Worst possible idea / reverse brainstorming

Status: Folded · Evidence: P · Family: Divergent ideation and idea expansion · Verdict: fold (2026-06-09)

Use instead: Assumption Reversal

What it is

Worst possible idea (also called reverse brainstorming, negative brainstorming, or the headstand method) is a divergent-ideation technique that inverts the brief: instead of asking “what is a good solution?”, the group is told to generate the most terrible, absurd, impractical, or actively harmful ideas it can - and is given explicit permission to revel in them. Then the worst ideas are mined: the facilitator flips each into its opposite, or pulls out the attribute that makes it bad and asks what the inverse of that attribute would look like, and the most interesting reversals are carried forward as candidate solutions. The emitted artifact is an idea shortlist - the same class of deliverable a forward brainstorm produces, reached by a backward route.

The technique has two distinct rationales in the literature, and separating them is the whole game for grading and placement:

1. The psychological (de-inhibition) rationale - the dominant one in the modern design-thinking write-ups. The claimed mechanism is that hunting for bad ideas removes the fear of being wrong, relaxes the participants, and lifts the self-censorship and performance anxiety that throttle an ordinary brainstorm; Bryan Mattimore, who coined the phrase, describes a stuck group that “loosened up” once it was searching for downright awful ideas. On this reading the worst-idea framing is a permission device, and the inversion at the end is almost incidental.
2. The mechanical (inversion) rationale - the older “reverse brainstorming” tradition. Whiting’s 1958 Reversal-Direct Method: enumerate the ways to make a product or process fail, then reverse those into solutions. On this reading the method is “produce a set of X, then invert X to get options” - a generation procedure whose engine is reversal, applied to whole candidate ideas (and their quality) rather than to a forward search.

Both readings land on the same artifact (an idea list) and differ only in what they think is doing the work - relaxation, or reversal. As the verdict section argues, neither rationale yields a durable move the catalog does not already own: the inversion engine is already carved up across the shipped skills, and the de-inhibition rationale is precisely the brainstorming mechanism that the strongest experiment in the field cast doubt on, and that an AI agent does not need in the first place.

When it helps / when it misleads

As a facilitation move with human groups, worst possible idea helps when a room is creatively frozen - the obvious ideas are exhausted, people are self-censoring, the session has gone quiet - and a permission-giving, low-stakes prompt is needed to restart flow. It is genuinely good at breaking the “we must say something smart” pressure, and the flip-the-attribute step can surface a non-obvious angle the forward search glossed (the worst idea often inverts a hidden assumption about who pays, when it happens, or what is allowed).

It misleads or wastes effort when:

• The “bad” ideas do not invert into anything informative. For many problems the worst ideas are simply bad - their inverse is the obvious good idea you already had, or a degenerate non-answer. The method guarantees a list of terrible ideas; it does not guarantee that flipping them yields more than the forward question would.
• The flip step is skipped or fudged. Without a disciplined “extract the load-bearing attribute, then negate that attribute” step, the exercise produces a venting list of jokes and never converts. The value, when there is value, is in the structured reversal at the end, and that reversal is exactly the move the shipped assumption-reversal already proceduralizes.
• It is run as the source of recommendations. Like every divergent generator, it produces candidates, not decisions; treating a flipped worst-idea as a vetted answer skips the convergence and viability work entirely.
• The de-inhibition rationale is imported into a context that has no inhibition to relieve. This is the decisive caveat for this library. The headline mechanism - “removes fear of evaluation and self-censorship” - is a claim about a human social-psychological failure mode (evaluation apprehension in a group). An AI agent generating ideas has no fear of ridicule, no performance anxiety, and no quiet-room dynamic to break; the permission device is solving a problem the agent context does not have. What is left, once the psychology is stripped out, is bare inversion-to-generate - which is already a shipped move.

What the evidence says

The honest grade for the candidate’s stated move - “generate the worst ideas, then flip them into good ones” - is P (practitioner), and this entry has to be careful in the same way the inversion entry was, because the popular write-ups attach a borrowed psychological rationale to a technique that has never itself been tested.

What the record supports. Worst possible idea / reverse brainstorming is a real, named, long-lived divergent-ideation technique with a clear lineage: the older “reverse brainstorming” / Reversal-Direct root (Whiting, 1958) and the modern design-thinking branding (Bryan Mattimore coined “worst possible idea”; the Interaction Design Foundation is its canonical contemporary teaching home). As a facilitation device for unsticking human groups it is widely taught and plausibly useful. That practitioner standing is the full extent of the directly-supported claim.

What the record does NOT support, and the laundering trap. There is no controlled or comparative study I can locate that tests reverse brainstorming / worst possible idea as a method against a forward brainstorm on idea quantity or quality; the design-thinking sources (IxDF, Atomic Object, Mattimore) are practitioner accounts and anecdotes, not experiments. The temptation is to borrow the brainstorming-productivity literature to make the method look evidence-backed - and here that borrowing does not merely fail to transfer, it points the other way:

• The technique’s headline mechanism is that it defeats evaluation apprehension / self-censorship. But Diehl & Stroebe (1987), the field’s foundational dissection of brainstorming productivity loss, found in their Experiment 3 that evaluation apprehension was not a major explanation of the loss; the dominant cause was production blocking (the turn-taking bottleneck), which a worst-idea framing does nothing to address. So the best-known experiment in the area undercuts, rather than supports, the rationale usually given for this method. Citing it as support would be backwards.
• The general brainstorming finding that loosely motivates the method - “go for quantity, defer judgment” (Osborn) - is real but is the property of brainstorming in general, already shipped here as brainwriting (S-tier). It does not single out the worst-idea inversion as the active ingredient, and the quantity-quality link is itself mixed in the literature.

Borrowing either body to lift this entry above P would be laundering a cousin’s (or in the Diehl & Stroebe case, an outright contradicting study’s) results onto a move neither tested. The conservative governing grade is therefore P: a recognized practitioner / design-thinking technique, with no direct controlled evidence for its own framing, and with the nearest rigorous brainstorming evidence explicitly not counted toward it (and noted as cutting against its stated mechanism).

Transfer caveat (required). All of the adjacent evidence is from human subjects in group and classroom settings; none of it studies worst possible idea performed by or with an AI agent. The one AI-context thread that is mechanically adjacent - “denial prompting” / constraint-based prompting, which incrementally forbids routine solutions to push an LLM into novel regions of its generative space (Lu et al., NeoGauge) - is a different operation (forbid-the-obvious, not generate-the-worst-then-flip) and does not validate this method on agents either. The evidence is transferred from human contexts and not validated for AI-augmented use.

Excluded figures (required). No specific effect size or percentage is claimed for this technique anywhere in the located sources; the design-thinking write-ups are qualitative (“loosened up,” “overcame the impasse”). There is therefore no widely-quoted statistic to exclude - but for the record, any figure attached to “reverse brainstorming improves ideation by N%” would have no primary source measuring this technique and could not move the grade.

Why it is / is not a skill here

Verdict: Fold into assumption-reversal. This overturns the catalog’s prior cand / build / P tag (“Candidate; relates to inversion”); the concrete reason follows, and it is the reason that prior tag already half-conceded by pointing at inversion.

The Build burden is to name one distinct, durable cognitive move that no shipped skill produces, and to show no existing skill (or short chain) already produces it. Worst possible idea fails that burden because, once its human-facilitation psychology is set aside, its mechanical core is inversion-to-generate - and the catalog has already deliberately partitioned that move:

• The mechanical move is the reverse-to-ideate engine that assumption-reversal already proceduralizes. assumption-reversal surfaces the load-bearing premises, negates each, and generates ideas from the reversed world, ending in a shortlisted assumptions-and-reversals sheet. Worst possible idea generates whole bad candidates, then negates the attribute that makes each bad, and generates ideas from the reversal, ending in a shortlist. Same engine (negate, then generate), same artifact-class (a reversal-driven idea shortlist), differing only in the object of the negation: assumption-reversal inverts the premises; worst-possible-idea inverts the quality judgment on candidate ideas. That is well within the ~20% overlap ceiling - it is one mode of the same negate-and-generate move, not a separable mechanism. Tellingly, assumption-reversal already defines itself against generic inversion in its own SKILL.md (“It is not generic inversion (‘how would we cause failure?’); it targets the foundational premises”), and the shipped inversion entry’s dossier names worst-possible-idea explicitly as inversion’s “ideation reading” - the catalog has, in two places, already located this move inside the reverse-to-ideate family. The schema target resolves: assumption-reversal is status: shipped.

• The generation muscle is already brainwriting, and the de-inhibition rationale does not transfer. The part of worst possible idea that is “produce lots of low-stakes candidates without fear of judgment” is the generic brainstorming move, shipped as brainwriting (S-tier, whose entire point is to defeat anchoring and apprehension by silent parallel generation). For the library’s AI-agent context the fear-of-evaluation rationale is moot anyway (an agent has no evaluation apprehension to relieve), so the psychological angle cannot rescue a standalone skill - it leaves only the inversion engine, already covered above.

• scamper’s “R = Reverse” prompt is the nearest single-prompt cousin, but it is not the home. SCAMPER reverses one attribute of an existing seed; worst possible idea reverses the quality of whole generated candidates. That is a real difference in object, but it cuts toward “this is a flavor of reverse-to-generate,” not “this is a new move SCAMPER lacks” - and SCAMPER does not carry the generate-bad-then-flip procedure, so it is the wrong fold target.

So there is no separable artifact that is uniquely “worst possible idea.” Its mechanical core duplicates assumption-reversal’s negate-and-generate engine most directly; its de-inhibition value is human-facilitation psychology that (a) the best brainstorming experiment doubts and (b) an AI agent does not need. That is a fold, not a build.

Why fold rather than recipe or reject: it is not a clean fixed chain of two shipped skills (it is one generation stance whose engine maps onto one existing move), so it is not a recipe. And reject would be less informative than fold - the technique is real, taught, and worth locating, so the honest service is to point the reader to where its durable move already lives, exactly as the library did when it folded inversion into premortem and steelmanning into red-team-light. Fold it into assumption-reversal as the reverse-to-ideate home, and let the dossier record that the pure generation muscle lives in brainwriting and the single-attribute reverse prompt lives in scamper. The learning value of the NO: a beloved, genuinely useful facilitation device is not automatically a skill. Worst possible idea is largely a way of getting humans to relax into generating - and a library that ships artifact-producing moves for AI agents, not room-warming devices for nervous groups, documents it and folds it rather than shipping a psychologically-motivated near-twin of assumption-reversal.

Lineage and who to read

The “reverse brainstorming” idea has an older root in Charles S. Whiting, whose 1958 Creative Thinking described the Reversal-Direct Method - enumerate the ways to make something fail, then reverse them into solutions (the same year Charles Hutchison Clark popularized brainstorming itself as a counter to “negative conference thinking”). The modern, design-thinking-branded form, “Worst Possible Idea,” was coined by Bryan Mattimore (co-founder of The Growth Engine Company; Idea Stormers, 2012), who tells the origin story of a stuck professional group that broke its impasse by hunting for awful ideas. In German design-method catalogues it appears as the Kopfstandmethode (headstand method). Its canonical contemporary teaching home is the Interaction Design Foundation (Rikke Friis Dam and Yu Siang Teo), which frames it as a start-of-ideation technique that relaxes participants by inverting the search. For the rigorous brainstorming evidence that this entry deliberately does not count toward it (because it tests a different operation, and in part contradicts the method’s stated mechanism), read Diehl & Stroebe (1987) on production blocking versus evaluation apprehension; for the generic generation move it leans on, read the brainwriting lineage (Osborn; Rohrbach’s 6-3-5); and for the move it folds into, read the shipped assumption-reversal. “Reverse brainstorming” and “worst possible idea” are generic descriptive terms in common use - no trademark, attribution to Whiting and Mattimore is courtesy not requirement - so this entry is documented descriptively and is not flagged as branded.

Named sources

• Charles S. Whiting, Creative Thinking (Reinhold, 1958). Introduced the “Reversal-Direct Method”: list the ways to make a product or process fail, then reverse them into solutions - the documented root of reverse brainstorming. Practitioner / foundational. (P)
• Bryan Mattimore, Idea Stormers: How to Lead and Inspire Creative Breakthroughs (Jossey-Bass, 2012). Coined “Worst Possible Idea”; the origin anecdote and the de-inhibition rationale (“the group loosened up”) come from here. Practitioner / popular. (P)
• Interaction Design Foundation (Rikke Friis Dam & Yu Siang Teo), “What is Worst Possible Idea?” and “Learn How to Use the Worst Possible Idea Method.” The canonical contemporary teaching write-up; qualitative, cites no controlled study. Practitioner / popular. (P)
• Atomic Object (Spin blog), “Design Thinking Toolkit, Activity 26 - Worst Possible Idea.” A representative practitioner facilitation guide (generate worst ideas, list their bad attributes, flip them). Practitioner. (P)
• Michael Diehl & Wolfgang Stroebe, “Productivity Loss in Brainstorming Groups: Toward the Solution of a Riddle,” Journal of Personality and Social Psychology 53(3) (1987): 497-509. Four experiments; Experiment 3 found evaluation apprehension is NOT a major cause of brainstorming productivity loss, Experiment 4 isolated production blocking as the dominant cause. Cited to show the method’s stated “removes fear of evaluation” mechanism is not supported (and is partly contradicted) by the strongest adjacent evidence; measures group brainstorming, NOT worst-possible-idea. (M, for production-blocking - explicitly not counted toward this entry)
• Charles Hutchison Clark, Brainstorming: The Dynamic New Way to Create Successful Ideas (Doubleday, 1958). Locates the period framing of brainstorming as a counter to “negative conference thinking”; context for the reverse-brainstorming lineage. Practitioner / foundational. (P)

Excluded under the evidence rule: no primary-source effect size exists for worst possible idea / reverse brainstorming; the design-thinking sources are qualitative. Any “improves ideation by N%” framing would trace to no study of this technique and is not counted toward this entry’s grade. The Diehl & Stroebe (1987) M-tier result is recorded but deliberately NOT counted toward the grade - it measures a different operation and, where relevant, contradicts the method’s stated mechanism rather than supporting it.