Discover how to build and share customised GPTs to boost your creativity and productivity.
Using GPTs for learning design
GPTs are a feature of ChatGPT Plus, allowing you to build customised chatbots without needing to code. You can train GPTs with your content and instruct them to perform a task in a particular way. These customised assistants can help you speed up your learning design and come up with creative ideas tailored to you and your learners.
Here are just a few ideas of how GPTs can be used in learning design:
Learning needs analysis — Come up with questions and analyse information to identify the learning needs, gaps, environment, etc.
Learner persona — Construct personas, empathy map and roleplay with a ‘learner’.
SME assistant — Help an SME craft content, then the learning designer to analyse and edit it.
Outcome builder — Create outcomes from project and learner information.
Course outline — Combine project information and content to create a course outline.
Script/activity writer — Take content and rewrite it for a particular activity, media or format.
Assessment developer — Develop assessments to measure outcomes.
Evaluation and data analysis — Analyse data and evaluate instructional resources/methods.
Check out my blog post, GPTs - What and Why?, to learn more about why GPTs are an important development and how they can be used for learning.
Building and sharing GPTs
GPTs are quick and easy to build, especially after a little practice. Here’s how you can create and share a GPT for learning design.
Step one: Define the GPT's audience and purpose
Just like when designing learning, you should start by considering the purpose of the project and who you’re designing for:
Who will the user be? — Identify the typical user of your GPT (if it’s not you).
What problem will the GPT solve? — Consider why the user will use the GPT and how it will help solve a problem.
This will help you build a GPT that is appropriate and useful to you or your audience. You could also include this information in your GPT instructions.
UX (user experience) tip: Consider how well your user can prompt ChatGPT and their expertise in the GPT topic/area. Some users may need more support to get the best results with your GPT.
Step two: Create your GPT
Start by selecting ‘Explore GPTs’ from the left menu and then the ‘Create’ button at the top right of the screen to access the GPT builder screen (shown below).
You can create your GPT in two ways: 'Create' or 'Configure'.
The Create option is best for simple GPTs or if you’re not sure how to write the instructions yourself. As you chat with the GPT it will update the name, image, instructions, etc., which otherwise you’ll need to enter manually yourself in the ‘Configure’ section — as shown below.
The Configure option is most suitable if you have a well-defined process or have some prompting expertise.
Here, you can input the following information:
Profile image (‘+’ icon) — upload a photo or use DALL.E to create a profile picture for your GPT
Name
Description
Instructions
Conversation starters — shown as buttons when the user opens the GPT. Selecting these buttons enters a prompt, which starts a new conversation for the user.
Knowledge — you can upload files to train your GPT
Capabilities — decide whether or not your GPT will be able to browse the web, generate images and interpret code or data files (I generally leave these switched on)
Actions — link to an API or external app (here's a guide on how to do this).
When you chat with the AI in 'Create’, it will automatically fill in the information in ‘Configure’ window.
Below is an example for a Branching Scenario Writer GPT I've made.
I told it: ‘I need help writing branching scenarios for learning resources’ and it wrote the name, description, instructions and conversation starters (as shown below):
This is a great start if you’re not sure how to write the instructions or want to create a simple GPT in minutes.
I named my GPT 'Branching Scenario Writer' and kept the description it wrote, but decide to write my own instructions (covered in the next section).
Although the conversation starters it created looked quite helpful, I wanted to guide the user through a specific process to elicit relevant context to help the GPT. With this in mind, I used one Conversation starter, ‘SELECT THIS BUTTON TO START’, and instructed the GPT to respond in a certain way to this prompt.
UX tip: The conversation starter buttons only show the first few words of the prompt. So, make sure each one is clear to the user in a few words.
Step three: Train and instruct your GPT
The two most important components to program your GPT are Knowledge and Instructions.
Knowledge
If you want the AI to include certain information or create content in a particular format or style, it helps to provide examples.
For my GPTs, I upload relevant examples of articles, activities or resources I’ve created in the past. If you don’t want to upload your content in a shared or public GPT, you could upload it in a normal chat thread, ask ChatGPT to create similar examples and then use these examples in your GPT.
You can upload research papers, theories, frameworks, etc., to train your GPT, though it's good practice to cite and/or ask permission from the original author (learn more about Responsible Learning Design with AI in this post).
I wanted my Branching Scenario Writer GPT to follow a particular format and style, so I uploaded a few examples of branching scenarios I’d designed previously.
Instructions
Writing instructions is similar to prompting, as how you write your instructions affects the GPT’s output.
Make sure your GPT instructions are SIC:
Specific — state exactly what you want
Informative — provide helpful context and relevant information
Clear — are easy to follow (AI is highly intelligent but not a mind reader!)
The length and detail of your instructions will depend on how specific you want the GPT’s output to be.
The instructions for the Branching Sceanrio Writer include context, rules and a process to follow, as shown directly below.
Select the arrow (left) to reveal/hide the instructions I used.
CONTEXT
You are an expert at writing engaging, exciting and realistic branching scenarios for e-learning resources.
RULES
You always use the same structure and tone of voice as the uploaded knowledge files, unless told to do otherwise.
You will use simple language that is accessible to a range of learners, including those who aren’t native English speakers.
You will incorporate the topic or information that the user gives you.
INSTRUCTIONS / PROCESS
In response to the conversation starter or any other prompt the user starts the conversation with, you will give the following response:
Hey there! To get started, please provide me with the following information:
1. Resource topic and/or purpose
2. Relevant performance outcome
3. Learner information
4. Number of steps required in the branching scenario (if applicable)
5. Any information you need me to include or would like me to refer to
Once I have this, we can start writing the branching scenario. You can also ask me to write the end feedback and to create alternate branches at any time.
You will generate each section of the branching text one at a time. After generating a section, give the following response to the user:
**Please give me feedback to rewrite this text better or tell me ‘OK’ and I’ll continue on. Or, let me know if you’d like to work on an alternate branch of the scenario or end feedback**.
If given feedback or a request to edit, you will re-write the section accordingly. If you are told ‘OK’ or something similar you will generate the next section of the branching scenario.
If you are told you are finished or if asked to write end feedback, ask the user what they’d like to include or whether they’d like suggestions. When writing the end feedback you are summarising the key points covered in the branching scenario in simple and concise language. This feedback is to help the learner apply the learning in relation to any performance outcomes provided.
If asked to create an alternate branch in the scenario (or a similar request), this means you need to create a branch in the same scenario, but where different decisions are made with different consequences.
When doing an alternate branch, ask the user how/where they’d like to start this alternate branch.
When writing the alternate branch scenario text, do not use the word ‘alternate’ or reference the fact that this is an alternate branch.
If you are ever unsure of what to do or are missing information you must tell the user.
FORMAT
Structure each step of the branching scenario as:
##Step number
One or two paragraphs describing the scenario
**A short question**.
- Option 1
- Option 2
One or two short paragraphs of correct and incorrect feedback.
UX tip: Markdown is a simple way to specify how you want the GPT to format its output.
Step four: Test your GPT
Before sharing the GPT, or using it yourself for something important, it’s best to test it out. You can do this in the preview half of the screen while building it, which allows you to quickly make edits.
Below, is my test of the Branching Scenario Writer:
Step five: Share your GPT
Once your GPT is complete you can choose to share your GPT, make it public or keep it private.
If you want to share the GPT with another ChatGPT plus user, set it to 'Anyone with a link' and copy the link using the three-dot icon.
If you want to make it available to everyone, select a category and confirm. Anyone will be able to search for your GPT in the GPT store — the store will be explained in the next section.
You can personalise your GPT Builder profile and connect a domain (you’ll need to verify the domain).
Share in the GPT store
The GPT Store has finally been released and is available to ChatGPT Plus users. The store is a digital marketplace where you can discover and use custom GPTs created by others. It's like an app store, but for AI models.
Here's how it works:
Browse and discover — Users can explore a variety of GPTs, each tailored for specific tasks or industries.
Easy access — Once you find a GPT that suits your needs, you can easily add it to your ChatGPT Plus environment.
Community-driven — The store allows developers to share their creations with a wider audience.
OpenAI have shared the following guidance for those wanting to share in the GPT store:
Review updated usage policies and GPT brand guidelines to ensure that your GPT is compliant (most of these guidelines focus on preventing misuse and ensuring respectful, legal content).
Verify your Builder Profile (settings > builder profile > enable your name or a verified website)
Publish your GPT as ‘Public’ (GPT’s with ‘Anyone with a link’ selected will not be shown in the store)
Earn money from your GPTs
You may be able to make money from your GPTs. At the time of writing, OpenAI are yet to release exact details, though they have stated that you will ‘be able to earn money based on how many people are using your GPT’.
The barrier for entry is low, as GPTs can be created quickly and easily. With competition so high and the GPT store only showing the most popular GPTs, it may be difficult for people to find your GPT through the store. There are also reports of GPTs being copied and shared by others as their own.
If you do want to earn money from your GPTs, you’ll need to provide unique value to your user and promote it. One way of doing this is to give your GPT API access through the Actions setting in the GPT builder (see next section), linking it with external data and apps, such as Google Apps — here is guide to using Actions in GPTs.
What next?
If you have a ChatGPT Plus subscription, start building and testing your own GPTs. Search for others’ GPTs in the store or by using GPT Tracker. If you don't have a subscription, why not subscribe for a month and see how GPTs can make a difference to your learning design.
Want to learn more about using AI for learning design? Follow me on LinkedIn and subscribe to this blog by selecting the button in the footer below.
Want help building a customised GPT? Or to streamline your learning design processes? Contact me on LinkedIn.
ChatGPT was used to help write the sections: ‘What are GPTs’ and ‘How does the GPT store work?’.
Comments