LLM-Driven Adjustments in Serious Games: A Feasibility Analysis

Area of Science:

• Neuroscience
• Artificial Intelligence
• Rehabilitation Medicine

Background:

• Serious games (SGs) and telerehabilitation are crucial for neurological recovery.
• Personalization and adaptive difficulty are key features for effective rehabilitation.

Purpose of the Study:

• To explore integrating a large language model (LLM) into an Assessment Serious Game (ASG).
• To analyze exercise data and generate personalized rehabilitation recommendations.
• To assess the LLM's capability in real-time SG adjustments and parameter suggestions.

Main Methods:

• Acquired medical knowledge from professionals to define target pathologies and parameters.
• Integrated the ASG with GroqCloud, utilizing a specialized LLM prompt.
• Designed the prompt to function as a virtual physiotherapist and SG developer for real-time adjustments.

Main Results:

• The LLM demonstrated effectiveness in recognizing real-time adjustments and following parameter setting instructions.
• Limitations were observed in the precision of adjustments and numerical parameter recommendations.
• Preliminary testing confirmed the system's basic capabilities.

Conclusions:

• A designed LLM prompt is feasible for adjusting SG difficulty and recommending setup parameters in telerehabilitation.
• Further improvements are needed in LLM reliability and accuracy for clinical application.
• This approach shows potential for enhancing personalized neurological rehabilitation through AI.