Skip to content

Prompt Templates for Content Items

Pre-production Process and Guidelines

GPT can produce a large amount of content in a relatively short amount of time. This makes it a good tool for brainstorming during the pre-production phase. The speed of GPTs output can be used to curate a breadth of course topics and outlines for most content items.

As with any output from GPT, verification is necessary. This can be done using GPT output alongside more traditional research content such as textbooks, other courses, and job skill reviews.

When generating topics for content brainstorming, it is important to be aware of bounded rationality. This is the bias toward a result seeming sufficient when it may only be “good enough”. In terms of GPT output, we may feel the output covers the correct topics, but some important topics might be omitted and are now hard to identify since the work feels complete.

For example, the GPT output for an introductory C course has the following variables topics: - Define and differentiate between the various data types in C. - Declare and initialize variables in C.

This might look good for variable learning objectives, but it omits variable manipulation.

One way to avoid bounded rationality is to first generate for breadth. This would output more than needed but allow for an exploration of topics and allow for a subtractive process that gets you closer to your goal.

Breadth First

It's OK to start with a simple prompt, but you will probably find after one or two output generations that certain adjectives will get you closer to the specific topics you want. Adjectives from the following groups can be included with your prompt: - Content Type: article, topic list, learning objective - Content difficulty: beginner, advanced, 3rd grade, high school - Content specificity: high-level, general - Content delivery: online vs. written vs. classroom - Output format: nested list, JSON object, single-level list

Example Prompt: Create a topic list for a high-level course in introductory video game development.

Note: When working with a breadth-first approach, increasing the max token size is usually necessary.

Reduce Manually

In some cases reducing the GPT output is as easy as copying what you do what, or erasing the things you don’t want. The following prompt output a list of 15 topics, where only the first 8 were pertinent.

Example Prompt: Create a beginner topic list of C programming language.

Reduce Automatically using Constraints

You can attempt to create more specific output by identifying topics to include or exclude. Simply use separate sentences on what to include and what to exclude.

Example Prompt: Create a topic list for a high-level course in introductory video game development. Include topics: art assets, sound assets, game engines. Exclude topics: programming, video game history.

Note: Since this process could require some trial and error, you should reduce the chance of variations in output by bringing the Temperature down toward 0.

Expand

Include your desired topic list or outline in a prompt to expand each prompt into learning objectives or written content.

In some cases, it may also be beneficial to expand each list item individually.

Example Prompt: Create a set of learning objectives based on the following list of C programming language topics.

  1. Introduction to C programming language
  2. Data types and variables in C
  3. Operators in C
  4. Control structures in C (if-else, switch-case, loops)
  5. Arrays in C
  6. Functions in C
  7. Pointers in C
  8. Structures in C

Conclusion

Prompt generation can be a trial and error process. The purpose of using content generation for preproduction tasks is to add another source of data that can be created quickly.

Learning Objectives

Generating a list of learning objectives/standards from a Course Topic and Module Topic:

A learner is taking an online course on [course topic]. After completing a module on [module topic], the learner should be able to

✅ This prompt is effective for generating a list of possible learning objectives

Example Prompt:

A learner is taking an online course on Python. After completing a module on Classes, the learner should be able to

Output:

  1. Understand the concept of Object-Oriented Programming (OOP) and how it is implemented in Python.
  2. Define and create classes in Python, using the class keyword.
  3. Explain the purpose of the __init__ method and create constructors for custom classes.
  4. Create and use instance variables and methods within a class.
  5. Understand the concept of inheritance and create subclasses that inherit properties and methods from parent classes. ...

Articles

Generating an article using a brief description + outline:

Write an article about [topic]. [Additive and subtractive logic, e.g. "include information on available frameworks", "do not mention Vue.js"] [Info about the audience]

[outline]

✅ This prompt is effective for generating a basic article that fulfills the outline provide ❗ If the outline is too complex (too many nested layers), GPT may duplicate the outline instead of writing an article from it 🔀 An alternative is to exclude the outline and include more information about your expectation in the first part of the prompt. This still works, but the flow is generally much more sensible when passing a human-reviewed outline

Example Prompt:

Write an article about Single-Page Web Applications (SPA) using the template below. The reader has some experience building basic web pages using HTML, CSS, and JavaScript.

Title: What is a SPA?

A. Introduction B. Multi-Page Applications C. Single-Page Applications D. SPA Frameworks E. SPA Pros and Cons i. Pros ii. Cons F. Conclusion

Syllabi

Generating a syllabus for a given topic:

Create a syllabus for an online course on the [topic]. It should be a bulleted list where each top-level bullet is one high-level concept. Each point should have 3-4 subpoints detailing the most important learning objectives for that section. [additive and subtractive prompts — e.g. “the course will cover using React and Express to build a web application”, “The course will not cover Vue.js”] [Info about the learner]

✅ This prompt is effective for generating a basic syllabus and some potential learning objectives ✅ This prompt is effective for generating syllabi on very technical topics ❓This prompt is less effective (but still effective!) when generating syllabi for more conceptual topics — generally needs more instruction to yield what you want

Example Prompt:

Create a syllabus for an online course on the C programming language. It should be a bulleted list where each top-level bullet is one high-level concept. Each point should have 3-4 subpoints detailing the most important learning objectives for that section. The learner has little to no prior experience with computer programming.

Output:

  1. Introduction to C Programming
  2. Overview of C programming language
  3. Importance and applications of C
  4. Setting up the development environment
  5. Writing, compiling, and running a simple C program
  6. Basic C Syntax and Structure
  7. C program structure and components
  8. Variables, data types, and constants
  9. Basic input/output functions (printf, scanf)
  10. Comments and code documentation
  11. Control Structures and Flow Control
  12. Conditional statements (if, if-else, switch)
  13. Looping structures (for, while, do-while)
  14. Break and continue statements
  15. Nested loops and conditions
  16. Functions in C
  17. Introduction to functions and their benefits
  18. Function declaration, definition, and calling
  19. Pass by value and pass by reference
  20. Scope and lifetime of variables …

Quiz/Assessments

Multiple Choice Questions

“Standard”

Write a multiple-choice question for the following learning standard: [LEARNING STANDARD GOES HERE] Provide a right answer and three wrong answers. Provide an explanation for why each wrong answer is wrong, each explanation should be on a separate line and be labeled with the incorrect answer choice. Have the first answer be the correct answer.

To add variety to this type of question consider adding to the prompt something about:

  • using a scenario
  • spot the error in a code snippet
  • guess the output of a code snippet

For example:

Write a multiple-choice question that uses a javascript code snippet and asks the user what the output should be for the following learning standard: "Scope is a programming concept referring to the visibility of variables or data in different contexts. Variables can have various scopes including global scope (the entire program), file or module scope, function scope, or code block scope." Provide a right answer and three wrong answers. Provide an explanation for why each wrong answer is wrong, each explanation should be on a separate line and be labeled with the incorrect answer choice. Have the first answer be the correct answer.