Computational thinking, first introduced by Professor Jeannette Wing of MIT in 2006, refers to the mental processes involved in problem-solving, system design, and understanding human behaviors through the application of fundamental concepts of computer science ( Brennan & Resnick, 2012 ). It emphasizes leveraging logic, abstraction, and algorithms to analyze and resolve issues, transcending the conventional physical and temporal perspectives. It is a way of thinking that represents abstract concepts through symbols, using data as the foundation for problem-solving. The core elements of computational thinking encompass abstraction and modeling, automation and algorithm design, data analysis and processing, as well as iteration and optimization. This approach enables individuals to distill critical information from complex problems, construct computational models, devise efficient algorithms, automate problem-solving processes, and reveal patterns and trends hidden within data, thereby providing robust support for decision-making. Throughout the problem-solving journey, continuous experimentation with new methodologies and strategies, as well as adjustments and refinements to algorithms and models, are essential for achieving optimal outcomes.

Deep learning utilizes the multi-layered structure of networks to learn and represent abstract features within data, automatically extracting these features from data through continuous learning and practice, thereby enhancing the model’s generalization capability. By constructing and training multi-layered neural networks, deep learning achieves data comprehension and interpretation. It is characterized by high automation, robust expressive power, strong adaptability, multi-domain applicability, and scalability. Deep learning models can adapt to diverse tasks and data types, continually improving their performance through iterative training, and have proven to be particularly effective in various fields such as image recognition, natural language processing, and predictive analytics.

A questionnaire survey was employed for empirical research, targeting undergraduate students majoring in Electronic Information Engineering at Nanchang Normal University. After designing the survey questionnaire, peer experts were invited to review and validate it, resulting in a final questionnaire comprising 29 measurement items and structuring 9 latent variables. These variables are decomposition ability, abstraction ability, modeling ability, algorithmic thinking, creativity, collaboration ability, iterative optimization, transfer ability, and evaluation ability. The definitions of these nine dimensions are as follows.

Decomposition Ability refers to the capacity to break down complex problems or systems into smaller, more manageable parts or components that are easier to understand ( Wing, 2006 ).

Abstraction Ability is the ability to extract universal laws, essential characteristics, or core elements from specific situations ( Shute, Sun, & Asbell-Clarke, 2017 ).

Modeling Ability refers to the ability to construct mathematical models, physical models, or conceptual models based on actual problems or scenarios.

Algorithmic Thinking is a methodology of solving problems by applying logic and steps ( Chuechote, Nokkaew, Phongsasithorn & Laosinchai,2020 ; Dagli & Tokmak, 2022 ; Peel & Friedrichsen, 2018 ).

Creativity is the capacity to generate novel, unique, and valuable ideas, products, or solutions ( Kafai, 2016 ; Runciman, 2011 ).

Collaboration Ability is the ability to work together with others to achieve common goals ( Prabawa, Rosjanuardi, & Nurlaelah,2024 ).

Iterative Optimization is the process of continuously improving a product or solution through repeated experimentation, feedback, and adjustments ( Cafarella & Vasconcelos, 2024 ).

Transfer Ability refers to the ability to apply knowledge, skills, or experiences gained in one domain or situation to another, different but related, domain or situation.

Evaluation Ability is the capacity to assess things or proposals objectively, comprehensively, and accurately.

Among them, decomposition ability, abstraction ability, modeling ability, algorithmic thinking, creativity, collaboration ability, iterative optimization, and transfer ability each have 3 measurement items, while evaluation contains 5 measurement items, as shown in Table 1 .

Questionnaires were distributed via the Wenjuanxing platform. A total of 110 questionnaires were sent out, with 104 questionnaires returned and all 104 were valid, resulting in an effectiveness rate of 100%. The basic characteristics of the samples are comprised of four parts: gender, hometown, grade, and family economic status (annual income). The basic characteristics are as follows:

1) Gender characteristics: The frequency of males is 76, accounting for 73.08%; the frequency of females is 28, accounting for 26.92%. Among the respondents, males account for over 70%, while females account for nearly 30%.

2) Hometown characteristics: The frequency of those from rural areas is 77, accounting for 74.04%; the frequency of those from urban areas is 27, accounting for 25.96%. The majority of respondents come from rural areas.

3) Grade characteristics: The frequency of freshmen is 1, accounting for 0.96%; sophomores, 53, accounting for 50.96%; juniors, 49, accounting for 47.12%; seniors, 1, accounting for 0.96%. The respondents are mainly sophomores and juniors, almost evenly split.

4) Family economic status (annual income) characteristics: The frequency of those with an annual income exceeding 500,000 yuan is 1, accounting for 0.96%; between 100,000 and 500,000 yuan, 15, accounting for 14.42%; between 50,000 and 100,000 yuan, 41, accounting for 39.42%; between 20,000 and 50,000 yuan, 29, accounting for 27.88%; and below 20,000 yuan, 18, accounting for 17.31%. The largest proportion is those with an annual income between 50,000 and 100,000 yuan, accounting for nearly 40%, followed by those with an annual income between 20,000 and 50,000 yuan, accounting for nearly 30%. In other words, those with an annual income between 20,000 and 100,000 yuan account for nearly 70% ( Table 2 ).

As can be seen from Table 3 , the Cronbach’s α values of the nine dimensions of the scale, namely decomposition ability, abstract ability, modeling ability, algorithmic thinking, creativity, cooperation ability, iterative optimization, transfer ability, and evaluate, are all above 0.88, indicating good internal consistency. Additionally, the Corrected Item-Total Correlation (CITC) values are all greater than 0.6, suggesting a high correlation among the items. The standardized factor loadings are all above 0.6, the Composite Reliability (CR) values are all greater than 0.7, and the Average Variance Extracted (AVE) values are all above 0.5. Therefore, the scale demonstrates good convergent validity.

According to Table 4 , there is a significant Pearson correlation at the 0.01 level (bilateral) between decomposition ability, abstract ability, modeling ability, algorithmic thinking, creativity, cooperation ability, iterative optimization, transfer ability, and evaluation. The correlation coefficient ranges from 0.800 to 1.000, indicating a high correlation; from 0.600 to 0.799, indicating a strong correlation; and from 0.400 to 0.599, indicating a moderate correlation. According to Table 5 , R² is 83.9%, indicating that the model has a good degree of explanation. In the ANOVA analysis, F is 61.801, P is less than 0.01, indicating that the model is statistically significant.

Notes:**. Significant correlation at the 0.01 level (bilateral).

* p < 0.05, ** p < 0.01.

An in-depth analysis of the data in Table 6 allows us to further explore the specific manifestations of gender differences in various abilities. Firstly, the results of the Levene’s test for homogeneity of variance (p > 0.05) ensure the validity of subsequent t-tests, indicating that the variances of different gender groups in various abilities are equal. This provides a solid foundation for comparing the mean differences between the two groups. Significant differences between genders were observed in decomposition ability, abstract ability, modeling ability, creativity, iterative optimization, transfer ability, and evaluation, with males scoring significantly higher on average than females. This finding may reflect the combined effects of multiple factors. Notably, algorithmic thinking and cooperation ability did not exhibit significant differences between genders, suggesting that in certain cognitive and social skills, males and females may possess similar potential and performance. The lack of gender difference in algorithmic thinking, a crucial ability for solving complex problems, may reflect the universal emphasis placed on logical thinking and problem-solving abilities in modern education, as well as the cultivability of these abilities across genders. Similarly, the insignificant difference in cooperation ability underscores the importance of teamwork and communication skills for all students, regardless of gender.

Firstly, through the Levene’s test for homogeneity of variance, it was confirmed that when comparing the performance of males and females in various abilities, the variances of the data groups were homogeneous (all P-values were greater than 0.05). Subsequently, the results of the t-test revealed significant effects of gender differences on multiple key abilities. Specifically, significant differences were observed between genders in decomposition ability, abstract ability, modeling ability, creativity, iterative optimization, transfer ability, and evaluation, with males consistently scoring higher on average than females in these areas. However, it is noteworthy that algorithmic thinking and cooperation ability did not exhibit statistically significant differences based on gender. This suggests that in terms of algorithmic logical thinking and teamwork collaboration skills, males and females possess similar potential and room for development, and gender is not a decisive factor.

1) Personalized Education and Training: Given the significant gender differences observed in specific abilities, universities should consider implementing more personalized education and training programs. Tailored curriculum content and learning paths should be designed based on the strengths and weaknesses of males and females in different abilities, promoting the comprehensive development of each learner.

2) Cultivation of Gender Equality Awareness: While the data reveals gender differences in certain abilities, this does not imply that one gender is inherently superior to the other in these abilities. Therefore, efforts should be intensified to promote gender equality awareness through education and advocacy, encouraging males and females to learn from each other, leverage each other’s strengths, and mutually improve.

3) Encouraging Cross-Gender Collaboration: In areas where algorithmic thinking and cooperation ability do not exhibit significant gender differences, active promotion of cross-gender collaboration and communication should be pursued. Through inter-gender teamwork, not only can innovative thinking be stimulated, but also team diversity and inclusiveness can be enhanced, thereby driving the production of higher-quality outcomes.

4) Data-Driven Decision-Making: The integration of deep learning and computational thinking enables in-depth mining and efficient utilization of data to knowledge. Deep learning models, through large-scale data training, can automatically learn and optimize the processing capabilities for complex tasks. A data-driven decision-making process can ensure that students receive fair opportunities and treatment based on their actual abilities.

5) Continuous Monitoring and Evaluation: The relationship between gender and abilities is a complex and dynamically evolving topic. Therefore, it is necessary to continuously monitor and evaluate the manifestation of gender differences across different periods and fields, allowing for timely adjustments to educational strategies to adapt to changing times and demands.