1 of 21

USING GENERATIVE AI IN RESEARCH

AI & ACADEMIA CONFERENCE

PANEL: AI AS FACULTY ASSISTANT

JOE GRECO

FINANCE DEPARTMENT

CSUF

​

2 of 21

INTRODUCTION

Generative AI is transforming academic research

Enhances literature reviews, idea generation, and writing

Reduces time spent on repetitive research tasks

Supports faculty in publishing research and textbook creation

3 of 21

HOW LARGE LANGUAGE MODELS (LLMS) WORK

Training on Large Datasets: LLMs are trained on vast amounts of text from books, articles, and websites

Neural Network Processing: Uses deep learning to predict the next word

Fine-Tuning & Optimization: Adjusted for specific tasks (research, coding, etc.)

Generating Responses: Uses probability to generate coherent and context-aware text

4 of 21

AI BASICS

Choose a reliable LLM: ChatGPT, Claude, Perplexity, etc.

Choose

Design well-structured prompts

Design

Provide clear context, details, and desired output

Provide

Avoid vague language to get accurate responses

Avoid

5 of 21

CREATING RESEARCH ARTICLES

6 of 21

AI-ASSISTED RESEARCH PROCESS

Select and expand on

- Select and expand on a selected theme

Generate

- Generate 10 themes, issues, and concepts

Guide

- Guide GPT(generative pre-trained transformer) with clear directions

Create

- Create a project: Upload research files

7 of 21

DEEPENING RESEARCH WITH AI

- Prompt GPT to list 10 key references

- Summarize 5-7 core ideas from each reference

- Identify key common issues, content, and trends

- Select one key issue and create a structured outline for a research article

8 of 21

WRITING WITH AN AI ASSISTANT: THE HUMAN SIDE

- Develop an article outline with references

- Write the text (the human side)

- Use Grammarly and Quilbot for refinement

- Ensure clarity and academic tone

9 of 21

CREATING A TEXTBOOK WITH AI

10 of 21

TEXTBOOK CREATION WITH AI: CREATE A GPT

- Upload lectures, PPTs, and SLOs

- AI generates a structured table of contents

- Assign SLOs to chapters and sections

- Draft text content with AI assistance

11 of 21

12 of 21

13 of 21

BASIC REQUIREMENTS FOR AI-ASSISTED RESEARCH

Choose a reliable LLM: ChatGPT, Claude, etc.

Design well-structured prompts

    • Task – mandatory: generate, write, analyze, etc
    • Context - a textbook chapter with references

​

    • Format – what does the end result look like

14 of 21

A WELL-STRUCTURED PROMPT

Prompt:

Imagine you are an AI researcher writing a text book for the course entitled Global Risk Management and International Finance. Your task is to generate a chapter of 2,500-words on the topic of exchange rate risk that provides a comprehensive analysis while maintaining a clear scope and avoiding unnecessary digressions.

15 of 21

CREATING THE TEXT CONTENT

STEPS TO CREATING CONTENT

    • CREATE A TABLE OF CONTENTS THAT INCLUDES SECTIONS FOR EACH CHAPTER (human side)
    • GENERATE REFERENCES FOR EACH SECTION OF THE CHAPTER
    • GENERATE AN OUTLINE FOR THE TEXTUAL CONTENT FOR EACH SECTION OF EACH CHAPTER
    • WRITE THE TEXT(human side)

16 of 21

ENHANCING LEARNING WITH AI-GENERATED CONTENT

- Create question banks in Canvas for each section of a chapter

- Use Bloom’s Taxonomy for assessment

- Develop quizzes and exams by sections

- Automate pooling for test generation

17 of 21

TEXT BOOK QUESTION BANK PROMPT

  • Using the attached file create 10 multiple choice questions with answers that include 6 at the recall, comprehension, and application level of Bloom's taxonomy and 4 that are at the analysis level of Bloom's taxonomy for Part Two of Chapter 15

18 of 21

SUMMARY: AI AS A RESEARCH ASSISTANT

19 of 21

MAJOR LIMITATIONS OF LARGE LANGUAGE MODELS (LLMS)

Hallucination: May generate false or misleading information

Bias & Ethical Issues: Reflects biases in training data, potentially reinforcing stereotypes

Limited Understanding: Recognizes patterns and may lack true comprehension or reasoning

Dependence on Training Data: Cannot generate knowledge beyond what it was trained on

20 of 21

CONCLUSION

AI accelerates research and writing

Streamlines literature reviews and idea generation

Facilitates structured content creation

Enhances efficiency but requires human oversight

21 of 21