What this covers
David Weinberger and Russ Roberts explore how the internet has fundamentally reorganized what counts as knowledge and how we attain it. Weinberger argues that knowledge is migrating from the traditional model—facts selected and arranged into linear arguments within books—toward a networked, web-based model where truth emerges through ongoing conversation among experts who grapple with complexity. The central claim is that the world is too fractally complex for book-length arguments to capture it adequately; the internet's messy web of links, by contrast, reflects reality better because it mirrors how human interest and attention actually work.
The conversation spans several distinct territories. Weinberger traces how facts themselves are a recent construction, not timeless foundations—he shows this through the evolution of Thomas Robert Malthus's work across editions and connects it to Jeremy Bentham's Utilitarianism as a catalyst for empirical investigation. He argues that real knowledge advances through conversation among people who largely agree, not through staged confrontations between opposites, a point that directly addresses the Echo Chamber worry. The discussion turns contentious on several fronts: whether multivariate social science estimates deserve to be called knowledge at all, whether the internet makes us smarter or merely plays to our biases, and whether online discourse quality can be engineered through small Choice Architecture decisions. Weinberger and Roberts also examine examples ranging from Stack Overflow as a model of Public Learning to recommendation systems built on Theory-Free Correlation, and debate how education will change as platforms like Udacity reshape which fields can be taught online.
Weinberger argues that knowledge is shifting from a fact-based, linear, book-bound model to a web-shaped, networked one in which truth lives in the ongoing argument among experts rather than in resolved narrative conclusions, because the world is too fractally complex to be captured by long-form arguments.
- Facts are a recent, constructed invention whose selection depends on human interest, so accumulating facts does not by itself yield knowledge
- Knowledge advances through conversation among people who largely agree, not chiefly through confrontation between opposites
- The internet expresses knowledge as a messy web of links that better reflects a fractally complex world than the linear book argument
Knowledge emerges through active thinking and public dialogue, not passive consumption of fixed information.
- Knowledge never resided in libraries or books themselves; you do not become wise or knowledgeable by consuming books but by chewing, thinking, and relating what you read to your own life and experience—and the internet makes us smarter precisely because it lets us do this processing in public and leave the trails of our thinking as a growing web of connections.
“There was never knowledge until you tried to read those books and process it through your own life and experience. You don't get wise or knowledgeable by consuming. You get wise and knowledgeable by chewing and thinking and relating.”
- A blog is a platform to think out loud where, unlike a published book, you do not have to write your final word—you can evolve, learn new things, and change your mind by editing—so the web erases the fixity of print and paper, though imperfectly, since some readers may see only the earlier erroneous version.
“it's a place to think out loud. Because you are the editor, you don't have to write your final word. It can evolve. You can learn new things. You can change your mind.”
Multivariate social-science estimates presented as facts are largely fake knowledge disguising unresolvable causal questions.
- Multivariate statistical estimates in social science—such as whether the stimulus created jobs or the multiplier on a dollar of debt-financed government spending—are not facts and not even close to facts; they get quoted as precise numbers like 1.57 or 0.6, but questions about causal social relationships are not really scientific questions even though scientific techniques are used, and treating such estimates as knowledge is largely fake knowledge.
“Have we established a fact about, say, the relationship between a dollar of government spending financed by debt and the impact on national income? I'd say: No, it's not a fact. It's not even close to a fact.”
- The 1980s decline in U.S. home ownership was treated as evidence that housing had become unaffordable and used to justify pro-homeownership policy, but the rising divorce rate of the 1970s increased the number of households (the denominator), mechanically lowering the ownership rate without telling us anything about housing affordability—illustrating how a true fact can be wrongly explained by an unexamined theory.
“the divorce rate was rising dramatically in the United States throughout the 1970s. And so when you went to measure the home ownership rate you had a lot more entries in the denominator, because there were a lot more households all of a sudden.”
Concerns about open information access are longstanding, and internet's impact on learning is discipline-dependent rather than uniform.
- Fears that open access to information is dangerous because people can't handle it or will only consort with the like-minded are not new; people have always complained about open access for many reasons, as seen in the brutal negative campaigning against Thomas Jefferson, so the internet merely adds a few new complaints to old ones.
“People have always complained about open access being dangerous, for a hundred different reasons. So, now we have a few new ones. But we have the old ones, too.”
- Education will change radically over the next 5-10 years, with online platforms like Udacity coming to dominate teaching in computer science, math, and similar fields, but the collaborative online learning tools are harder to bring to bear in other areas, so the internet's effect on knowledge is highly discipline-dependent.
“I think education is going to change radically in the next 5-10 years. I'm very excited about what Sebastian Thrun is doing with Udacity.com”
Major scientific debates may be unresolvable through traditional methods yet still yield learning through argument.
- Major debates like whether Keynesian stimulus works are probably not resolvable through the traditional methods of science, yet both sides still learn something from arguing about it even without convincing each other the way a replicable experiment would—though what they may learn is to become more entrenched in their existing views.
“I've come to believe it's probably not resolvable through the traditional methods of science. But you do learn something from arguing about it, even though you don't convince each other in a way that you do through a normal replicable experiment.”