Real-world creative processes ranging from art to science rely on social feedback-loops between selection and creation. Yet, the effects of popularity feedback on collective creativity remain poorly understood. We investigate how popularity ratings influence cultural dynamics in a large-scale online experiment where participants ($N = 1\,008$) iteratively \textit{select} images from evolving markets and \textit{produce} their own modifications. Results show that exposing the popularity of images reduces cultural diversity and slows innovation, delaying aesthetic improvements. These findings are mediated by alterations of both selection and creation. During selection, popularity information triggers cumulative advantage, with participants preferentially building upon popular images, reducing diversity. During creation, participants make less disruptive changes, and are more likely to expand existing visual patterns. Feedback loops in cultural markets thus not only shape selection, but also, directly or indirectly, the form and direction of cultural innovation.
Many instances of herd behavior such as fads and fashions are attributed to positive feedback loops amplifying and propagating potentially arbitrary choices in collectives (Bikhchandani et al., 1992;Kameda & Hastie, 2015). In cultural evolution, examples of such feedback loops include positive frequencydependent selection, when the adoption rate of cultural variants increases with their incidence among a population, due to e.g. conformity biases (Newberry & Plotkin, 2022), or increased utility, as in the case of "network goods" (Björkegren, 2018). A related phenomenon is cumulative advantage, when popular cultural goods that are already popular attract more attention, independently from their intrinsic worth. Cumulative advantage can amplify arbitrary differences in cultural success: for instance, Salganik et al. (2006) have shown that access to popularity information in artificial markets of songs ultimately led to more unequal and unpredictable market shares. However, the effects of such feedback loops in cultural markets involving both consumers and producers of cultural goods has been far less studied experimentally, with the exception of Balietti et al. (2016), who demonstrated that competitive incentives influence peer-review feedback processes involving both creators and reviewers. However, this study did not consider popularity information (a common source of social feedback), and it did not directly examine how social information affects the creative output itself. In fact, cultural evolutionary literature has often neglected the incidence of feedback loops on the form and direction that innovations take, reflecting a broader neglect of the "producer's standpoint" (André et al., 2023).
Yet, in culture, feedback loops acting on selection and creation together determine outcomes such as the amount of diversity that is ultimately generated and sustained or the quality of cultural innovations. For instance, phenomena such as cumulative advantage may distort perceptions of innovations’ fitness (Salganik & Watts, 2008;Salganik et al., 2006) or long-term potential, and alter the space of possibilities that producers are willing to explore. In fact, social feedback may alter innovation strategies in ways unappreciated by blind variation models (Campbell, 1960). For instance, we hypothesize that access to social feedback encourages producers to amplify or exaggerate arbitrary features present in successful cultural goods. This could create a “cultural runaway” (Boyd & Richerson, 1988;Prum, 2012) analogue to biological runaway selection, an evolutionary process responsible for exaggerated ornaments across several species and driven by a positive feedback between a trait and preference for this trait (Lande, 1981;Lehtonen & Kokko, 2012).
We explore these hypotheses through a large-scale online experiment in which participants (𝑁 = 1 008) iteratively observe a cultural market of images and populate it with their own drawing after having selected an image to improve. Our experiment thus implements both selection and creation steps, therefore capturing the two major forces of cultural evolution (Mesoudi, 2021). We find that access to popularity information (about how many times items were copied) leads to less diverse cultural markets; slower innovation; and initially less aesthetic and creative images, although this trend may be reversed in the long-run, possibly due to social feedback protecting high-quality innovations. These effects are mediated by alterations of both selection and creation. During selection, popularity information triggers a process of cumulative advantage, whereby participants preferentially choose to build upon popular ideas, which undermines diversity and amplifies popularity inequalities. Moreover, in line with our cultural runaway hypothesis, individuals make less disruptive changes when provided with popularity information, and they are more likely to extend existing visual patterns instead.
We develop a large-scale collective experiment in which participants observe collections of images (16×16 grids with black or white pixels), select one image to modify, and contribute their creation to the collection (Fig. 1). The editing interface was similar to other pixel-based drawing experiments in the literature (Hart et al., 2017;Kumar et al., 2022;Tchernichovski et al., 2025). In the treatment group, during the selection step, participants observe the popularity of each item in the market (i.e. how many times the item had been previously selected to be edited). Participants are asked to behave so as to maximize the chance that their own work is selected by subsequent participants. This creates a strategic tension: while participants may choose to capitalize on ideas that have proven successful in the past, doing so implies competing with many similar images for attention, decreasing the chance of being selected. This dilemma thus invites either conformist strategies (building on popular
This content is AI-processed based on open access ArXiv data.