A Bilingual Synthetic Dataset for Elderly Emotional Wellbeing and Agentic AI Support in Low-Resource Contexts

The dataset consists of 2,000 synthetic records, divided equally into two subsets: 1,000 English records and 1,000 Roman Urdu records. Each record represents a simulated elderly participant aged between 50 and 90 years. Both subsets share an identical schema comprising 13 attributes, enabling direct cross-lingual comparison and analysis.

No missing values are present in the dataset. Summary statistics confirm identical distributions across both language versions, indicating controlled generation and minimizing unintended linguistic bias.
Dataset Attributes

Each record includes the following attributes:

Scenario_ID: Unique identifier representing the contextual scenario

Age: Participant age (50–90)

Gender: Male or Female

Emotional_Tone: Annotated dominant emotional state

Cognitive_Difficulty: Self-reported difficulty level (1–5)

Coping_Strategy: Emotional or cognitive regulation behavior

Learning_Outcome: Outcome of AI interaction

Support_Needed: Type of AI assistance required

Technology_Interaction: Mode of technology use

Reflection_Tag: High-level reflective keyword

Depression_Level: Emotional difficulty indicator (0–10)

Perceived_Usefulness: AI usefulness rating (0–10)

AI_Utilization_Purpose: Primary intent of AI usage

Participant_Response: Free-text response in English or Roman Urdu