Comprehensive survey · 2026

Social Intelligence Modeling A Comprehensive Survey from Social Perception to Social Simulation

Zikai Song*, Xiajie Li*, Yunyao Zhang, Xinglang Zhang, Wei Yang, Junqing Yu

Huazhong University of Science and Technology

We organize social intelligence modeling as four organically connected stages—perception, understanding, generation, and simulation—forming a coherent path from observable social signals to evolving social systems. A searchable collection aligned with this taxonomy makes related work easier to locate across stages.

Community

Join the conversation

Discuss the survey or find the current WeChat group entry.

GitHub Discussions

Join the survey discussion

Ask questions, suggest papers, and discuss the taxonomy with the AI4SS community.

Open discussions

WeChat Group

Join the AI4SS survey group

Open the dedicated thread for joining information maintained by the AI4SS community.

View WeChat group

Definition

What is SIM?

AI has studied society for years, yet the field remains fragmented. Research on social intelligence appear in different areas such as sentiment analysis, misinformation detection, user modeling, content generation, multi-agent systems, and social simulation. Because these studies use different task names and keywords, it is difficult to see how they connect or to form a clear view of the field as a whole.

Our Social Intelligence Modeling (SIM) brings these fragments together in one framework. By organizing nearly 1,000 papers across 20+ tasks, we show how AI models social content, people, relations, and social evolution. SIM structures the field into four connected levels: Perception, Understanding, Generation, and Simulation, providing a unified view of how AI engages with social systems.

Perception

Recognize social signals and states in content, users, and multimodal interactions.

Understanding

Model social structure, discourse, networks, communities, and behavioral patterns.

Generation

Produce socially situated content, dialogue, counter-speech, and interventions.

Simulation

Reconstruct dynamic social processes and explore plausible collective futures.

Overview of the Social Intelligence Modeling framework from social perception to social simulation
Survey framework A continuous path from observable signals to simulated social dynamics.

Taxonomy and statistics

Task distribution

Counts are generated from the selected survey database. Multi-category papers appear in every relevant task, while the headline total counts each paper once.

Four field-level views

Loading taxonomy…

Tasks by field

Fixed full-survey counts · click a task to filter the paper list

Chronological map

Literature by field over time

Each dot is one paper positioned by publication date. Rows correspond to the four survey fields; select a row, year, or paper to update the explorer.

Open the complete taxonomy architecture
Complete category architecture for the Social Intelligence Modeling survey

Paper collection

Search and filter papers

Search by title, author, venue, task, abstract, or summary.

Browse all papers
/
Loading the paper atlas…

Paper submission

Suggest a paper

Submit structured metadata for review. A suggestion is added to the collection only after validation and maintainer approval.

  1. 1
    Submit

    Paste Zotero metadata or fill only the paper basics.

  2. 2
    Review

    A maintainer checks scope, duplicates, and completes the taxonomy.

  3. 3
    Publish

    Accepted data updates the collection and website.

Paper intake

Required fields are marked *

Optional · quickly fill available metadata from one copied Zotero item
Optional supporting files Pipeline figures or the paper PDF
Pipeline images optional · choose, drop, or paste · 4 × 10 MiB
0 / 4 images selected
Paper file optional · choose or drop PDF · 50 MiB
No PDF selected

This creates a public GitHub issue that remains open as the submission record. After submitting, select GitHub's Subscribe control; approval status, validation results, processing updates, and maintainer feedback are posted there. Do not include a private email address.

Prefer GitHub's native issue form?

Upload interrupted

Choose how to continue

Completed uploads are kept. Continue retries only the unfinished files; alternatively, open the prefilled GitHub Issue and attach the unfinished files manually.

Metadata helper

Import from Zotero

  1. Select one item in Zotero.
  2. Use the one-click metadata plugin to copy its JSON.
  3. Paste below; recognized fields are filled without removing existing form fields.
Zotero context menu with Copy Meta to Json Format highlighted
Right-click the selected Zotero item and choose “Copy Meta to Json Format”.
Plugin not installed? Download the Zotero plugin