processing system that harnesses distributed intelligence.

The Soviet Planning Problem

Soviet central planners faced an impossible task: setting prices and production targets for millions of goods. They couldn’t gather enough information or process it fast enough. Result: chronic shortages, surpluses in the wrong places, and massive inefficiency.

Markets solve this problem not through central intelligence but through distributed price signals—millions of people making billions of decisions based on local information.

The Information Argument for Centralization

Yet centralization has its own logic. Some decisions require coordination that distributed systems struggle to provide:

• Collective Action: Building infrastructure, fighting wars, responding to crises
• Long-term Planning: Democracies often struggle with problems beyond the next election
• Overcoming Resistance: Entrenched interests can block necessary reforms
• Speed: Centralized systems can decide and act faster

The Authoritarian Advantage?

Some argue that complex modern challenges (climate change, AI governance, pandemics) require centralized coordination that democracies can’t provide.

China’s handling of COVID—draconian but initially effective—is often cited. But Harari notes this argument ignores the information costs of centralization: suppressing early warnings, punishing whistleblowers, and making corrections impossible.

The Conversation Requirement

For Harari, the key to healthy democracy is conversation—genuine dialogue between people with different views. Democracy isn’t just voting; it’s the process by which citizens inform each other, debate, and reach (temporary, revisable) consensus.

This requires certain conditions:

Prerequisites for Democratic Conversation

• Shared Reality: Participants must agree on basic facts, even if they disagree on values
• Good Faith: Participants must be willing to change their minds based on evidence
• Common Language: Terms must mean approximately the same thing to all parties
• Moderated Forums: There must be spaces where diverse voices can engage without being drowned out
• Time: Real deliberation can’t happen at social media speed

The Conversation Breakdown

Harari argues that these conditions are eroding. Social media creates parallel realities where people don’t share facts. Algorithms optimize for engagement, not understanding. Political tribalism makes good faith impossible. The common language fractures into tribal dialects.

Without conversation, democracy becomes mere tribal warfare dressed up in electoral procedures.

A Brief History of Political Information Systems

• Ancient Democracy (Athens): Small scale; citizens spoke directly to each other; limited to property-owning males
• Print Revolution (16th-17th century): Pamphlets, newspapers enabled larger-scale public debate; both democratizing and destabilizing
• Mass Media (20th century): Radio, TV created shared national conversations but also enabled propaganda
• Social Media (21st century): Everyone can speak but nobody listens; conversation fragments into echo chambers

The AI Inflection Point

AI represents a potential inflection point in this history. It could enhance democratic conversation—providing better information, facilitating translation across perspectives, identifying common ground. Or it could destroy it—optimizing for engagement over truth, enabling unprecedented propaganda, replacing human judgment with algorithmic decisions.

Will AI Decide for Us?

The deeper question: if AI can process information better than humans, why bother with democratic deliberation at all? Why not let the algorithm decide?

Harari’s answer: because values can’t be computed. AI can tell us the most efficient way to achieve a goal, but it cannot tell us what goals to pursue. That’s a human question—one that requires human conversation.

Key Takeaways

• Decision Structures: How societies make decisions depends on how their information networks are structured
• Democracy as Information Processing: Distributed decision-making harnesses diverse knowledge and enables error correction
• Conversation Is Essential: Democracy requires genuine dialogue, not just voting
• Conversation Is Threatened: Modern information technology is fragmenting the shared reality that conversation requires
• The Values Question: AI can optimize for goals but cannot choose goals—that requires human deliberation