Human-AI Co-Design Study for Stroke Rehabilitation Agents

564.3 KB PDF details a Design-Based Research methodology combining synthetic patient generation and real-code execution for developing Embodied Conversational Agents (ECAs). The study, authored by Anna Lekova, presents results from a pilot deployment using a Furhat robot, including lexical diversity (MTLD ≈ 134) and usability scores (M = 86.67 for relatives).

Use Cases

• Analyze lexical diversity (MTLD) and repetition (Yule's K) metrics from LLM responses to evaluate conversational quality.
• Compare usability scores and Intrinsic Motivation Inventory scores between patient relatives and rehabilitation professionals.
• Examine the methodology of synthetic patient generation for prototyping LLM-based conversational pipelines in rehabilitation contexts.

Strengths

• Includes specific quantitative results like lexical diversity (MTLD ≈ 134) and usability scores (M = 86.67).
• Presents a novel co-design methodology validated through a pilot deployment with a physical Furhat robot.
• Released under a permissive CC BY 4.0 license for reuse.

Limitations

• Dataset is a 564.3 KB PDF document, not a structured tabular dataset, limiting direct computational analysis.
• Sample size for the pilot deployment is not specified, limiting statistical generalizability.
• Data is confined to a specific research context of post-stroke rehabilitation, not a broad clinical dataset.

Provenance

Source

Anna Lekova via figshare.

Collection Method

Design-Based Research combining synthetic patient generation with a pilot deployment using a Furhat robot.

Time Range

null

Freshness

Last updated March 2026.

Geography

null