Creating Educational Materials with AI: A Step-by-Step Guide

Contents:

• Key aspects of the GAIDE framework
• Preparation
• Creating a preliminary version of educational material
• Improving content at the global level
• Refining and improving content at the detailed level
• Integrating developments into a harmonious system
• Key aspects of using neural networks in creating educational material

The process of integrating generative artificial intelligence into the creation of educational materials and entire educational courses is gradually gaining popularity. However, new scientific research on the benefits and potential threats of using neural networks in instructional design presents a complex picture. In the same areas of work, AI can both significantly assist and negatively impact the final results.

How can artificial intelligence be effectively integrated into the process of developing educational programs, minimizing possible risks and maximizing benefits? Various frameworks, which represent structured action plans, are widely used in the field of instructional design. American researchers Ethan Dickey and Andres Bejarano have created a unique framework called GAIDE, which enables a thorough and informed approach to the creation of educational materials using any neural network technologies.

Key aspects of the GAIDE framework

GAIDE is an acronym for Generative AI for Instructional Development and Education. This is the name given by the scientists to their algorithm for creating educational materials, which includes the stages of preparation, content generation, and subsequent refinement.

The framework is based on an iterative method: the user receives content created by artificial intelligence, analyzes it, makes adjustments for the neural network, and continues this process until a satisfactory result is achieved. An accessible chatbot is proposed as an AI tool. Dickey and Bejarano tested GAIDE using ChatGPT 3.5 and 4.0, Bard, LLaMA, and Copilot in Microsoft Bing. Their results showed that ChatGPT 4.0 demonstrated the greatest consistency in producing high-quality content.

The framework is based on outcome-oriented design principles and constructivist theory, which posits that students do not simply receive knowledge pre-packaged but construct it through interaction with the learning environment. Therefore, GAIDE provides an adaptive structure that allows for the modification of learning materials depending on different contexts and the needs of different learner groups.

The framework's creators, who are university professors, developed it with the specifics of higher education in mind and have also used it in the development of teaching materials for information technology courses. However, according to them, GAIDE was conceived as a universal tool that can be used regardless of educational level and subject area.

In the next section, we will consider the stages of creating educational content based on the GAIDE framework.

Preparation

The first stage of preparation is carried out without the use of neural network technologies. This process is of great importance, since it ensures the efficiency of further generation of educational materials and prevents unnecessary waste of time.

The creators of GAIDE advise first of all to define the main goal, which will serve as a guideline for content creation and will help the instructional designer maintain focus on a given direction.

As Dickey and Bejarano note, the goal can be formulated in general terms, for example: "To create meaningful, interesting, and relevant educational material for students in the Computer Science and Technology course." However, it is necessary to link this broad goal with specific, measurable results. So, the wording "students will learn sorting in programming" is not a good example, while "students will be able to use basic sorting algorithms" sounds much better.

The prompt usually begins with a description of the learning situation and learner characteristics. The main rule is to provide artificial intelligence with the information that would be useful to a specialist creating educational materials for a new audience. The description of the context often includes the following elements:

• definition of the role that the neural network should take (e.g., "You are a computer science lecturer at a higher education institution");
• topic of the training;
• initial information available to the target audience;
• Demographic parameters of the target audience in the context of universities usually boil down to an indication of the course in which the students are studying.

Due to the fact that the results-based approach to designing the educational process was chosen as a framework, learning objectives, which represent the expected learning outcomes, play a key role in GAIDE. The capabilities of a neural network can be used to formulate them.

When formulating a query, experts recommend using specialized terms. This, firstly, allows you to accurately and clearly convey your intentions to the model, and secondly, helps the neural network begin to "reason" in the spirit of a professional educational software developer, which, in turn, improves its performance.

Considering the results achieved during the preparation stage, the prompt will look as follows:

1. Determine the main characteristics and properties of a binary search tree, including its structure and operating principles.

2. Analyze algorithms for inserting and deleting nodes in a binary search tree, comparing their efficiency and time costs.

3. Apply binary search tree traversal methods (pre-order, symmetric, and post-order) to retrieve data in different orders.

4. Compare a binary search tree with other data structures, such as arrays and linked lists, to identify their advantages and disadvantages depending on specific problems.

5. Write code to implement a binary search tree, including functions for inserting, deleting, and searching for elements, and test it on various datasets.

The initial results may not be of high quality, which is taken into account in the GAIDE framework. Experts recommend asking the model more options than necessary, and then highlighting both successful and unsuccessful answers in the next query. This will help the neural network improve the final results.

For example, if the neural network has suggested a number of educational goals, some of which are not entirely relevant to you, and you want to focus on the practical application of knowledge, you can use the following prompt: "Options 2, 3, and 7 are relevant to me, but goal 2 should be reformulated so that it corresponds to a higher level in the Bloom's pyramid. Additionally, change Goal 3 to emphasize practical application rather than understanding."

If you plan to make changes to the neural network's output yourself, Dickey and Bejarano recommend communicating in the same chat what exactly you changed and how, even if you don't currently require feedback from the AI. This will allow the model to better take into account your preferences in the future.

Read also:

Advantages and disadvantages of using ChatGPT when creating online courses.

Creation Preliminary version of educational material

Once the learning objectives have been refined, you can begin creating the learning materials for the course.

Within GAIDE, learning materials are divided into two main groups: lectures and student assignments. Assignments include not only questions, but also exercises, problems, case studies, and other activities. Separate algorithms have been developed for each of these groups, assisting authors in content creation and subsequent improvement.

The lecture draft is a preliminary plan that can be refined in the following stages. A sample request might be formulated as follows: "Develop an outline for a 50-minute lecture on the topic 'Binary Search Tree', considering the following learning objectives: ...".

The assignment draft is a preliminary list of options from which the teacher can select the most appropriate ones. As mentioned earlier, it is recommended to request a more extensive list from the neural network to ensure a variety of choices. Example query: "Provide 20 assignment options on the topic 'Binary Search Tree,' focusing on the following educational goals: ...."

Improving Content Globally

After generating draft materials using a neural network, it is necessary to carefully analyze the results and begin general revision. At this stage, it is important to consider the content in its entirety, paying attention to the main sections. The key is to assess the draft's alignment with the overall course goal, as well as the specific learning objectives, student level, and context. If certain elements are generally moving in the right direction, but the quality of their development leaves much to be desired, then this stage should not be focused on this - the next stage will be devoted to detailing and revision.

When making changes to the lecture plan, it is worth considering a number of important points:

• What is the preferred length of the lecture?
• Does the process of preparing the lecture involve additional responsibilities for the instructor, such as creating assessment and support materials?
• Is there an opportunity to participate in learning activities before or after the lecture?
• Which subtopics should be considered most significant for discussion?
• Is it necessary for the lecture to contain certain elements, such as testing or group work?

If any of the elements from this list are important for the development of the lesson, appropriate recommendations should be added to the prompt for its revision.

The authors argue that there are two approaches to Refining the plan: You can make changes one by one, adding one new sentence at a time, or you can provide the neural networks with a full list of necessary adjustments. For example, you could say, "Change the plan so that the lecture takes no more than 45 minutes and include a group discussion." As a result, a reliable basis should be formed for more thorough elaboration and clarification of details.

The purpose of high-level refinement of tasks is to compile a list of various and practical solutions. When developing feedback for a neural network, the following aspects should be considered:

• Assignment types may include such forms as solving equations or problems, analyzing specific cases, open-ended questions, multiple-choice questions, and other options.
• The degree of proficiency in knowledge and skills is assessed using Bloom's Taxonomy or another selected model.
• The learning focus implies attention to specific skills within the overall topic of the lesson.
• The topic implies an original presentation of the assignments.

Instructions for the neural network depend on the options received at the draft stage and what exactly you are not satisfied with them. For example, if only some of the assignments meet your educational goals, it makes sense to mark them in the request as suitable and ask the neural network to suggest similar options. If a neural network begins to repeat itself or deviate from established requirements, experts recommend creating your own example and using it as a template.

The authors note that when creating a set of tasks, the boundary between the macro- and micro-levels of refinement becomes quite blurred. As soon as you begin to focus not on finalizing the entire set as a whole, but on improving individual selected options, this will indicate a transition to a new stage of work.

Read also:

Creating online courses using Neural Networks: Practical Examples

Refining and Improving Content at a Detailed Level

Once you've ensured that the structure of the material meets your requirements and aligns with your learning goals, you can begin working in depth on individual elements—such as subsections, subtopics, and specific questions and tasks. At this stage, prompts can be very detailed, including pointing to specific words or phrases and asking the neural network to rephrase specific parts of the text.

However, as the neural network generates one version after another, difficulties can arise, such as loss of focus and contextual inconsistency.

Dickey and Bejarano describe a phenomenon they call loss of focus, where the neural network begins to perform poorly on its instructions, as if losing track of both previous requests and its own responses. In this situation, it's recommended to start a new session with the neural network by opening a new chat window. It's important to re-introduce the context and target audience, as well as list the learning objectives. Be sure to indicate what stage of development you're at; for example, this could be a previously created lecture outline that will help you move on to detailed development of individual sections. Context confusion occurs when the neural network uses the same parameters to generate different elements. For example, if you specify that the first task should focus on comprehension and memorization of information, there's a risk that the neural network will apply this requirement to tasks 2 and 3, where the emphasis may shift, for example, to practical skill practice. In this regard, the framework's authors recommend formulating requests in more detail, specifying which section you are editing and the specific learning objective for the text or assignment.

Depending on the type of content, the GAIDE authors also distinguish between various methods for improving materials at the micro level.

For each element of the lesson plan, the neural network is capable of generating both key ideas and expanded texts on various subtopics. It is important to analyze the extent to which the resulting text meets the stated learning objectives and evaluate its ability to engage students. It is equally important to check the material for factual errors. By gradually leaving comments in the prompts with each new iteration, you will ultimately be able to generate high-quality texts for the upcoming lecture.

After editing certain parts of the text, you can ask the neural network to combine them into a single scenario. In addition, upon appropriate requests, artificial intelligence will be able to help create a presentation to support the lecture, offer various diagrams and images, as well as active task formats for students, which will significantly improve their educational process.

Read also:

Eight ways to Teachers and methodologists can implement the Perplexity neural network in their work.

When working on questions and assignments, you can change their difficulty level and introduce narrative elements. Everything depends on your idea: the neural network can either "wrap" each assignment into a short story or create a full-fledged narrative, for example, with a character who explores a topic together with students.

To improve the wording, it is effective to engage the neural network in the role of a student, allowing it to complete a specific assignment. The authors note that this approach helps identify shortcomings in instructions that may confuse real students. This method is especially useful when refining problem situations and case studies, as well as open-ended questions—these are the formats with which students most often work independently.

In addition, the suggestion generated by the neural network can serve as the basis for creating an assessment rubric. It can serve as an example of correct and complete execution of the task, provided that the AI ​​has completed the task at the proper level, and also help in formulating criteria for evaluation.

Integrating Developments into a Harmonious System

The GAIDE framework uses a segmentation-based approach, in which different elements of educational materials are created and refined independently from each other, often over several sessions. While it's possible to ask the neural network to combine these developments into a single lesson scenario, one shouldn't expect a high-quality result. Therefore, as Dickey and Bejarano note, the optimal option would be to create a separate document where successful generation results can be transferred. This would then allow for manual assembly of all the elements, careful organization of the material, and final refinement.

The authors note that generative AI can not only develop individual educational materials but also assist in organizing entire courses. Once you have defined the learning objectives and outlined the basic structure of the course, including assessment methods, the neural network can suggest a detailed lesson plan that will indicate the distribution of time across different topics.

Read also:

Four key principles for creating effective prompts that will help teachers in developing educational materials.

Key aspects of using neural networks in creating educational materials

The authors provided additional comments regarding the use of generative social networks. While these observations are not directly related to the GAIDE framework, they may be quite useful for those interacting with neural networks in their work:

• Artificial intelligence is much like a diligent elementary school student. Dickey and Bejarano argue that this analogy is very useful for understanding the optimal ways to interact with AI. Like children aged six to eight, neural networks successfully complete tasks when given clear instructions. However, like young children, AI lacks the independence and initiative to find solutions without clear guidance. As a result, unless a human monitors the process closely, the neural network may begin to deviate from predetermined goals.
• One of the significant advantages of artificial intelligence is its ability to generate ideas. According to the framework's developers, rather than setting strict parameters for the creation of specific content from the outset, it is wiser to use a neural network for brainstorming. The suggestions that AI can generate are often diverse and creative, which in turn contributes to the improvement of the final learning material.
• Artificial intelligence can provide feedback from a variety of perspectives. Dickey and Bejarano recommend paying attention to prompts that assign specific roles to the neural network, such as "Imagine yourself as a first-year student" or "You are a first-year student." Using such approaches, we can better understand how different students with different levels of knowledge and experience will interact with the materials, allowing us to tailor the educational process to the needs of different groups of learners.
• It is important to recognize the limitations of neural networks. When it comes to creating learning materials, there may be a desire to achieve perfection from AI—repeatedly requesting changes and attempting to improve the generated content with each new version. However, such attempts do not always lead to the expected results. A more effective approach would be to recognize that there are areas where AI isn't at its best and to use the opportunities it can offer wisely.

Read also:

• Five skills that are becoming relevant for instructional designers in the face of the growing influence of neural networks.
• The IDEAS framework for online learning: what is it and how does it work?
• Five common mistakes when integrating artificial intelligence tools into training and staff development processes.
• List of neural networks for creating images: ways to improve educational materials.
• List of online platforms where you can develop infographics for educational presentations for free.