Many cheap ideas before one expensive one: divergence and design fixation

The most quotable rule in the studio is the least well evidenced. The rule is to make several candidates before you refine any of them. The rule rests on one outcome experiment, and on a much larger and divided literature about the failure it is meant to prevent.

This unit argues that producing several candidates before refining any improves design outcomes, because it counteracts fixation. The evidence is strongest for parallel prototyping, thinner and contested for the mechanism by which examples inspire or fixate: the outcome evidence comes from a few short experiments with novices, while the mechanism evidence is larger, older and divided. It is the mechanism, not the outcome, that usually justifies the rule. This is the fifth of ten units; it takes the candidates unit four drew and hands unit six one of them to prototype.

One experiment tests the rule by measuring the finished design

Alex Osborn (1953) set out the procedure the studio inherited. Brainstorming deferred judgement, put quantity before quality and combined other people's ideas. Osborn was an advertising executive, not a researcher. The doctrine therefore entered design education as a habit rather than a finding.

The habit's strongest test compares two ways of working. Dow, Glassco, Kass, Schwarz, Schwartz and Klemmer (2010) had thirty-three novice designers make banner advertisements for one client, a student-run design magazine, each producing five prototypes and a final advertisement and receiving a descriptive critique of each. The study held the number of prototypes, the feedback and the total time equal, and changed only the arrangement: the serial group was criticised after each prototype, while the parallel group made three, took criticism on all three, then made two more and took criticism again. The finished advertisements were judged twice, by independent experts and by a real campaign. The parallel advertisements drew 445 clicks per million impressions, against 398 for the serial group, and the experts rated them higher.

The authors call the serial failure fixation. It is the only study here that compares finished designs. Its limits are novices, a short task and brief sessions.

Dow, Fortuna, Schwartz, Altringer, Schwartz and Klemmer (2011) moved the comparison to teams. Participants designed advertisements alone, then met a partner. They shared three designs, the best of three, or one. Sharing multiple designs improved the outcome, exploration, critique quality and rapport.

Fixation is measurable, and a warning does not remove it

If parallel prototyping works, the reason is usually said to be that it blocks fixation. Jansson and Smith (1991) ran experiments to test whether design fixation is a measurable barrier. They define it as a blind adherence to a set of ideas that limits conceptual design.

Smith, Ward and Schumacher (1993) supplied the demonstration. Participants were asked to invent creatures for an uninhabited planet just like Earth, drawing front and side views and labelling the parts; some saw three examples beforehand, sharing three features, four legs, two antennae and a tail, which as Marsh, Landau and Hicks (1996) describe them have no necessary connection with one another. The designs of those who had seen the examples carried those features more often than the designs of a control group. The conformity did not fade when a twenty-three minute unrelated task intervened. Telling participants to produce ideas very different from the examples did not reduce it, while telling them to conform increased it.

Chrysikou and Weisberg (2005) asked whether the effect can be addressed. In one experiment 89 participants worked under control, fixation or defixation instructions. People followed the example even when it contained inappropriate elements. Defixating instructions reduced the effect.

Purcell and Gero (1996) widen the frame. Design fixation is a case of the general failure to restructure a problem, the family that includes functional fixedness. Any move that forces a fresh structure, including rival candidates, should help.

Examples both fixate and inspire, and the two effects coexist

Marsh, Landau and Hicks (1996) tested the conformity effect directly. Across three experiments they observed conformity to the examples. They could not confirm that it constrained creative output, because total output, elaborateness and noncritical features did not differ as predicted. A delay increased conformity.

Sio, Kotovsky and Cagan (2015) pooled the field, meta-analysing 43 studies and 1229 participants. Examples led people to generate more example-related ideas, and fewer categories of ideas. The ideas were rated more novel, and quality rose with the degree of copying. The benefit was largest for few and uncommon examples.

Perttula and Sipila (2007) reached a compatible conclusion from inside design. Exposure to other people's ideas could both stimulate and constrain what was produced, depending on how similar the exposure was to the task.

Vasconcelos and Crilly (2016) reviewed 25 inspiration and fixation studies. They identified 14 manipulated variables, and found great variety in methods and results. Shared constructs and stricter reporting, they argue, are needed before firm conclusions are possible.

Groups generate fewer ideas than the same people working alone

The rule is usually taught as if the ideas should come from a group. The evidence is more awkward than that.

Diehl and Stroebe (1987) ran four experiments on brainstorming groups. Nominal groups, working alone and later pooled, outperformed interacting groups. By manipulating blocking directly, the authors established that production blocking, waiting for a turn, accounted for most of the loss. Evaluation apprehension did not. A group assumes participants who can speak, hear and share one language.

Nijstad, Stroebe and Lodewijkx (2002) refined the mechanism. Participants generated ideas at a terminal. The authors manipulated whether the displayed ideas were diverse or homogeneous, and whether they arrived organised or at random. Exposure reduced the latencies of category changes. Diverse ideas widened the range produced, and many ideas from a few categories deepened production within them.

The image generator fixates more strongly than a plain example

Wadinambiarachchi, Kelly, Pareek, Zhou and Velloso (2024) ran a between-participants experiment with 60 people. They sketched chatbot avatar ideas while receiving no inspiration, an image search, or an AI image generator. The generator group fixed on the example more, and produced fewer, less varied and less original ideas than the no-inspiration group. The authors call the mechanism fixation displacement: attention moves onto the generator's output. With novices and one short session, this is a risk to manage rather than a rule.

Contested question: does the example fixate, or does the rule work for another reason?

One position holds that examples and initial ideas fixate. Jansson and Smith (1991) named the effect and showed it measurable. Smith, Ward and Schumacher (1993) showed that a warning does not remove it. Chrysikou and Weisberg (2005) showed that its size depends on framing. The other position holds that examples are as likely to inspire. Sio, Kotovsky and Cagan (2015) found ideas produced with examples more novel and of higher quality. Marsh, Landau and Hicks (1996) could not confirm that conformity constrained output. Perttula and Sipila (2007) found stimulation and interference in a single paradigm.

Vasconcelos and Crilly (2016) argue that the two sides are not measuring the same thing, which is why the dispute has lasted. What would settle it is a shared measure of fixation, a standard task, and the systematic manipulation of example number and commonness.

The second disagreement is closer to the rule. The parallel result of Dow and colleagues (2010) is usually read as evidence for making many candidates, and the experiment does not show that. Both conditions made five prototypes, so what differed was when the criticism arrived and how many unfinished prototypes stood beside each other when it did: the parallel designers had three candidates open at once, so criticism of one landed while the other two were still available for comparison. A design that varied the number of candidates and the grouping of feedback independently would decide whether plurality does any work of its own. That design would need both a fixation measure and a finished outcome, and that study has not been run.

The two sides differ in evidence strength as much as in conclusion. Jansson and Smith and Smith, Ward and Schumacher ran controlled experiments, but on small samples and short tasks. Sio, Kotovsky and Cagan pooled 43 studies and 1229 participants, so their finding that examples can raise novelty outweighs any single fixation demonstration. Marsh, Landau and Hicks show that the two effects can coexist. The strongest objection is that the rule's headline experiment does not test the rule: Dow and colleagues held the number of prototypes equal, so what they varied was when criticism arrived. While the dispute is open, a designer keeps several unfinished candidates visible at once, and measures fixation rather than counting ideas.

Where the argument is heading

The first is the move from whether to show examples to which examples, when and to whom, where Sio, Kotovsky and Cagan (2015) identified the moderators and Vasconcelos and Crilly (2016) the reporting a test would need. The second is the move from the individual to the group, where Dow and colleagues (2011) show that sharing several designs changes how a team reasons together. The third is the entry of the generator, where Wadinambiarachchi and colleagues (2024) show that an AI image generator can fixate more strongly than a plain example, so the rule is heading towards a conditional one: not many ideas, but which ideas, arranged how, offered to whom, and against which tool.

The sources this unit rests on

• Chrysikou, E. G. and Weisberg, R. W. (2005). Following the wrong footsteps: fixation effects of pictorial examples in a design problem-solving task. Journal of Experimental Psychology: Learning, Memory, and Cognition. doi:10.1037/0278-7393.31.5.1134 Chrysikou and Weisberg ran two experiments on pictorial examples; people followed the example even when it contained inappropriate elements, though defixating instructions reduced the effect.
• Diehl, M. and Stroebe, W. (1987). Productivity loss in brainstorming groups: toward the solution of a riddle. Journal of Personality and Social Psychology. doi:10.1037/0022-3514.53.3.497 Diehl and Stroebe ran four experiments comparing real and nominal brainstorming groups. Nominal groups were more productive, production blocking accounted for most of the loss, and evaluation apprehension did not. The group is not the best place to diverge.
• Dow, S. P., Fortuna, J., Schwartz, D., Altringer, B., Schwartz, D. L. and Klemmer, S. (2011). Prototyping dynamics: sharing multiple designs improves exploration, group rapport, and results. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI). doi:10.1145/1978942.1979359 Participants designed advertisements and then met a partner, sharing three designs, the best of three, or one. Sharing multiple designs improved the outcome, exploration, critique quality and rapport.
• Dow, S. P., Glassco, A., Kass, J., Schwarz, M., Schwartz, D. L. and Klemmer, S. R. (2010). Parallel prototyping leads to better design results, more divergence, and increased self-efficacy. ACM Transactions on Computer-Human Interaction. doi:10.1145/1879831.1879836 Novice designers created web advertisements in parallel or in series, each producing five prototypes and a final advertisement, with only the order of critique differing. The parallel advertisements performed significantly better.
• Jansson, D. G. and Smith, S. M. (1991). Design fixation. Design Studies. doi:10.1016/0142-694X(91)90003-F Jansson and Smith ran a series of experiments testing whether design fixation, a blind adherence to a set of ideas or concepts that limits conceptual design, is a measurable barrier. The results demonstrated the effect, and their definition is still the field's.
• Marsh, R. L., Landau, J. D. and Hicks, J. L. (1996). How examples may (and may not) constrain creativity. Memory and Cognition. doi:10.3758/BF03201091 Marsh, Landau and Hicks tested the conformity hypothesis in three experiments in which participants sketched new category members after seeing examples. Conformity appeared, but the authors could not confirm that it constrained output.
• Nijstad, B. A., Stroebe, W. and Lodewijkx, H. F. M. (2002). Cognitive stimulation and interference in groups: exposure effects in an idea generation task. Journal of Experimental Social Psychology. doi:10.1016/S0022-1031(02)00500-0 With participants generating ideas at a terminal, Nijstad and colleagues manipulated stimulus diversity and sequence. Exposure reduced the latency of category changes, and diverse ideas widened the range produced.
• Osborn, A. F. (1953). Applied Imagination: Principles and Procedures of Creative Problem-Solving. Charles Scribner's Sons. https://archive.org/details/appliedimaginati0000alex_p4f3 Osborn set out the divergence-then-converge procedure: deferred judgement, quantity before quality and the combination of others' ideas. The book is the practitioner origin of the studio habit, with no experimental evidence.
• Perttula, M. and Sipila, P. (2007). The idea exposure paradigm in design idea generation. Journal of Engineering Design. doi:10.1080/09544820600679679 Perttula and Sipila found that exposure to others' ideas in design idea generation could both stimulate and constrain what was produced, depending on how similar the exposure was to the task.
• Purcell, A. T. and Gero, J. S. (1996). Design and other types of fixation. Design Studies. doi:10.1016/S0142-694X(96)00023-3 Purcell and Gero review experiments on design fixation and connect it to functional fixedness in general problem solving. Design fixation, they argue, is the general failure to restructure a problem rather than a peculiarity of design.
• Sio, U. N., Kotovsky, K. and Cagan, J. (2015). Fixation or inspiration? A meta-analytic review of the role of examples on design processes. Design Studies. doi:10.1016/j.destud.2015.04.004 Sio, Kotovsky and Cagan meta-analysed 43 studies with 1229 participants. Examples produced more example-related and fewer categories of ideas, but the ideas were more novel and quality rose with copying.
• Smith, S. M., Ward, T. B. and Schumacher, J. S. (1993). Constraining effects of examples in a creative generation task. Memory and Cognition. doi:10.3758/BF03202751 In three experiments participants sketched new category members after seeing examples. The example group's designs were more likely to carry the examples' features, and neither an intervening task nor an instruction to differ reduced the conformity.
• Vasconcelos, L. A. and Crilly, N. (2016). Inspiration and fixation: questions, methods, findings, and challenges. Design Studies. doi:10.1016/j.destud.2015.11.001 Vasconcelos and Crilly reviewed 25 inspiration and fixation studies using a similar approach. They identified 14 manipulated variables and found great variety in methods and results, and argue that shared constructs and stricter reporting are needed.
• Wadinambiarachchi, S., Kelly, R. M., Pareek, S., Zhou, Q. and Velloso, E. (2024). The effects of generative AI on design fixation and divergent thinking. Proceedings of the CHI Conference on Human Factors in Computing Systems (CHI). doi:10.1145/3613904.3642919 Wadinambiarachchi and colleagues ran a between-participants experiment with 60 people sketching chatbot avatar ideas with no inspiration, an image search, or an AI image generator. The generator group fixed on the example more and produced fewer, less varied ideas.