Generative AI Empowering Innovation in Educational Models

Generative AI Empowering Innovation in Educational Models

by
The report from the 20th National Congress of the Communist Party clearly states, “Promote the digitalization of education, build a lifelong learning society and a learning-oriented country for all,” which strengthens the strategic guideline of promoting educational modernization through digitalization. Accelerating the digitalization level of basic education schools is crucially dependent on cultivating students’ good digital literacy, with computational thinking, which constructs abstract models as a problem-solving approach, being considered an essential component of digital literacy. Currently, generative artificial intelligence (AIGC) is rapidly gaining popularity worldwide, and its features of structured and hierarchical text generation align well with the concept of computational thinking. Therefore, there is an urgent need to integrate generative artificial intelligence into basic education to construct a scientifically sound new teaching model, thereby enhancing students’ computational thinking abilities.
Related Research
As the cornerstone of national education, the teaching forms of basic education should reflect continuous stability, often maintaining a cautious attitude towards the application of new technologies. However, this does not mean being rigid; once new technologies have been scientifically validated and proven to effectively improve teaching quality, they can be comprehensively integrated into basic education. As the latest technology, generative artificial intelligence has already been implemented to a certain extent in some basic education schools, achieving valuable results. Li Yingzhe explored the innovative functions of ChatGPT 4.0 in middle school mathematics lesson preparation, concluding that it can help teachers clarify teaching objectives, assist in improving activity design, showcase unique teaching styles, and provide personalized learning support. Luo Heng employed a quasi-experimental research method to examine the effectiveness of generative teacher comments, indicating that personalized teacher comments obtained through artificial intelligence can effectively assess and provide feedback in middle school classrooms, enhancing learners’ motivation and self-regulated learning abilities. Computational thinking, proposed by Zhou Yizhen from Carnegie Mellon University, is defined as applying fundamental concepts of computer science to solve problems, design systems, and understand human behavior, with its essence being abstraction and automation. Following Zhou Yizhen, Chen Guoliang pointed out that computational thinking should advance to the 2.0 era, where each scientific field has its unique characteristics of computational thinking. Ren Youqun proposed strategies for effectively integrating computational thinking into primary and secondary school information technology courses, suggesting a combination of method acquisition, tool application, and thinking transfer to implement computational thinking education.
Generative artificial intelligence, represented by ChatGPT, can answer various questions posed by people, and its characteristics of precision, inspiration, structure, and logic in answering questions align with the features of computational thinking. Therefore, if students possess computational thinking, they can better utilize generative artificial intelligence for learning. However, existing literature severely lacks research on the correlation between generative artificial intelligence and computational thinking. Thus, how to use generative artificial intelligence to enhance students’ computational thinking abilities should be an important goal of current educational digitalization research.
Evolution Path of Generative AI Technology
Generative artificial intelligence falls under the category of large-scale language models and belongs to the field of artificial neural networks. Generative artificial intelligence did not appear out of nowhere but has undergone several iterations of artificial neural networks. Based on its technical characteristics, its developmental history primarily follows the evolution path of Multi-Layer Perceptrons (MLP) → Convolutional Neural Networks (CNN) → Recurrent Neural Networks (RNN) → Generative Artificial Intelligence (AIGC).
Multi-Layer Perceptrons (MLP) are practical foundational artificial neural networks, consisting of an input layer, several hidden layers, and an output layer, with each layer containing multiple neurons, where each neuron is fully connected in a forward manner, updating the connection weights of each neuron through the BP algorithm iteratively. However, MLP consumes enormous energy due to the full connectivity of all neurons. Scholars have proposed improvement strategies for the shortcomings of MLP, namely local connections and feature extraction, which have been practically applied in Convolutional Neural Networks (CNN).
CNN extracts local features by applying convolution operations on input data, mainly consisting of convolutional layers, pooling layers, and fully connected layers. The convolutional layer is the core component of CNN, generating feature maps; the pooling layer reduces the spatial dimensions of feature maps; the fully connected layer outputs final classification or regression results. Although CNN also employs full connectivity in the final part, the advancement in reducing network scale compared to MLP, where all neurons are fully connected, is undoubtedly significant. Although Convolutional Neural Networks have somewhat overcome the enormous energy consumption of Multi-Layer Perceptrons, this architecture primarily corresponds to recognizing data with spatial structures, and it performs poorly in processing natural language, which is time-series data. The emergence of Recurrent Neural Networks (RNN) provided a feasible solution for natural language processing.
The most significant feature of RNN is that the current output depends not only on the current input but also on all previous inputs. RNN has an initial state, which can be viewed as RNN’s “memory,” used to store information from previous times. At each time point, RNN uses the current input information and the state from the previous time to produce the current output information, allowing RNN to learn patterns and rules in time-series data, thereby improving its prediction and classification accuracy. However, in practical applications, unfortunately, only very nearby historical information can be retained, despite the theoretical output of RNN being related to all historical information. In real human contexts, the meaning of a word is related not only to the preceding word in sequence but also to many preceding and succeeding words. To overcome this limitation, generative artificial intelligence (AIGC) using the Transformer algorithm was born.
In the Transformer algorithm of generative artificial intelligence, each word is computed with all words in the sentence to determine the relevance of that word to each word, thereby identifying a more accurate meaning of that word in the sentence. This algorithm consists of three steps: the first step is encoding, the second step is positioning, and the third step is the self-attention mechanism. The Transformer algorithm breaks through the barriers of time-series sequences, focusing instead on the value weight of each word in relation to every other word in the sentence, making the generated sentences more aligned with the characteristics of human language, thus enabling generative artificial intelligence to be rapidly recognized and accepted by the general public.
New Teaching Model Supported by Generative AI – An Example of Cultivating Computational Thinking
1. Model Construction
The International Society for Technology in Education and the Computer Science Teachers Association provide an operational definition of “computational thinking”: defining problems, logically organizing and analyzing data, abstractly representing data; generating automated solutions through algorithmic thinking; identifying, analyzing, and implementing various possible solutions to find the most effective solution; summarizing the problem-solving process and transferring it to other related problems. The above operational definition is a general model, and to implement it in real scenarios, specific improvements should be made. Therefore, based on the operational definition of computational thinking, researchers have developed a new teaching model for basic education supported by generative artificial intelligence, focusing on problem-solving and reasonably utilizing generative artificial intelligence to achieve the teaching goal of enhancing students’ computational thinking abilities.
The first step of the new teaching model is to propose actual problems discovered in teaching activities, which should be contextual or ill-structured problems, reflecting deep learning in Bloom’s taxonomy of educational objectives, corresponding to the definition of “defining problems” in computational thinking. The second step is to analyze the possible causes of the problem in depth, requiring the assistance of AIGC for problem analysis. Thanks to the precision and logic of AIGC in problem analysis, students only need to adjust the identified possible causes slightly, such as optimizing wording or changing the order of statements, and since the documents generated by AIGC are often structured, this step corresponds to the computational thinking operational definition of “logically organizing and analyzing data.” The third step is to visualize the possible causes of the problem. The previous step has combined AIGC’s auxiliary analysis to identify the possible causes, but typically this is expressed in pure text, which may be complex. To distill key factors, visualization techniques can be used to represent them graphically, such as mind maps or tree diagrams. The advantage of visual representation is that it can present the key factors causing the problem in a structured model, removing unnecessary descriptive text and retaining only the core key points, providing a good data model for proposing feasible solutions. This step corresponds to the computational thinking operational definition of “abstractly representing data.” The fourth step is to propose a certain number of feasible solutions, again requiring AIGC technology to process the previously analyzed causes of the problem and generate what it considers feasible solutions. Students then deeply reflect on the generated solutions, forming a series of problem-solving strategies that are well-expressed and logically clear, corresponding to the computational thinking operational definition of “generating automated solutions through algorithmic thinking.” The fifth step is that the proposed solutions should be theoretically feasible and should be validated through practice to determine which solution is authentic and reasonable, corresponding to the computational thinking operational definition of “identifying, analyzing, and implementing various possible solutions to find the most effective solution.” The sixth step requires students to summarize and generalize all the steps above to form a relatively stable problem-solving strategy, preparing for transfer to similar problems, corresponding to the computational thinking operational definition of “summarizing the problem-solving process.” The seventh step is to apply the stable problem-solving strategy to attempt to solve more similar problems to test whether the strategy is generalizable, corresponding to the computational thinking operational definition of “transferring to other related problems.”
The fundamental goal of this new teaching model is to cultivate students’ computational thinking abilities so that they can efficiently solve real problems encountered in teaching contexts. The involvement of AIGC as an auxiliary technology is its greatest feature, including assisting in analyzing possible causes of problems and assisting in proposing feasible problem-solving solutions, promoting the enhancement of students’ computational thinking abilities through a “human-machine collaboration” approach.
2. Teaching Implementation
To apply the new teaching model supported by generative artificial intelligence in actual teaching activities, the application of this teaching model needs to be divided into several stages, gradually cultivating students’ computational thinking abilities in a step-by-step manner.
(1) Non-AIGC Stage
The teaching objective of this stage is to familiarize students with the general computational thinking operational process. The main process involves teachers proposing contextually relevant real problems, students analyzing the problems, constructing models, proposing several feasible solutions, and submitting what they believe to be the optimal solution to the teacher. Teachers evaluate the quality of the solutions based on a rubric to assess students’ computational thinking abilities.
(2) AIGC Stage
After the previous stage of teaching practice, students have a certain understanding of the general computational thinking operational process. To promote “human-machine collaboration,” AIGC is introduced in this stage. The main process involves teachers proposing contextually relevant real problems, and students, with the assistance of AIGC, analyze the causes of the problems, construct models, propose solutions, and submit the optimal solution to the teacher. Teachers evaluate the quality of the solutions based on a rubric to assess students’ computational thinking abilities.
(3) Autonomous Learning Stage
In the previous two stages, the problems to be solved were proposed by teachers. In future practical life and work, students will face problems independently, so in this stage, students should discover real problems existing in their daily lives. The main process involves students independently proposing problems, while the remaining parts are the same as in the previous stage.
(4) Cooperative Inquiry Stage
“Autonomy, cooperation, and inquiry” are the current advocated directions for teaching reform, aiming to cultivate students’ awareness and ability for cooperative inquiry in the AIGC technology environment. In this stage, students with different characteristics are grouped based on the principle of heterogeneous grouping, and then they discuss and determine a relatively complex real problem, using AIGC to analyze the causes of the problem and propose a certain number of feasible solutions. Each group member practices the feasible solutions, assesses the practical effects of each solution, and through group discussions, they identify the optimal solution, collectively summarizing the problem-solving strategies for future learning practices.
Conclusion
Generative artificial intelligence typically possesses characteristics of structure and hierarchy in problem-solving, aligning well with the features of computational thinking. Therefore, with the aim of cultivating students’ computational thinking abilities, this article constructs a new teaching model for basic education supported by generative artificial intelligence, including model construction and teaching stages. Looking ahead, generative artificial intelligence will become an important component of technology application in the field of education, promoting the personalized and intelligent development of basic education.
Author Institution │ Sichuan University of Arts and Science, Jiangsu University Smart Education Research Center
Source │ “Information Technology Education in Primary and Secondary Schools” 2024 Issue 9

Generative AI Empowering Innovation in Educational Models

Generative AI Empowering Innovation in Educational Models

Click the above image to enter the online subscription page of the magazine
Important Notice

Recently, some teachers received the following manuscript notification and certificates, both of which were fake documents and seals.

Special Reminder! Please do not believe it! Do not transfer payment!

↓↓↓

Generative AI Empowering Innovation in Educational Models

Generative AI Empowering Innovation in Educational Models

Generative AI Empowering Innovation in Educational Models

For submissions, please recognize the official submission email ([email protected]) and the WeChat mini-program “Information Technology Education Magazine“.

Typesetting | Yi Dan Preliminary Review | Xiao An
Re-examination | Lu Qiuhong Final Review | Su Jin Zhu
Generative AI Empowering Innovation in Educational Models
If you like this article, don’t forget to click “See More” for support~Generative AI Empowering Innovation in Educational Models
What You Want to SeeWonderful
Generative AI Empowering Innovation in Educational Models

New Issue Delivery│2023 Issue 9Magazine Directory

Benefits Are Here! Welcome to Subscribe to “Information Technology Education” 2024 Magazine

Call for Papers! Welcome to Participate in the 2024 “Digital Empowerment of Primary and Secondary School Teaching Practice” Call for Papers Activity

The 2024 Primary and Secondary School Students’ Information Technology Innovation Application Works Collection Activity Has Started!

Leave a Comment