LAB2FAB - Human-in-the-Loop Expert System for SME Sustainability

The repository provides the datasets, prompt templates, and evaluation artefacts used in the development and assessment of LAB2FAB, a human-in-the-loop expert system that combines sustainability assessment, knowledge-based inference, generative AI, expertise retrieval, and human supervision.

Repository contents: • README.md - Description of the dataset package, structure, and anonymization procedures. • 01_sdg_criteria.md - Seven SDG-aligned sustainability assessment criteria in Portuguese, together with their associated Sustainable Development Goals and targets. • 02_prompt_templates.md - Prompt templates used during the experimental evaluation, including keyword extraction, keyword refinement, gap generation, and recommendation generation prompts across the evaluated prompting strategies. • 03_sus_item_distribution.csv - Item-level response distribution for the System Usability Scale (SUS) evaluation (N=10). • 03_sus_summary.md - Summary statistics of the SUS evaluation, including overall usability measures reported in the study. • 04_insitu_ratings_by_organisation.csv - Aggregated organisational ratings for generated artefacts collected during the in-situ evaluation. • 04_insitu_ratings_by_dimension.csv - Aggregated ratings by evaluation dimension (keywords, gap statements, and recommendations) covering 113 assessed gaps.