AI Prompting
AI/ML Prompting Guide for Research, Processing, and Analysis
return to ml
Introduction
This guide helps users employ AI/ML technologies effectively to enhance research collection, processing, and analysis. It provides structured prompts with placeholders for direct use in conversational AI applications like ChatGPT.
Prompting Strategies
1. Write Clear Instructions
Ensure your instructions are specific and direct to get the most relevant and accurate responses from AI.
Prompt Template:
"Analyze [detailed description of the topic or data] focusing on [specific aspects]. Summarize the findings in [desired length or format]."
2. Provide Reference Text
Giving reference texts helps the AI understand the context or background of your query better.
Prompt Template:
"Considering the information provided in [reference text or document], discuss [specific question or topic]. Include any necessary citations from [specific sources or databases]."
3. Split Complex Tasks
Complex tasks should be divided into simpler, more manageable sub-tasks.
Prompt Template:
"Step 1: Outline the main arguments found in [document or text]. Step 2: Evaluate the implications of these arguments on [specific field or issue]. Step 3: Predict future trends based on this evaluation."
4. Give the Model Time to “Think”
Prompt the AI to consider the response more deeply by asking it to explain its reasoning process.
Prompt Template:
"Before answering, consider the factors affecting [topic or issue]. What are the potential impacts of [specific factor]? Explain your reasoning."
5. Use External Tools
Leverage the AI’s ability to integrate with external tools for enhanced data retrieval or analysis.
Prompt Template:
"Use the data from [external database or API] to calculate [specific calculation]. Summarize the results and explain how they impact [related topic or issue]."
6. Test Changes Systematically
Regularly evaluate the effectiveness of your prompts and the accuracy of the AI’s responses.
Prompt Templates:
Create an advanced query for $SearchEnginePlatform using $Language incorporating the information below: Searching for: Region: CountryX Language: Local Language of $Region SearchEnginePlatform: Google Data Type should include: PDF Media Type: News Database Report WIKI Repository Date Range: a month