Creating a Custom GPT for Educators: A Step-by-Step Guide
And an example of the book recommendation GPT I created.
With the exciting news that ChatGPT has recently made all custom GPTs free, there has never been a better time to create your own AI tools tailored to your specific needs. This guide will walk you through creating your own custom GPT using a real example of one that I made for my students called "Book Buddy."
Step 1: Define Your Purpose and Audience
Before diving into the technical aspects, it's essential to clearly define the purpose of your custom GPT and your target audience. For instance, the "Book Buddy" GPT is designed to help my high school students find books that interest them and are part of our collection.
Step 2: Access the GPT Platform
To create a custom GPT, you'll need to access a platform that allows for customization. The good news is that you don't need a ChatGPT Pro account; these tools are now available for free. Just go to https://chatgpt.com/gpts.
Sign Up/Log In: If you don’t already have an account, sign up for one. If you do, log in using your credentials.
Step 3: Start a New Project
Once you are logged in, navigate to the section where you can start a new project.
Create a New GPT: Look for an option to create a new GPT or custom AI. Click on it to begin.
Step 4: Input Basic Information
You will need to input some basic information about your GPT.
Name Your GPT: Give it a meaningful name, such as "Book Buddy" for this example.
Description: Provide a brief description of what your GPT will do. For instance, “A book recommender using the books we have in our library. Make sure to ask Mrs. M for help finding the book.”
Step 5: Define Capabilities and Features
Specify what your GPT will be able to do. This is a crucial step where you define the interactions and functionalities of your custom GPT.
Book Recommendations: Include a feature where the GPT can suggest books based on student interests.
Library Management: Add capabilities for managing book inventories and tracking borrowed books.
Student Assistance: Enable the GPT to answer common questions students might have about library services.
In Book Buddy, I focused on Book Recommendations, so I asked it to do the following:
Book Buddy is a friendly yet informative book recommender for students, focusing primarily on the books listed in the UHS catalog document that the user has uploaded. This GPT uses the contents of the document as its primary dataset to provide a mix of fiction and non-fiction genres, offering detailed book information, including synopses, genres, themes, and book formats (e.g., eBook, audiobook, hard copy) with reasons tailored to the student's academic level and interests. When providing recommendations or information, Book Buddy will first consult this document to ensure the suggestions are relevant and are part of the UHS Library collection. If a query cannot be answered using the document, Book Buddy will then use its baseline knowledge or other sources to provide the best possible assistance, always noting the source of the information.
Step 6: Train Your GPT
Training involves feeding your GPT with specific data and instructions to ensure it performs as expected.
Data Input: Provide a dataset or a series of prompts and responses that the GPT can learn from. This might include book summaries, FAQs, and common library inquiries. In my case, I asked TLC to provide me with a custom report that included the titles, formats, genres, and descriptions of all of the books in our collection. Note that ChatGPT cannot read MARC records, so the format had to be either Excel or PDF.
Fine-Tuning: Adjust the responses based on initial testing to ensure accuracy and relevance.
Step 7: Test Your GPT
Before deploying, it’s important to test your GPT to see how it performs in real scenarios.
Simulate Interactions: Conduct several test interactions with your GPT to ensure it responds appropriately.
Adjust and Refine: Based on the testing results, make any necessary adjustments to improve accuracy and user experience.
Step 8: Deploy and Share
Once you are satisfied with your GPT's performance, it’s time to deploy it.
Deployment: Follow the platform’s instructions to deploy your GPT. This might involve integrating it with your existing systems or creating a standalone application.
Share with Your Audience: Provide access to your target audience. In this case, my target audience is just my high school students and staff. You can share a link, embed it in your library’s website, or include it in your digital communications.
Step 9: Monitor and Update
After deployment, continuously monitor the performance of your GPT and update it as needed.
Feedback Loop: Collect feedback from users to understand how the GPT is being used and where it can be improved.
Regular Updates: Periodically update the data and training to keep the GPT relevant and useful.
Example: "Book Buddy" GPT
To give you a concrete example, here’s a brief overview of how the "Book Buddy" GPT was created:
Purpose: Assist high school students with book recommendations.
Capabilities: Book suggestions.
Training Data: Included book summaries, library FAQs, and common student questions.
Deployment: Shared via a custom link to students and staff.
Things to Remember
Here are some things to think about while experimenting with creating your own GPTs:
They will not be perfect. At times, it may recommend books that are not in your collection or get some information wrong.
Availability Check: Since the GPT is looking at a fixed dataset, there is no way for it to tell if the book is currently available in our library. So you would need to make sure to tell the patrons to check with the librarian to see if the book is currently available.
User Training: Provide training or a brief guide for users on how to interact with the GPT effectively.
Continuous Improvement: Regularly update the GPT with new data and refine its responses based on user feedback.
By following these steps, you can create your own custom GPT tailored to the specific needs of your educational environment. This tool can be a valuable addition to your resources, allowing students to get more customized recommendations.
