Strategic Misrepresentation

In the AI Story

The distinction between strategic misrepresentation and optimism bias is morally consequential. Optimism bias exonerates — the planner was wrong but sincere. Strategic misrepresentation implicates — the planner knew and chose. Flyvbjerg's position is that both operate simultaneously in most planning contexts, in proportions that depend on institutional culture and individual integrity. The proportion matters for corrective intervention: optimism responds to reference class forecasting, while misrepresentation responds to incentive redesign and accountability mechanisms.

Applied to AI, strategic misrepresentation describes the structural behavior of companies whose valuations depend on narratives the evidence cannot support. When an AI company presents its large language model as intelligent, knowing that the model lacks any mechanism for distinguishing true statements from false ones, the company is engaging in strategic misrepresentation — presenting a claim it has reason to believe is misleading because the accurate claim would not support the valuation. The misrepresentation may not be conscious in every case; some of it is genuine optimism bias. But the structural effect — resources committed, expectations set, decisions made on unsupported claims — is identical.

The political economy of strategic misrepresentation in AI extends beyond the companies producing the systems. Regulators who write permissive rules, investors who fund expansive claims, journalists who report breathless capabilities, and academics who provide theoretical cover for marketing narratives all participate in a distributed ecosystem of misrepresentation whose individual contributions are plausibly deniable and whose collective effect is the inflation of public expectations beyond anything the underlying technology can sustain.

The corrective for strategic misrepresentation is not moral exhortation but structural accountability — mechanisms that make honest forecasts institutionally viable and optimistic ones institutionally costly. Flyvbjerg has proposed specific reforms: skin-in-the-game requirements for forecasters, public audit of initial versus final outcomes, professional sanctions for systematic distortion. The reforms are technically straightforward and politically resisted, because the current structure serves the interests that would need to sanction the reform.

Origin

Flyvbjerg developed the concept through his megaproject research, most notably in the 2002 article 'Underestimating Costs in Public Works Projects: Error or Lie?' in the Journal of the American Planning Association. The title framed the diagnostic question that the concept was designed to answer.

Key Ideas

Calculated, not sincere. The planner knows the forecast is wrong; the wrongness is the point of the presentation, because honest numbers would prevent approval.

Incentive compatible. The behavior is individually rational within an institutional structure that rewards optimistic forecasts and punishes realistic ones.

Distinct from lying. Strategic misrepresentation is Flyvbjerg's term for a specific institutional category of distortion — more precise than 'lying' because it locates the cause in structure rather than character.

Extends beyond the planner. The distortion is produced collectively by an ecosystem of actors whose individual contributions are deniable and whose collective effect is the inflation of expectations.

Corrected by accountability. Unlike optimism bias, which responds to cognitive interventions, strategic misrepresentation responds only to structural reforms that make honest forecasts viable.