The First Idea Problem

In the AI Story

The problem is structural rather than personal. It is not that people are too hasty or too lazy to consider alternatives. It is that the self-organizing system actively produces the first idea as a feature — the pattern's function is to make recognition and response efficient, which means the pattern must produce a default answer quickly. The first idea is the system doing what it evolved to do. The failure is not in producing the first idea but in treating the first idea as the best idea without the deliberate lateral work that would have revealed alternatives.

AI systems have a permanent first-idea problem. Every output is, in de Bono's terms, a first idea — the response that follows most naturally from the statistical patterns of the training data. The output can be refined, elaborated, nuanced — but the refinement is vertical. The output stays within the same framework, the same set of assumptions, the same pattern. To reach a different framework requires a lateral move that the system cannot make from within its own dynamics.

The builder who prompts an AI for a solution and refines the AI's output has replicated the human first-idea problem at computational scale. The machine generates its first idea. The builder evaluates and directs refinement. The refinement stays within the framework. The output converges on a polished version of the default the pattern naturally produced. The collaboration feels productive — the output is competent and often impressive — but nothing has been produced that could not have been reached by following the pattern.

The escape is the same escape de Bono identified for the biological version: deliberate lateral intervention before accepting the first idea. A provocation, a random entry, a domain shift, a demand for alternatives that are not refinements of the first one. The discipline costs time — time the fluent first output makes it tempting to skip — but the discipline is what differentiates output the pattern would produce anyway from output the pattern could never have reached.

Origin

De Bono articulated the first idea problem across his creativity works, with particular emphasis in Serious Creativity (1992). The concept generalizes beyond creativity into judgment and decision-making — it is a specific case of the broader phenomenon that Kahneman and Tversky later formalized as the default operation of System 1.

Key Ideas

The first idea is the pattern's default. Not lazy or hasty thinking — the structural output the self-organizing system produces when doing its job.

Refinement is not creativity. Polishing the first idea produces a better version of the default; it does not produce anything the pattern could not have reached.

AI makes the problem permanent. Every model output is a first idea at computational scale; prompting for refinement locks the collaboration within the default pattern.

Demand alternatives before evaluation. De Bono's APC tool — Alternatives / Possibilities / Choices — forces the generation of three alternatives before evaluation begins.

Lateral intervention as discipline. Provocation or random entry before accepting the first idea is what separates genuinely novel output from polished convergence.