From semantic memory to collective creativity: A generative cognitive foundation for social creativity models

TL;DR

The paper tackles the problem of mechanistically modeling collective creativity by linking cognitive-level idea generation to social structure. It introduces a minimal, generative framework where agents share a semantic vocabulary and a common substrate, but differ in semantic topology via a single knob, the Watts–Strogatz rewiring probability $p_i$, which governs semantic modularity $Q(G_i)$. Idea generation uses a fixed length-$T$ random walk, and agents exchange ideation traces, enabling emergent cognitive stimulation and network redundancy without exogenous creativity traits. Key findings include that higher modularity reduces ideational breadth, lower pre-interaction overlap predicts larger stimulation gains, and shared inspiration sources increase network-level redundancy. The framework provides a principled sandbox for counterfactual experimentation on cognition–social structure interactions in collective creativity and supports mechanistic theory-building beyond traditional trait-based modeling.

Abstract

Simulation-based theory development has yielded powerful insights into collective performance by linking social structure to emergent outcomes, yet it has struggled to extend to collective creativity. Creativity is hard to capture purely at the social level, as novel ideas are generated through cognitive mechanisms. To address this gap, we introduce a multi-level socio-cognitive agent-based framework in which agents share a common semantic vocabulary and substrate but differ in semantic network topology. A single generative parameter tunes semantic modularity, yielding emergent individual differences in ideational breadth. When agents exchange ideation traces, two canonical social-creativity phenomena arise without being imposed: lower pre-interaction ideation overlap predicts larger stimulation gains, and shared inspiration sources induce network-level redundancy. The framework enables mechanistic theory-building about cognition and social structure in collective creativity.

Figures (4)

• Figure 1: Modularity $Q$ decreases as the Watts--Strogatz rewiring probability $p$ increases. We generate a shared $G_0$ with $|V|=100$ nodes and average degree $k=4$. Points show the mean $Q$ across $15$ graphs per $p$; shaded bands denote 95% bootstrap confidence intervals over graph replicates.
• Figure 2: Semantic modularity constrains exploratory access. Each point is a semantic graph ($N_G=500$). Higher modularity $Q(G_i)$ predicts lower expected ideational breadth ($\widehat{B}_i$). We use $30$ walk replicates per sampled prompt per graph. Line shows the fitted linear trend; shaded band the 95% CI.
• Figure 3: Lower pre-interaction overlap predicts larger stimulation gains. We plot the fixed-effects partial relationship between pre-interaction overlap and stimulation benefit after residualizing both variables by ordered-pair fixed effects and prompt fixed effects. We partition residualized overlap into $100$ quantile bins and plot, for each bin, the mean residualized overlap against the mean residualized gain. The fitted line has a slope equal to the estimated $\beta$ from Eq. \ref{['eq:stimulation_fe']}.
• Figure 4: Shared inspiration increases redundancy in recipients’ semantic exploration. Bars show mean post-inspiration recipient--recipient overlap across $495{,}000$ matched instances; whiskers denote 95% CI computed with SEs clustered by inspiration source; ***$p<0.001$.