Shuichiro Ogawa
日本語

Notes

Research, experiments, and reflections on AI and design, published along with the process of making them.

Back to the monthly list

Creativity (7)

Where new ideas come from.

  • Does Quantity Breed Quality? Hits and Hit Rates, Idea Counts and the Best Idea, How Prototypes Are Compared, Practice Time, and the Origin of the Ceramics Class Story

    2026-11-12

    The saying that quantity breeds quality was examined separately for four kinds of quantity (number of works, number of ideas, prototype iteration, and practice time), using 35 academic sources and 5 documents on the origin of a well-known anecdote. People who produce more works also produce more hits. In data on about 1.5 million inventors, 1,742 psychologists, and 20,296 scientists, the correlation between total output and total citations was .59 to .77. The correlation between total output and average citations per item, however, was −.0004 to .07, so prolific producers did not make better individual works (Forthmann et al. 2020). The most influential work appears at a random position within a career (Sinatra et al. 2016; Janosov et al. 2020), and no change in productivity was detected during hot streaks (Liu et al. 2018). When quality is measured as the share of papers in the top 1% by citations, researchers with more papers had higher shares, with diminishing returns in recent cohorts (Larivière and Costas 2016, 28 million researchers). In idea generation, a quantity goal increased the number of good ideas (Paulus et al. 2011), and later ideas were more creative (Beaty and Silvia 2012). However, people had difficulty selecting original ideas from their own output (Rietzschel et al. 2010), and the quality of the best idea depended on the average, the variance, and the ability to discern quality as well as on the number of ideas (Girotra et al. 2010). In design prototyping, a parallel condition outperformed a serial condition even though both conditions produced five prototypes (Dow et al. 2010). Practice time explained 14% of the variance in performance after the 2018 correction, ranging from 1% to 24% across domains and falling to 1% among elite athletes (Macnamara et al. 2014, 2016). Ericsson himself defined deliberate practice by its design under a teacher rather than by its amount. The ceramics class story was adapted from a method that the photographer Jerry Uelsmann used in his beginning photography class, which Orland heard from Uelsmann and rewrote as a ceramics class (Orland's email, quoted in a footnote by Clear). No grade records or comparisons were found. The book explains the result by learning from mistakes, but Uelsmann himself described it as producing enough images and then editing back to the best ones. The academic literature strongly supports the claim that producing more yields more hits. Whether producing more makes each work better depends on the domain, the measure, and the content of the practice.

  • What Does It Take to Produce Novelty Now? Separating Traits, Attitudes, Stances, Abilities, and Skills

    2026-09-26

    LLM-generated research ideas outscored experts on novelty, yet after execution they lost more ground and the rankings flipped (Si et al.); scientists who use AI publish and are cited far more, while the range of topics science covers shrinks (Hao et al., Nature 2026). Now that ideas that look new are no longer scarce, this note asks what a person who produces novelty needs, separating what is asked of people into traits, attitudes, stances, abilities, skills, and career. It draws on 56 newly collected academic works (36 of which are used in the text, together with 13 other works, most of them verified in sister notes) and 70 recent sources from AI providers, public bodies, research firms, and leading researchers (22 used in the text). On traits, curiosity correlates with creativity at r = .41 but only .16 with creativity rated by others; openness correlates at .22 and a pro-risk attitude at .08. Of the two facets of grit, perseverance of effort positively predicts creative achievement while consistency of interest predicts it negatively. On abilities, divergent thinking predicts creative achievement only weakly (r = .216); people can largely identify their own best ideas, but they recognize creative ideas less well when they want to reduce uncertainty. Problem-finding tasks elicited more original responses than divergent-thinking tasks (g = 0.887). On careers, high-impact work clusters in periods that follow exploration and then focus, and the pivot penalty (impact falling the further one moves from prior work) has grown over five decades. The average effect of training is g = 0.53, and methods that strengthen convergent thinking outperform those that strengthen divergent thinking. With the spread of generative AI, weight has shifted away from generation toward choosing ideas that survive execution, verifying outputs, confidence grounded in one's own domain competence, and choosing directions that do not erode collective diversity. AI providers' own materials leave to humans the setting of research goals, formulating problems so they can be verified automatically, selection, and checking against reality, and funding agencies have begun writing rules that keep judgment from being delegated to AI. No empirical study measuring the 'taste' that leading researchers keep invoking was found; this note splits it into choosing questions, evaluating and selecting, and holding domain competence, and maps each onto existing evidence.

  • What Social Coordination Costs Arise from Collaborating with Others on Creative Work

    2026-09-22

    Adding people does not make work progress in proportion to headcount; the difference disappears into coordination loss and motivation loss (Steiner 1972), and into a relational loss from feeling that it is harder to get help (Mueller 2012). Coordination cost is not a single kind; it needs to be seen as six kinds that arise for different reasons: synchronizing logistics and information, loafing, the awkwardness of being evaluated, conflict, reconciling differences in interpretation, and negotiating names and shares. In creative work, beyond the wait for one's turn in brainstorming (Diehl & Stroebe 1987), the content of the output itself — word choice, how a problem is framed — becomes an object of negotiation, which makes it hard to isolate coordination as purely a matter of logistics. Conventions and transactive memory make coordination cheaper, but in Broadway's collaboration networks the relationship between the degree of cohesion and success is an inverted U (Uzzi & Spiro 2005), and past a certain point cheaper coordination and better outcomes stop moving in the same direction. Replacing the partner with AI doesn't remove this either: a robot facilitator does not soften the effect of evaluation apprehension, loafing still occurs in human-AI teams, and while AI-sourced ideas raise the evaluation of individual works, the works come to resemble each other more (Doshi & Hauser 2024) — so it cannot be read as "with AI, social coordination cost is zero."

  • Recombination of the Known and Nonhuman Reasoning: Can Hallucination Be Treated as a Resource for Creativity?

    2026-07-22

    An essay that reorganizes the argument for treating hallucination as a resource for creativity through two lenses. The first is the lineage of combinatorial innovation and creativity, which holds that innovation is an unexpected recombination of known elements (Schumpeter's new combinations, Weitzman's recombinant growth, Arthur's combinatorial evolution, Uzzi's atypical combinations, Fleming, Campbell and Simonton's BVSR, Boden's three types, and Runco and Jaeger's standard definition). The second is the lineage of collective intelligence and diversity, which holds that because LLMs operate on reasoning principles different from those of humans, adding them to human groups may add diversity along a new, orthogonal axis (Hong and Page's diversity-trumps-ability, Page, Rahwan's Machine Behaviour, Krogh and Vedelsby's ambiguity decomposition, and, as counterevidence, Bender's stochastic parrots, Shumailov's model collapse, Peterson's knowledge collapse, Lorenz on social influence, and Thompson, Grim, and Singer on the Hong-Page theorem). After setting up both lenses in their strongest forms, the essay shows that the integrating thesis that binds them (a single mechanism of recombination on a nonhuman reasoning basis) runs into an internal contradiction between interpolation and a new axis unless an operation that moves sampling outside the original distribution is added, and it reduces the interlock to one falsifiable question rather than an established mechanism. It is a sequel (an essay) to the neutral literature map hallucination-as-creativity-resource-literature. 37 references. It is an argumentative reorganization, and primary empirical evidence for the affirmative side is currently missing (the limits of scope are gathered at the end).

  • An Academic Map of Research That Treats Hallucination as a Resource for Creativity

    2026-07-21

    A literature map that surveys peer-reviewed research and preprints positioning generative-AI hallucination (confabulation) not as a defect to be suppressed but as a starting point for ideation and creativity. It connects, cluster by cluster: the positive reframing of hallucination (the value of confabulation in Sui et al., valuable hallucination in Chen & Wang, the creativity-perspective survey by Jiang et al., the decoding-layer tradeoff in He et al., the usefulness in drug discovery in Yuan et al.); LLM creativity evaluation and output diversity (Stevenson et al., Bellemare-Pepin et al., the homogeneity of Wenger & Kenett, diversity reduction under RLHF in Kirk et al., diversity collapse at the SFT stage in Karouzos et al., min-p sampling in Nguyen et al.); ideation support and design fixation (AI-augmented brainwriting in Shaer et al., design fixation in Wadinambiarachchi et al., CST in Shneiderman); scientific hypothesis generation (Shahhosseini et al.); idea homogenization and prompt-driven diversity recovery (Girotra et al., Meincke et al.); and the classics of creativity psychology (Guilford, Mednick, Beaty & Kenett). A neutral prior-work review. Thirty-one references (bibliographies confirmed reachable; caveats aggregated at the end). The view that treats the deviations of older, lower-precision models as a resource is thinly represented in explicit research, and is treated here as an implication drawn from research on the performance/diversity tradeoff.

  • The Creativity of Subtraction: The Additive Bias, and the Hypothesis of Designing Absence

    2026-07-19

    A single deep argument that inverts one asymmetry creativity and design quietly assume, namely that addition (multiplying options, adding elements) is generative while subtraction (removing, withholding, not-making) is a secondary work of convergence, using the same symmetry-completion move by which de Broglie derived matter waves. The inversion says subtraction is a first-class generative operation symmetric to addition, and that absence and the un-made carry structure not reducible to filtering. This is offered not as a settled fact but as a falsifiable hypothesis, testable by comparative experiments and tool audits. The additive bias has already been demonstrated on the problem-solving side (Adams, Converse, Hales & Klotz 2021, Nature); the core proposal treats that as the 'photon shown on one side' and moves the target to the symmetry of generative operations in design theory. Novelty is strictly limited to two points: the theoretical symmetrization that raises subtraction to a first-class generative operation, and the methodological audit of the additive bias in creativity metrics and authoring tools. 24 references (20 with confirmed bibliography, 4 flagged for primary verification).

  • Creativity, Situated Cognition, and Environment: A Literature Map of 4E Cognition, Affordances, and Situated Learning

    2026-07-19

    A survey of published scholarship that treats creativity and cognition not as inner individual traits but as distributed across body, environment, situation, and relations with others. It runs from Gibson's affordances (1979) and Norman's adaptation for design (1988), through the embodied mind of Varela, Thompson, and Rosch (1991) and the extended mind of Clark and Chalmers (1998), to Lave and Wenger's legitimate peripheral participation (1991), Hutchins's distributed cognition (1995), Vygotsky's zone of proximal development (1978), and Wood, Bruner, and Ross's scaffolding (1976), and on to a critical view of measuring ability apart from environment (Ross 1977, Merton 1968, Bourdieu & Passeron 1977, Sandel 2020). 14 references.