Benchmarking Dataset for AI Communication in Cosmetic Dentistry

Alaa Al-Haddad's proof-of-concept study provides a dataset of 160 AI-generated responses to 80 patient prompts across four cosmetic dentistry domains. The dataset, last updated in 2026, includes comparative evaluations of two large language models using a 20-point rubric assessing readability, practicality, empathy, structure, and safety. The data supports the feasibility of a multi-metric framework for evaluating patient-facing AI communication.

Use Cases

• Benchmarking AI communication quality based on the described multi-metric rubric
• Comparing model performance on patient-facing tasks based on the four cosmetic dentistry domains
• Analyzing the relationship between instruction configuration and output characteristics like readability and empathy

Strengths

• Includes 160 AI-generated responses from two distinct models for direct comparison
• Evaluations were performed by two board-certified specialists, suggesting expert review
• The 20-point rubric covers multiple communication dimensions beyond factual accuracy

Limitations

• The dataset is described in a PDF report; the underlying structured data files are not specified
• The dataset is small in scale at 781.5 KB, limiting its scope
• Row count and column-level documentation are absent, requiring manual inspection after download

Provenance

Source

Alaa Al-Haddad via figshare

Collection Method

Responses were generated from an 80-prompt test set derived from patient inquiries and evaluated by specialists.

Freshness

Last updated 2026-05-11 05:28:15