Skip to content

Assumption Reversal and Recombination

A generative method — instantiates Divergence-Convergence Cycle Orchestration

Generates structurally different options by naming a design's load-bearing assumptions, negating them, and recombining the fragments — so breadth comes from reframing, not from decorating one idea.

Most brainstorming produces variations on a single hidden template: everyone is quietly agreeing on what the thing is and only arguing about its trim. Assumption Reversal and Recombination attacks that template directly. It makes the group write down the assumptions the current framing takes for granted — the ones so obvious nobody states them — then deliberately negates each one and asks what design would have to exist if the reversal were true. The recombination step is what makes it more than a provocation exercise: fragments from several reversed assumptions are crossed to assemble whole candidates that sit in a genuinely different part of the solution space. Its defining move is that novelty is manufactured by reframing, not hoped for from a bigger idea list.

Example

A public library system is losing foot traffic and the team's first ideas are all the same shape — nicer furniture, longer hours, a better app. Assumption Reversal starts by surfacing what "a library" silently assumes: the collection is physical; you come to the building; silence is the norm; we lend things and you bring them back. Each is negated in turn. Physical → the collection is skills and tools, not books. You come to us → the library travels, as lending lockers on transit platforms. Silence → the space is loud and social, built around making. Lending → nothing is returned; you take a seed, grow it, return two.

Recombining those fragments yields candidates that no amount of polishing the current model would have reached: a library of things run from neighborhood lockers, a noisy maker-and-repair commons, a seed-and-tool exchange with no due dates. The output is a small set of structurally distinct option families, each tagged with the reversal that produced it — raw material the convergence side can later screen, not a decision.

How it works

What distinguishes it from ordinary ideation is the two-step forcing function:

  • Name the load-bearing assumptions. Not peripheral preferences — the propositions the whole current framing rests on. If negating an assumption changes nothing, it wasn't load-bearing.
  • Reverse each one. Fully negate or sharply relax it, and describe the world in which the reversal holds.
  • Recombine across reversals. Cross fragments from different reversed assumptions into whole candidates, so options differ along several axes at once rather than one.
  • Keep provenance. Tag each candidate with the reversal(s) that generated it, so the inventory records why each option is structurally different.

Tuning parameters

  • Assumption depth — target load-bearing assumptions or surface ones. Reversing peripheral assumptions is safe but yields trivial novelty; reversing core ones is where structural difference comes from and where discomfort lives.
  • Reversal aggressiveness — full negation versus partial relaxation. Full negation reaches further but produces more infeasible fragments; relaxation stays nearer the possible.
  • Recombination breadth — how many reversed fragments you cross into one candidate. More crossing widens the space but multiplies half-formed hybrids.
  • Feasibility deferral — how long "impossible" reversals are kept alive before pruning. Killing them early keeps the list tidy but re-imposes the very assumptions you were trying to escape.

When it helps, and when it misleads

Its strength is escaping functional fixedness — the pull that makes an existing solution feel like the only shape the problem can take.[1] By forcing the frame to move, it reliably clears the diversity floor that a normal ideation session quietly fails, and it surfaces option families a scoring method could never have ranked because they were never on the table.

Its failure mode is novelty for its own sake: reversals can manufacture options that are structurally different and also useless, and a big pile of contrarian fragments can feel like progress while containing nothing viable. The classic misuse is running it backwards — reversing assumptions to rationalize a contrarian idea someone already wanted to pursue, rather than to open the space honestly. The discipline that guards against this is to pair it with an explicit diversity floor and a downstream screen, and to treat its output strictly as raw candidates for convergence to test, never as a verdict.

How it implements the components

This method sits entirely on the divergence side — it produces breadth and hands off before any narrowing:

  • exploration_dimension_map — the named assumptions become the axes of exploration; each reversal defines a direction the space can be pushed.
  • option_diversity_floor — its whole reason to exist is guaranteeing that the option set spans structurally distinct families, not near-duplicates.
  • candidate_option_inventory — it emits candidates tagged with their originating reversal, giving the inventory both options and provenance.

It does not set convergence_criterion_set or build an option_comparison_frame — the reversed options are unranked on purpose; those belong to Concept Screening Funnel and Multi-Criteria Decision Matrix.

Notes

Reversal must exempt the non-suspendable guardrails. You negate assumptions about form, channel, and business model — not the safety, legal, or rights constraints that stay fixed no matter how the design is reframed. Skipping that exemption is how an assumption-reversal session drifts into "options" that are novel only because they quietly discard a non-negotiable; the guardrail set (held by Divergence-Convergence Workshop) is what keeps the reversal honest.

References

[1] Functional fixedness — the cognitive tendency to perceive an object or design only in terms of its established use — is the specific bias this method is built to break, which is why it targets the assumptions a framing treats as fixed rather than the details it already varies. ↩