Guidelines for Responsible AI Use

Intro ’Metrics, Spring 2025

Overview

This policy outlines acceptable use of Large Language Models (LLMs) like ChatGPT/Claude/Gemini/Kimi/Doubao/Deepseek and many others in our class. Our goal is to ensure these tools enhance rather than hinder your learning experience.

Note

Remember: The goal of using LLMs is to develop your own understanding and analytical skills. When in doubt, ask yourself: “Am I using this tool to better understand the material, or am I using it to avoid learning?”

Core Principles

  • Focus on learning and understanding over completion.
  • Be honest about LLM use when relevant.
  • Use LLMs as learning aids, not substitutes for thinking.

1. Writing Assignments

Permitted Uses

  • Checking your reasoning approach (without getting solutions)

  • Understanding general concepts related to problems

  • Verifying mathematical steps you’ve already completed

  • Getting hints when stuck (after making genuine attempts)

✅ Acceptable Use Cases

  • Grammar and Spelling:

    "Please help me to check sentence structure"
"Please help me to polish the language with word clarity"

  • Brainstorming:

    "What are some potential topics related to \[economic concept\]?"
"What are some key points to consider when analyzing \[topic\]?"

Unaccepted Uses

  • Getting direct solutions to problems

  • Having LLMs solve equations or derivations

  • Using LLMs to check answers before submission

  • Generating problem-specific explanations

❌ Unacceptable Use Cases

  • Getting direct solutions
"Please complete this problem set for me"
"Write an analysis of these regression results"
"What is the answer to the question, please write down steps by steps"

2. Programming Work

Permitted Uses

  • Using LLMs for debugging

  • Getting help with syntax

  • Learning coding patterns

✅ Examples of Acceptable Use:

  • ✅ Debugging:

    # Acceptable: Asking about specific errors
"Why am I getting this error in my regression analysis?"
lm(y ~ x, data = df) # Error: object 'y' not found

  • ✅ Learning Patterns:

    # Acceptable: Understanding code structure
"How do I create a function to calculate summary statistics?"

Unaccepted Uses

  • Submitting LLM-generated code without understanding it

  • Using LLMs to complete entire programming assignments

  • Copying code directly from LLMs without modification or documentation

❌ Examples of Unaccepted Use:

  • Like

    "Write a complete program to analyze this dataset"

3. Analysis and Writing for a Research Project

Acceptable Use Cases

✅ “How do I handle missing values in this column?” ✅ “What’s the best way to rename these variables?”

✅ “My regression output shows NaN values. What might cause this?” ✅ “How do I fix this error in my ggplot code?”

✅ “Can you explain this statistical concept in simpler terms?” ✅ “What are some good resources for learning more about panel data?”

Unacceptable Use Cases

  • Asking Chatbots to “write a 500-word analysis of the reading”
  • Using LLMs to expand your rough draft into a full essay
  • Asking LLMs to generate examples or case studies

❌ “What conclusions should I draw from this data?” ❌ “Interpret these results for me”

❌ “Summarize this research paper” ❌ “What are the key points from this week’s readings?”

4.Transparency Requirements

Important

You should acknowledge LLM use by:

  1. Including a brief statement describing how you used LLMs
  2. Specifying which parts of your work received LLM assistance
  3. Explaining how you verified and modified any LLM-generated content

Examples for LLM assistance note:

  • “Used Deepseek R1 Model to help debug loop structure in lines 45-50”
  • “Used Chatgpt o1 model for syntax help with ggplot visualization”

A Real Example of an LLM Creation Statement

Important

These guidelines were developed with the assistance of Claude 3.5 Sonnet.

Specifically:

  • The content was developed through systematic consultation with an LLM;
  • Case studies and examples were iteratively refined using LLM assistance;
  • Document structure and Quarto formatting were optimized using LLM-generated code;
  • All content has undergone thorough review and received approval from Prof. Zhaopeng Qu to ensure academic integrity and pedagogical alignment.