Generative agent simulations operate at two scales: individual personas for character interaction, and population models for collective behavior analysis and intervention testing. We propose a third scale: meso-level simulation - interaction with group-level representations that retain grounding in rich individual experience. To enable this, we present Synonymix, a pipeline that constructs a "unigraph" from multiple life story personas via graph-based abstraction and merging, producing a queryable collective representation that can be explored for sensemaking or sampled for synthetic persona generation. Evaluating synthetic agents on General Social Survey items, we demonstrate behavioral signal preservation beyond demographic baselines (p<0.001, r=0.59) with demonstrable privacy guarantee (max source contribution <13%). We invite discussion on interaction modalities enabled by meso-level simulations, and whether "high-fidelity" personas can ever capture the texture of lived experience.
Figures & tables
Figure 1. The central artifact of Synonymix is the "unigraph", a semantically unified knowledge graph constructed from each participant’s narrative persona. This artifact, which captures the shared personal history of the collective, can be explored for greater understanding or sampled for synthetic representation, and offers a discrete structure upon which formal privacy guarantees can be introduced for the participants. It is visualized here with n=4 for readability. This figure depicts the central artifact of Synonymix, which is a 3-tiered graph. The three layers, from bottom to top, are comprised of subject nodes, factual nodes, and interpretive nodes respectively. Each node is colored according to the participant from which it originated, and linked to nodes from all similar concepts across the entire agent bank. There is a sidebar for this simple visualization with n=4, which notes categories for the colors: respondent_1498 (48 unique), respondent_2529 (55 unique), respondent_2574 (47 unique), respondent_2742 (55 unique), Merged (multiple personas). It also unigraph stats: \epsilon(epsilon): N/A | DP: No; Total nodes: 226 (S:76, F:83, I:67); S merge rate: 10.5%; F merge rate: 0.0% I merge rate: 17.9%
Figure 2. Unigraph Traversal on 4 Personas
Figure 3. TOP: Box-plot comparison of aggregate enrichment (D→L) and transformation (L→F) distances across all GSS items. Diamonds indicate means; horizontal lines indicate medians. Two enrichment outliers ( 0.4 , 0.5 ) for nominal questions not shown for scale. BOTTOM: Per-item comparison of transformation vs. enrichment distance. Points below the diagonal indicate items where the pipeline loses less signal than narrative expansion adds.