Random Entry

In the AI Story

De Bono's canonical demonstration: a 1976 advertising workshop on improving a pencil. Dictionary opened at random. The word was 'nose.' The audience laughed, then started thinking. Within fifteen minutes they had generated more genuinely novel ideas than two hours of conventional brainstorming had produced — pencils that released scents when sharpened, pencils with textured grips, pencils that detected chemicals in paper. None of these ideas would have been reached through conventional prompting, because 'pencil improvement' activates the established pattern (better eraser, smoother graphite) and the pattern does not include noses.

The arbitrariness is non-negotiable. A word selected because it 'seems interesting' is already inside the pattern and will not disrupt it. The technique requires a word the thinker would not have chosen — from a random generator, a dictionary opened at random, the first noun in a newspaper headline. The discomfort of working with a truly random element is the signature of the technique working. If the word connects easily, it was not random enough.

Applied to AI, random entry gains an advantage de Bono could not have anticipated: the machine's associative reach vastly exceeds the human's. A human given 'archaeology' and 'notification system' might find two or three connections before strain sets in. The AI finds dozens — across information theory, excavation methods, preservation techniques, stratigraphy of dig sites. The human provides the disruption; the machine provides the associative depth. The combination generates a volume and diversity of novel ideas that neither could produce alone.

A second application introduces randomness into the process itself, not the content. Random domain shifts: 'Solve this software architecture problem as a landscape architect. Now as a choreographer. Now as an epidemiologist.' Each domain activates a different region of the model's pattern space. The landscape architect sees pathways and sight lines. The choreographer sees sequences and rhythms. The epidemiologist sees networks of transmission. Each lens reveals what the original framework excluded — not because the connections are absent from training data, but because the original pattern suppressed them.

Origin

De Bono developed random entry as part of the broader lateral thinking toolkit in the late 1960s and 1970s, demonstrating it in thousands of workshops across four decades. The technique has been validated in creativity research as producing higher rates of divergent output than unstructured brainstorming, though controlled studies specifically on de Bono's method remain limited.

Key Ideas

Arbitrariness is the mechanism. Only genuinely random elements are guaranteed to be outside the pattern; selection defeats the purpose.

Associative bridging produces novelty. The cognitive work of connecting unrelated elements is where new paths are carved.

Discomfort is diagnostic. If the connection is easy, the word was not random; if it feels impossible, the associative effort will likely produce a genuine lateral move.

AI multiplies reach. The machine finds associative paths between the random element and the problem that no individual human could cover.

Domain shifts as random entry. 'Solve this as a biologist / as a poet / as an urban planner' activates different regions of the model's pattern space.