metadata
dataset_info:
- config_name: opinion
features:
- name: question
dtype: string
- name: selections
struct:
- name: Andorra
list: float64
- name: Argentina
list: float64
- name: Armenia
list: float64
- name: Australia
list: float64
- name: Bangladesh
list: float64
- name: Bolivia
list: float64
- name: Brazil
list: float64
- name: Canada
list: float64
- name: Chile
list: float64
- name: Colombia
list: float64
- name: Cyprus
list: float64
- name: Czechia
list: float64
- name: Ecuador
list: float64
- name: Ethiopia
list: float64
- name: Germany
list: float64
- name: Greece
list: float64
- name: Guatemala
list: float64
- name: Hong Kong SAR
list: float64
- name: India
list: float64
- name: Indonesia
list: float64
- name: Japan
list: float64
- name: Kenya
list: float64
- name: Libya
list: float64
- name: Malaysia
list: float64
- name: Maldives
list: float64
- name: Mexico
list: float64
- name: Mongolia
list: float64
- name: Morocco
list: float64
- name: Netherlands
list: float64
- name: New Zealand
list: float64
- name: Nicaragua
list: float64
- name: Nigeria
list: float64
- name: Peru
list: float64
- name: Philippines
list: float64
- name: Puerto Rico
list: float64
- name: Romania
list: float64
- name: Russia
list: float64
- name: Serbia
list: float64
- name: Singapore
list: float64
- name: Slovakia
list: float64
- name: South Korea
list: float64
- name: Taiwan ROC
list: float64
- name: Tajikistan
list: float64
- name: Thailand
list: float64
- name: Tunisia
list: float64
- name: Turkey
list: float64
- name: Ukraine
list: float64
- name: United States
list: float64
- name: Venezuela
list: float64
- name: Zimbabwe
list: float64
- name: options
list: string
- name: source
dtype: string
splits:
- name: train
num_bytes: 204742
num_examples: 300
download_size: 120595
dataset_size: 204742
- config_name: persona
features:
- name: Country
dtype: string
- name: Language
dtype: string
- name: Count
dtype: int64
- name: Participants
list:
- name: Age
dtype: int64
- name: Education Level
dtype: string
- name: Gender
dtype: string
- name: Marital Status
dtype: string
- name: Occupational Group
dtype: string
- name: Political Preference
dtype: string
- name: Religion
dtype: string
- name: Social Class
dtype: string
splits:
- name: train
num_bytes: 6486294
num_examples: 50
download_size: 195709
dataset_size: 6486294
configs:
- config_name: opinion
data_files:
- split: train
path: opinion/train-*
- config_name: persona
data_files:
- split: train
path: persona/train-*
MAPS: Multilingual Agent Perspective Survey
Check out our Multilingual Society Simulation research paper at arXiv, and code implementation at GitHub.
Get Started
from datasets import load_dataset
def load_global_opinions_question():
ds = load_dataset("magicgh/MAPS", "opinion", split="train")
return ds
def load_global_opinions_persona():
ds = load_dataset("magicgh/MAPS", "persona", split="train")
return ds
Citation
@article{masim,
title={MASim: Multilingual Agent-Based Simulation for Social Science},
author={Zhang, Xuan and Zhang, Wenxuan and Wang, Anxu and Ng, See-Kiong and Deng, Yang},
journal={arXiv preprint arXiv:2512.07195},
year={2025}
}