The Pros and Cons of AI in Curriculum Development: Research by

Contents:

• Potential threats of introducing AI into the development of educational courses
• The contribution of artificial intelligence to the process of developing educational programs
• Key aspects for integrating artificial intelligence into educational design

Over the past two years, many scientific studies have been conducted regarding the impact of generative artificial intelligence on the educational process. Philippa Hardman, formerly a research fellow at the University of Cambridge and developer of the DOMS™️ learning design system, summarizes the key findings of these studies, highlighting both the risks and benefits associated with integrating AI into instructional design.

In discussing the research findings selected by Philippa, we will not delve into the commonly accepted shortcomings of neural networks, such as "hallucinations," the risks of data leakage when using open platforms, or complex ethical and regulatory issues. Instead, our focus will be on the less obvious advantages and disadvantages that directly affect the development of educational programs and the creation of learning materials.

Potential Threats of Implementing AI in Course Development

Recent research has identified vulnerabilities in artificial intelligence that are important to consider when developing educational programs based on neural networks.

Using AI-powered tools can significantly improve the process of designing educational programs and facilitate their scalability. However, researchers emphasize that AI solutions based on pre-developed templates significantly limit the freedom of instructional designers to make independent methodological decisions.

Research conducted by scientists from Norway and the UK has shown that the use of rigid frameworks in instructional design tools limits the creativity of methodologists. Such standardized approaches create a feeling of being "squeezed" within a framework, which complicates the ability to adapt learning materials to the needs of the target audience. At the same time, there are more flexible solutions on the market, such as iLUKS and ChatCLD. These tools offer a certain structure that the instructional designer can use, but they also provide the opportunity to modify, refine, and further develop various elements of the program.

Read also:

Competition: who can create more A high-quality course—an instructional designer using a neural network, or someone working without its help?

In the course of their research, scientists discovered that artificial intelligence often creates content without taking into account methodological norms and is unable to adapt materials to specific conditions. For example, a study conducted in China showed that mathematics lesson plans developed based on the GPT-4 model required significant revision in 78% of cases before they met established educational standards and student proficiency levels.

Research conducted by Belgian scientists has shown that artificial intelligence tools reproduce elements of instructional design to a greater extent than they actually use its principles for careful analysis and content development. At the same time, as the authors of the study note, many learning design specialists mindlessly accept AI advice without critical reflection, which negatively impacts the quality of the final product.

Conclusion: The use of neural networks allows for the creation of initial versions of materials that require careful review for compliance with instructional design principles and require significant revision. Therefore, the anticipated time savings may not be as apparent, given the additional effort required for review and revision.

Philippa Hardman notes that there is a diversity of opinions regarding the creativity of instructional designers. On the one hand, AI-based tools can assist in generating new ideas. On the other hand, there is a tendency toward automation and standardization of all processes where possible. However, relying too much on neural networks when making creative decisions can negatively impact the final result.

Research conducted in the United States has shown that instructional designers who actively use artificial intelligence in their work are significantly less likely to develop unique assignments for students compared to their colleagues who are less interested in neural networks. Similar conclusions were reached in a study from Belgium, which also noted that a lack of critical thinking about AI proposals leads to predictability and lack of originality in educational materials.

Conclusion: Using neural networks to create content is appropriate if you are satisfied with the standard, generic format of educational assignments and materials. If you strive for uniqueness, however, it is worth using neural networks consciously – using them as an auxiliary tool during group brainstorming sessions, and then refining the ideas yourself. We will discuss this aspect in detail later.

Conclusion: Using neural networks allows for the creation of "rough drafts," but such materials require careful evaluation for compliance with pedagogical principles and require significant revision. Consequently, the estimated time savings may be questionable, or at least not significant, given the time it will take to review and correct them.

The Contribution of Artificial Intelligence to the Instructional Design Process

Recent research has identified key areas where the use of generative artificial intelligence can make a real difference in instructional design. You might be surprised, but these are largely the same areas that are fraught with risk. Either the research findings are conflicting, or the issue lies in how exactly these technologies are applied.

It appears that artificial intelligence does indeed significantly speed up the process of creating educational materials. Researchers in South Korea reported that ChatGPT reduced the time spent on lesson preparation by 65%. In turn, their colleagues from China presented the TreeQuestion platform, which uses AI to generate multiple-choice tests, which allows to reduce the time for developing test items by 95%.

Nevertheless, we acknowledge the aforementioned risks and understand that the reduced timeframes may negatively impact the quality of educational content if we do not review and edit materials created using artificial intelligence. And if we do review and edit, the question arises as to how much the material preparation process actually speeds up, given the additional time spent.

Furthermore, artificial intelligence facilitates the scale of knowledge and skill assessment. For example, automated test development not only saves teachers time but also allows for more regular assessments. Research conducted in China has confirmed that such automation does not lead to a decrease in the accuracy of results. According to the experience of scientists, the use of Bloom's Taxonomy in prompts is recommended for creating high-quality test items. However, they emphasize that teachers need to ensure in advance that the questions generated by the neural network are appropriate for the students' level and truly measure achievement in specific academic areas.

Personalized learning is an approach that adapts to the student's level of knowledge and experience, and takes into account their interests and preferences regarding the pace and methods of learning. Contextualization of learning, in turn, is a current trend in training and development, which also involves customizing the curriculum, not to the personal characteristics of individual students, but rather to the specific environment in which they work or will work, as well as their professional context.

Numerous education experts emphasize that artificial intelligence opens new horizons for adaptive learning that can be tailored to the individual needs of students. Recent research only confirms this view. For example, chatbots configured to various collective personas created based on the characteristics of student groups (a process called personas) demonstrated a 29% increase in learning outcomes in a course on working with information. Scientists attribute this effect to the fact that adapted explanations, prompts, and tasks had a more effective impact on students’ assimilation of the material than standard educational materials.

Read also:

"At the moment, it is impossible to implement "Personalization using neural networks is impossible... However, in a year it will be a reality!"

Wait, what about the claim that artificial intelligence is incapable of adapting materials to specific conditions? In fact, everything depends on the system configuration and the level of human involvement in the process. For example, to create student-centered learning programs, a group of researchers from the United States developed a framework called ARCHED. Within this framework, one AI tool is responsible for generating educational content, while another evaluates it for compliance with instructional design principles and learning objectives based on Bloom's taxonomy. A human controls the entire process, setting the necessary parameters to adapt the program to different groups of students and making important methodological decisions.

Philippa Hardman argues that a number of studies demonstrate how AI chatbots and adaptive learning platforms can adapt to student needs in real time. They are capable of analyzing their behavior and current achievements, and then providing feedback, as well as suggesting relevant learning materials and tasks appropriate to the level of difficulty.

While one of the risks mentioned is that the use of neural networks can sometimes hinder creative thinking, Philip Hardman notes that everything depends on how the specialist applies these AI tools. When used correctly, such technologies can become an excellent assistant in creative work, capable of generating ideas that a person might not come up with on their own.

Research conducted in the United States shows that the use of artificial intelligence in brainstorming contributes to the generation of ideas that are 47% more diverse than discussions conducted solely within a human group. Another study found that ChatGPT can offer educators a variety of options for delivering information and engaging in learning activities, thereby enriching the educational process.

It's important to be mindful of the neural network's suggestions, as this can negatively impact the depth and variety of students' learning experiences. Artificial intelligence tools are especially useful in the early stages of development, when instructional designers need ideas for new courses, various assignment options, and drafts of learning materials. Research shows that using such technologies frees up time and mental resources that can be devoted to analyzing submitted ideas, selecting the most appropriate ones, refining them, and adapting them to specific learning conditions.

Key Aspects for Integrating Artificial Intelligence into Instructional Design

Philippa Hardman argues that current research shows that hopes that artificial intelligence will radically transform approaches to instructional design are likely exaggerated. However, the benefits that neural networks can bring to workflows should not be underestimated. Artificial intelligence is successfully coping with certain tasks, demonstrating significant results and highlighting those areas where human participation remains indispensable.

To maximize the benefits of artificial intelligence in the work of instructional designers and minimize potential threats that could negatively impact the quality of the educational process, Hardman offers the following recommendations:

• Improving digital literacy and mastering prompt engineering skills are becoming important tasks, since well-formulated prompts can increase the quality of generation by 58% compared to standard options.
• The use of AI tools that significantly limit the freedom of action of the instructional designer should be avoided. Instead, it is wiser to choose more adaptive solutions that provide the ability to refine and customize the results obtained from the neural network.
• Entrust artificial intelligence with tasks such as creating drafts and alternative versions, analyzing large data sets, scaling feedback and adapting materials for different groups of students, as well as other routine processes that can be automated. Free up time to make creative, methodological, and strategic decisions, where neural networks cannot compare to human thinking.
• It is necessary to carefully monitor the quality of the created materials and not use them in their original form. It is important to ensure not only that there are no "hallucinations," but also that the content meets the methodological requirements, educational objectives, characteristics of the target audience, and the context of the educational process.
• It is important to focus on the diversity of educational materials and try to minimize excessive standardization.

Read also:

• Five skills that educational designers should develop in the context of the influence of neural networks.
• Five common mistakes when integrating artificial intelligence tools into training and staff development.
• 8 ways to use the Perplexity neural network for teachers and methodologists.
• Advantages and disadvantages of ChatGPT when creating an online course.
• Creating online courses using neural networks: practical examples.