Cultural Data Annotator – MARS Project

About the role

The Cultural Data Annotator supports the MARS Project by contributing authentic, human-generated data that reflects the cultural values of Singapore. Working independently and remotely, the annotator completes a two-stage workflow: survey questionnaires followed by structured cultural value annotation of LLM-generated responses.

Key responsibilities

• Complete all survey questionnaires accurately and in full before beginning annotation
• Select culturally relevant topics that genuinely reflect current norms and values in Singapore
• Articulate clear, ranked cultural value principles for each selected topic
• Create, select, and refine value-related questions in a balanced mix across all three question types
• Review and annotate LLM-generated responses to align with both cultural and personal values
• Maintain strict consistency between value labels and their corresponding text throughout annotation
• Ensure all annotations reflect authentic, current cultural norms — not stereotypical or outdated generalisations
• Complete 12 sets of annotation tasks
• Comply with all quality, confidentiality, and anti-fraud requirements of the MARS Project

About you

• Native-level proficiency in English (Singapore) with deep knowledge of Singaporean cultural values (written and spoken)
• Ability to clearly distinguish between cultural perspectives and personal perspectives when evaluating AI-generated text
• Strong critical evaluation skills and attention to value-text consistency throughout annotation
• Ability to reflect current, realistic cultural norms without relying on stereotypes
• Familiarity with Large Language Models (LLMs) such as ChatGPT or similar is helpful but not required
• Self-directed and able to work independently in a fully remote setting

About us

Centific AI Data Solutions commissions the MARS Project, a large-scale cultural data annotation initiative that collects high-quality textual data reflecting the values of specific cultures, which is used to train and align Large Language Models (LLMs) such as ChatGPT, Gemini, Claude, and Sea-Lion. Contributors provide genuine human perspectives that reflect real, current cultural norms — enabling AI systems to understand and respond in culturally appropriate ways.