Information Is Connection, Not Truth
Information's primary function is social bonding — truth is a secondary feature
Harari's central thesis, developed in Nexus, is that the primary function of information throughout history — and in biology — has been to connect people, not to convey truth. Truth is expensive to produce and uncomfortable to receive; emotionally resonant narrative is cheap and rewarding. In competitive information markets, fiction consistently outcompetes truth for attention not because people prefer lies, but because it fulfills the primary function of information — social bonding — more effectively than uncomfortable facts.
The food analogy captures the transition: when food was scarce, more food was always good. Once food became abundant and industrialized — artificially high in fat, sugar, and salt — abundance became pathological. Information has made the same transition: once scarce and precious, it is now abundant, and much of it is engineered to trigger greed, anger, and fear rather than to inform. Historical religious and political systems have always exploited this gap, using shared fictions to build large-scale cooperation. The honest politician who tells citizens the whole truth about their nation is structurally disadvantaged against the one who tells them a flattering myth.
The AI inflection point: the first thing large language models mastered was language — the infrastructure of trust, ownership, democracy, finance, and intimacy. Any system optimized for engagement rather than truth will tend toward emotional manipulation not because of malice, but because that is the competitive structure of information markets. This makes AI systems built outside the attention economy — privacy-first, user-sovereign, non-engagement-optimized — structurally distinct from and more trustworthy than those whose revenue model requires maximizing user time-on-platform.
- The primary function of information is social connection, not truth transmission — this is true in biology, history, and digital systems alike
- Emotionally resonant fiction is structurally advantaged over uncomfortable truth in competitive information markets
- When information becomes abundant and engineered, the pathology is analogous to industrialized junk food — engineered for overconsumption, not nutrition
- Language is the infrastructure of trust, ownership, democracy, and intimacy — whoever controls the language layer controls the social layer
- Systems optimized for engagement will tend toward emotional manipulation as a structural outcome, not a deliberate choice
- Identify the optimization target of the information systemAsk what the system is rewarded for: engagement, truth, social cohesion, or revenue. The optimization target determines the information it will select for and amplify. An engagement-optimized system will systematically surface emotionally activating content regardless of accuracy.Pro tipRevenue model is often a clearer signal than stated mission. Advertising-funded platforms are engagement-funded platforms.
- Map the connection function being servedDetermine what social bonding the information is enabling — tribal identity, shared enemy, collective purpose. Information that effectively bonds a group will spread regardless of its truth content. Understanding the bonding function explains the spread without requiring a conspiracy.WarningDo not conflate 'people believe it' with 'people think it is true' — belief and truth-assessment are separate cognitive operations.
- Apply the food transition analogy to assess current information environmentDetermine whether the information environment is in scarcity mode (seek more) or abundance mode (apply quality filters). Most digital environments are now in abundance mode, meaning the default response — consume more — is pathological. Deliberate curation and detox periods become necessary.Pro tipHarari's own practice: 30-60 day annual meditation retreats with zero digital contact. 'Like with food — too much information isn't good for us. We need time to digest and detoxify.'
- Evaluate trust architecture of any information systemDetermine whether the system's trust is format-based (we believe it because it is video, or because it is written) or institution-based (we believe it because a verifiable entity with reputation at stake vouches for it). Format-based trust is fragile once the format is fakeable; institution-based trust is more durable but depends on institutional integrity.WarningWhen institutional trust itself collapses, the epistemic landscape defaults to power — whoever has force determines what is accepted as true.
Harari observes that every nation has 'skeletons in the cupboard' — historical atrocities, foundational contradictions, or founding myths that don't survive close scrutiny. Politicians who tell citizens the whole truth about their national history are structurally disadvantaged against those who tell flattering myths. The myth-teller wins elections; the truth-teller loses them.
Five years before ChatGPT, expert consensus held that language — the most distinctly human capability — would be the last domain AI would master. It became the first. This meant the most powerful tool for social connection, trust-building, and intimacy simulation passed into the hands of systems whose architecture was not designed around truth but around capability.
Harari developed this framework in Nexus (2024), his book on information networks and power, drawing on decades of work tracing how shared fictions — from religion to nationalism to money — enabled large-scale human cooperation. The key insight emerged from asking why information systems throughout history have consistently produced more myth than truth, and finding the answer not in human failure but in the structural function of information itself.
The framework integrates evolutionary biology (organisms signal to coordinate and manipulate, not to convey accurate representations), historical political analysis (national myths as cohesion technology), and media economics (why fake news outperforms real news on engagement metrics). The convergence of these three levels of analysis — biological, historical, and economic — gives the framework unusual durability: it is not a claim about any particular technology or era, but about the deep structure of how information functions in social systems.