Dataset: Perceived Reliability of an Explainable AI-Based Digital Educational Resource (Chatbot for Research Methodology Instruction)

This dataset contains the full evaluation data from a two-phase study assessing the perceived reliability and technical quality of an explainable, Generative AI-based digital educational resource (DER/GenAI) designed as a chatbot for research methodology instruction in higher education. The chatbot was built on the Dify platform using GPT-4o mini with a Retrieval-Augmented Generation (RAG) architecture. The study was conducted in two staged phases: (1) technical quality review by 10 experts (Phase 1), and (2) perceived reliability evaluation by 28 university faculty members (Phase 2). The dataset includes:- Likert responses (18 items in Phase 1; 18 items plus 2 global questions in Phase 2) on a 5-point scale assessing nine quality dimensions: explainability, accuracy, context understanding, clarity and language, relevance, coherence, detail, adaptability, and speed.- Demographic information (gender, education level, teaching experience, field of knowledge, prior experience with AI-based educational resources).- Optional free-text qualitative comments from participants, preserved in the original Spanish.- A complete data dictionary documenting all variables.- A README file with study design, reproducibility notes, and anonymization procedures. Instrument reliability: Phase 1 Cronbach's α = 0.929; Phase 2 α = 0.966. The dataset has been anonymized: no personally identifiable information was collected, timestamps have been removed, participant IDs are sequential codes (E01-E10 / F01-F28), and third-party personal names in qualitative comments have been redacted. This dataset supports a manuscript currently under peer review. The manuscript reference will be added upon acceptance.