Research Prompting

Introduction

Research prompts demand sources, balanced views, and explicit uncertainty. Pair with RAG or search tools for grounding.

Beginner analogy: Like briefing a research analyst — primary sources, dates, counter-evidence.

Understanding the topic

Core concepts to understand:

• Require sources with URLs and publication dates.
• Ask for confidence levels per claim.
• Demand counterarguments alongside main thesis.
• Pair with RAG / search tools — bare LLM is too hallucination-prone.
• Use 'I don't know' as a valid answer.

Syntax reference

Visual workflow / architecture:

Prompt:
"For each claim, cite a source with URL + date.
 State confidence: low / medium / high.
 List 2 counterarguments.
 If unknown, say 'I don't know'."

Real-world use

Perplexity, You.com, Glean — all research-style prompts wrapped around search/RAG.

Best practices

• Ground with RAG or live search.
• Reject answers without sources.
• Surface confidence to user.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. Why ground research prompts in RAG?
• Q2. How prevent hallucinated sources?
• Q3. Confidence calibration — how?
• Q4. Counterargument prompting — why?