Web 2.0 platforms don’t just reveal demand — they substantially create and shape it

The classical economic picture treats preferences as exogenous — markets reveal what people already want. But recommendation algorithms don’t passively surface existing preferences; they train preferences through controlled exposure. The observation mechanism and the influence mechanism are the same system. If preferences are substantially endogenous to the information environment, neither markets nor central planning can claim privileged access to “what people really want.”

Explanandum

Does the attention economy make demand more legible (strengthening the case for AI-assisted planning) or more manufactured (undermining the foundations of both market and planning-based systems)?

Substance

Hayek’s argument for markets assumes preferences are exogenous — people want things, markets reveal those wants through willingness to pay, prices coordinate supply. On this view, digital behaviour data makes preferences more articulable, potentially strengthening the case for central optimisation.

But the alternative view, anticipated by Galbraith’s “dependence effect” in The Affluent Society (1958), is that the production system creates the wants it then satisfies. Orthodox economists dismissed Galbraith as paternalist — Hayek himself argued there’s no principled line between “genuine” and “manufactured” preferences since all preferences are shaped by environment.

The Web 2.0 era makes this harder to dismiss. A television advert is a blunt instrument. A recommendation algorithm is individually targeted, continuously adaptive, and optimised against real-time engagement data. It runs something like gradient descent on your attention, iteratively adjusting outputs to maximise time-on-platform. The person being optimised over experiences this as a service that understands them well. But the end state — their revealed preferences — is as much product of the optimisation as input to it.

This creates a genuinely novel epistemic problem. If preferences are substantially endogenous to the information environment, the “signals” that markets or planners read are partly reflections of the reading apparatus itself. This isn’t Goodhart’s Law (targets corrupting measures) — it’s more fundamental: the measurement process altering the thing being measured. Closer to Heisenberg than Goodhart.

The implications ripple across all three articles. Google’s dominance looks productive on Olley-Pakes, but if Google shapes what users expect from search, the high covariance is partly an artefact of manufactured evaluative criteria. China’s cheap water policy doesn’t just meet demand — it constructs a political economy where demand for cheap water is a cultivated dependency. Land readjustment’s value uplift is real but endogenous to a policy environment.

Supports

• Galbraith’s dependence effect, dismissed in 1958, looks more plausible in the algorithmic era
• Facebook’s internal research (the “teen mental health” documents) showed the platform shaped, not just served, user preferences
• The replication crisis in behavioural economics suggests that even “revealed preferences” in lab settings are sensitive to framing effects

Challenges

• Hayek’s rebuttal has force: all preferences are environmentally shaped, so where do you draw the line between “genuine” and “manufactured”?
• People do exercise agency — they leave platforms, change habits, resist algorithmic nudges
• The strongest version of the argument (preferences are entirely constructed) implies there’s no ground truth to optimise toward, which may be nihilistically unhelpful

Open Questions

• If neither markets nor planners can claim privileged access to “real” preferences, what becomes the basis for economic policy?
• Does the question shift from “which system best reveals preferences” to “which system shapes preferences in directions we’d reflectively endorse” — and is that an answerable question?
• Can the Ise principle help — periodic rebuilding of information environments to prevent preference lock-in?

Source Context

Emerged from the observation that the “information economy” might be making demand more visible or might be substantially creating it — and that the distinction matters for the Hayekian knowledge argument and for all three articles’ implicit assumptions about what markets and institutions are optimising toward.