02 UIST2022 Social-Simulacra

PDF · 28 Aug 2026

The paper introduces the concept of social simulacra—large, synthetic populations of personas, posts, and replies generated by large language models—to support the design and evaluation of social computing systems. The authors present SimReddit, a prototype tool that leverages GPT‑3 to (1) GENERATE realistic user profiles and content, (2) explore WHAT‑IF scenarios by editing community rules or moderation policies and instantly observing simulated behavioral changes, and (3) create MULTIVERSE collections of alternative community evolutions for comparative analysis. The work details prompting strategies for scaling persona creation, methods for chaining prompts to produce coherent thread‑level interactions, and mechanisms for controlling demographic and behavioral attributes. Two empirical studies assess (a) technical plausibility—human judges rate generated posts and replies as plausibly human‑written—and (b) design utility—professional designers use SimReddit to iterate on moderation policies, reporting that the tool helped surface unintended consequences and refine community guidelines. The authors discuss ethical considerations, limitations of language‑model‑based simulation, and future directions such as integrating multimodal content, richer behavioral models, and user‑in‑the‑loop validation. Overall, the paper argues that synthetic social simulacra can accelerate early‑stage prototyping, reduce reliance on costly live deployments, and enable systematic exploration of design spaces in social computing.

Topics

Social Simulacra: Synthetic Populations as Design Prototypes

Introduces the notion of social simulacra—large, AI‑generated collections of personas, posts, and replies that emulate real‑world community dynamics—positioned as reusable design artifacts for exploring social computing systems.

SimReddit: Architecture and Core Functionalities

Describes the SimReddit prototype, detailing its three primary modes—GENERATE (persona and content creation), WHAT‑IF (policy editing with instant behavioral simulation), and MULTIVERSE (parallel community evolutions for comparative analysis)—and the overall system pipeline.

More from Joon

PDF

03_2024_LLM-Agents-Grounded-Self-Reports.pdf

1mo ago
PDF

04_CHI2022_Jury-Learning.pdf

1mo ago
PDF

01_UIST2023_Generative-Agents-Interactive-Simulacra.pdf

1mo ago
PDF

LLM Agents Grounded in Self-Reports Enable General-Purpose Simulation of Individuals

1y ago
PDF

Generative Agents: Interactive Simulacra of Human Behavior

3y ago
PDF

Social Simulacra: Creating Populated Prototypes for Social Computing Systems

4y ago
See all content