LLM-Based Heuristic Evaluations of High- and Low-Fidelity Prototypes (GPT-4o)

This dataset contains structured usability evaluation data for a range of user interface prototypes, assessed using Nielsen’s 10 heuristic principles. It includes:

• Prototype_Metadata.csv – Metadata for each prototype, specifying its name, fidelity level (high or low), domain, and source URL.
• IRR_Results.csv – Inter-Rater Reliability (IRR) results comparing human evaluator agreement across ten heuristics.
• EvaluationTemplate.rtf – A standardized evaluation template used by both human raters and GPT-4o to ensure consistent heuristic assessment.
• LLM_Evaluation folder – Structured evaluations conducted by GPT-4o across two phases (LLM Evaluation 1 and LLM Evaluation 2), each divided into high-fidelity and low-fidelity prototypes. Each prototype file contains issue identification, severity ratings (0–4), and heuristic-based recommendations.

This dataset supports research on AI-assisted usability evaluation, enabling comparison between LLM and human assessments, and analysis of prototype usability across domains and fidelity levels.