Source-Grounded Research Workflow: Extraction, Indexing, and Synthesis

Cómo usar esta colección

This collection provides a strict workflow for researchers to conduct source-based analysis using AI as an objective data processor. The goal is to prevent hallucination by limiting the AI to the text you provide. **Stage 1: Raw Data Extraction** Begin with the **Scientific Data Extractor Prompt**. Copy this prompt into any standard LLM chat interface. Paste a scientific abstract containing numerical data (e.g., sample sizes, p-values, dosages). The prompt instructs the AI to output a structured Markdown table with columns for 'Raw Text Snippet', 'Variable/Concept', 'Value', 'Unit', and 'Context Note'. *Critical Constraint*: The AI must not interpret the significance of the data or summarize the study's conclusion. It records only what is explicitly stated. If units are implicit, they must be stated explicitly. This creates an auditable baseline of raw facts. **Stage 2: Evidence Mapping** Once you have your reading list, use the **Question-to-Source Index Builder**. Prepare a numbered list of specific research questions and paste your reading list excerpts (labeled with Source IDs like S1, S2) below the prompt. The AI will generate a traceable index mapping each question to exact quoted spans in your text. *Critical Constraint*: The AI is forbidden from browsing live links or inventing findings. If a question is unanswered by the provided text, it will be listed as 'Unmatched' rather than hallucinating an answer. This ensures every claim in your report is directly traceable to your materials. **Stage 3: Ethical Synthesis & Discussion** For qualitative analysis or educational contexts, use the **Case-Study Discussion Guide Prompt**. Paste a scenario, subject, audience level, and time limit. The AI will produce a facilitator-ready guide with neutral summaries, a 'Facts to confirm' list, and open-ended questions grouped by themes. *Critical Constraint*: The AI must not provide legal advice, medical diagnoses, or professional assessments. It distinguishes between source claims and its own inferences, marking gaps clearly. This stage helps structure complex discussions without overstepping ethical boundaries. **Prerequisites & Limitations** - Access to a standard LLM chat interface (no API or coding required). - You must provide all source text (abstracts, excerpts, scenarios). The AI does not fetch external content. - Runtime verification has not been performed; always cross-check extracted data against original documents. - Do not paste sensitive personal data or copyrighted texts you lack permission to share.