Innovations in Introductory Programming Education: The Role of AI with Google Colab and Gemini

Abstract

This study explores the impact of artificial intelligence on the teaching of programming, focusing on the GenAI Gemini tool in Google Colab. It evaluates how this technology influences the comprehension of fundamental concepts, teaching processes, and effective teaching practices. In this research, students’ motivation, interest, and satisfaction are determined, as well as the fulfillment and surpassing of their learning expectations. With a quantitative approach and a quasi-experimental design, an investigation was carried out in seven programming groups in a polytechnic university in Guayaquil, Ecuador. The results reveal that the use of GenAI significantly increases interest in programming, with 91% of the respondents expressing increased enthusiasm. In addition, 90% feel that the integration of GenAI meets their expectations, and 91% feel that it has exceeded those expectations in terms of educational support. This study evidences the value of integrating advanced technologies into education, suggesting that GenAI can transform the teaching of programming. However, successful implementation depends on timely training of educators, ethics training for students, ongoing interest in the technology, and a curriculum design that maximizes the capabilities of GenAI.

1. Introduction

In a world where artificial intelligence (AI) technologies are making inroads, traditional educational barriers are being challenged as never before [1,2]. This transformation places the student at the center of the learning process where the role of the professor is also redefined [3,4]. The integration of technologies such as generative artificial intelligence (GenAI) in higher education promotes AI-assisted learning that optimizes teaching but, at the same time, raises important ethical questions [5,6].

On a more specific level, and referring to the training of engineering students, the understanding of programming concepts at an early introductory stage of university studies is essential for academic success, as it lays the foundation for learning more complex content [7,8,9,10]. In the development of this learning, however, many young people encounter difficulties in mentally operating with abstract ideas, such as, for example, instructions for designing control or repetition structures, as well as the storage of data in vectors and matrices [11]. In this context, the integration of GenAI in the classroom offers an effective response to overcome these barriers, providing personalized support and algorithmic resources that facilitate the assimilation of these concepts [12,13,14].

In this sense, this study explores how GenAI tools, such as Gemini in a specific context, can enrich the understanding of programming and the implications this has for teaching and learning [6,15]. The incorporation of GenAI into educational curricular content can transform the way in which various disciplines, including programming, are taught and learned [16]. Platforms such as Google Colab in combination with Gemini facilitate access to learning resources, and create interactive environments that foster dynamic and adaptive learning [17,18].

In this way, motivation and interest are fostered, which are fundamental for the students’ commitment and academic success; enabling learning strategies that favor individual and group performance [19]. This analysis also aims to determine whether the use of GenAI tools manages to enhance these aspects, as well as to examine whether students’ learning expectations are met or even exceeded [5,20,21].

The aim of this work is to evaluate the impact of generative artificial intelligence, specifically Gemini in Google Colab, in the teaching of programming and to contribute to the literature in specific scenarios. It analyzes how these tools influence the comprehension of fundamental programming concepts, in complete teaching processes and effective teaching practices where motivation, interest, and satisfaction of students, as well as in meeting and exceeding their learning expectations are evidenced, and favor educational development. Through classroom experience and student perceptions, we also seek to provide a comprehensive view of the impact that GenAI has on the teaching of programming, with the objective of identifying effective educational practices focused on the current needs of students [22]. To determine the impact of using GenAI tools in teaching programming, research questions are established in three relevant areas (see

Table 1. Research questions for the study.

Research Question	Area
RQ1: how does the use of generative artificial intelligence tools, such as Gemini in Google Colab, affect the comprehension of fundamental programming concepts among students?	Comprehension of the content
RQ2: to what extent do artificial intelligence tools facilitate the learning of complex subjects compared to traditional teaching methods?	Comprehension of the content
RQ3: how does the integration of artificial intelligence in the classroom influence students’ interest in programming?	Interest level
RQ4: what is the relationship between the use of artificial intelligence tools and students’ motivation to learn programming?	Interest level
RQ5: do you meet students’ learning expectations when using artificial intelligence in your programming classes, and to what extent do you exceed these expectations?	Learning expectations
RQ6: what are students’ perceptions of the support they receive for their programming learning through artificial intelligence tools?	Learning expectations

).

The first question focuses on the use of these tools that are integrated into the programming course, and improves the comprehension of basic concepts, such as the use of variables, control structures, selection, or repetition, as well as compound structures such as vectors and arrays. The purpose is to evaluate whether GenAI (Gemini) provides tangible support in the assimilation of these essential fundamentals [11]. The second question focuses on the ease of learning complex topics, exploring whether GenAI helps students to cope with and understand more advanced content for their academic development [6]. The third and fourth questions address interest in using GenAI tools, and how they influence motivation in the classroom [11]. The fifth and sixth questions focus on expectations about student learning when using GenAI compared to traditional methods [23]. This is fundamental to understanding whether the integration of advanced technologies, platforms, and intelligent environments can make learning more engaging and motivating [7]. In addition, it evaluates the fulfillment and exceeding of expectations in relation to student satisfaction with the use of GenAI with Google Colab, in a specific scenario and over time, opening spaces to contribute with new research and add to the recent ones [24].

2. State of the Art

The study by Perezchica-Vega et al. [25] investigated professors’ experience with generative artificial intelligence (GenAI) in education. Using a quantitative methodology, concerns about academic honesty are identified, and the benefits of GenAI for learning, such as data analysis and the creation of didactic materials, were highlighted. It is concluded that its integration could improve educational quality, although adjustments in evaluation mechanisms are still required. Also, in Hernández González et al. [26], the perceptions of university students about Generative Artificial Intelligence (GenAI) in education were analyzed. Through a qualitative and descriptive approach, opinions about its advantages and disadvantages were identified. Most students accepted its use, highlighting improvements in their learning, although they expressed concerns about ethics and privacy in the use of this technology. For their part, Padilla Piernas and Martín-García [27] investigated the perceptions of Spanish university professors about generative artificial intelligence (GenAI) in higher education, using the AETGE/GATE model. At the same time, the importance of training programs for an effective implementation of GenAI in learning is emphasized. The importance of training programs for effective implementation of GenAI in learning is emphasized, while Solano Hilario et al. [28], in their research, analyzed how the use of Generative Artificial Intelligence (GenAI) tools improve student learning compared to the use of traditional tools. Through a systematic review of the literature, the results indicate an equal preference between both styles, highlighting personalization and feedback as key factors to optimize learning.

On the other hand, the research conducted by García Peñalvo et al. [29] analyzed the impact of generative artificial intelligence, especially ChatGPT (2022 and 2023), on education. Through a systematic review of the literature, the advantages and disadvantages of its use in learning are identified. Although it presents a potential for improving educational processes, it also poses risks and limitations that require careful evaluation and an adequate response in teaching practice. In addition, Schiavo et al. [30] consider the use of the Technology Acceptance Model (TAM) to understand usage intention, individual perception, and adaptation in the age of AI literacy. In this regard, they identify factors that influence human behavior patterns, such as perceived usefulness, perceived ease of use, perceived security, perceived trust, perceived confidence, respect for privacy, perceived anxiety, perceived social influence, and perceived compatibility. Thus, these data help to facilitate behavioral patterns, improve educational inclusion, and support the development of the teaching and learning process. According to Solomovich and Abraham [31], the Technology Acceptance Model (TAM) contributes to the collection of data that provides insight into behavioral traits related to trust levels in chatbot users. These empirical results offer the opportunity to further improve these technology platforms for the benefit of all stakeholders. A necessary aspect to highlight is that of intrinsic motivation Verdugo-Castro et al. [15].

Likewise, Henrique-Sanches et al. [32], in their quasi-experimental research, studied the impact of hands-on activities in the Skill and Simulation Laboratory after the pandemic. Although motivation levels did not change significantly, the high intrinsic motivation and regulation identified suggest that active methodologies are key to effective learning in medical students. In the field of robotics, also based on AI in educational applications, Vitale and Dello Iacono [33], through explanatory research, studied the use of social robots as an inclusive educational technology to improve mathematics learning. The research showed that the Pepper robot significantly increased student engagement by offering personalized support. Its ability to create a stimulating learning environment tailored to diverse needs was highlighted.

Some studies, such as the one by Rossettini et al. [34], evaluated the accuracy of AI chatbot responses (ChatGPT-4, Microsoft Copilot, and Google Gemini) in health science input tests. The intelligent assistants or chatbots performed well overall, although with differences in accuracy and narrative coherence. In the study by Fabijan et al. [35], the ability of AI models to interpret MRI video sequences was analyzed to identify and analyze pediatric brain tumors. The models failed to accurately identify lesions, highlighting the need for better targeted integration of AI into medical diagnostics. While Rossettini et al. [34] evaluated the accuracy and narrative coherence of AI chatbots when using standardized exam questions in health sciences, Fabijan et al. [35] explored the limited capabilities of AI models (ChatGPT and Gemini Pro) when analyzing medical MRI videos, i.e., the need to improve the experience of constructive use of AI tools in medical domain activities requires incorporating reliable meaningful experiences that enhance the understanding of content for an AI.

Students in higher education show significant interest in AI chatbots, driven by their ability to provide immediate feedback, assist with academic tasks, and boost motivation. However, concerns regarding the reliability of AI responses have also been noted, as these tools sometimes produce inaccurate or inconsistent information. As Schei et al. [36] points out, the appeal of AI chatbots lies in their immediate usefulness and accessibility, making them particularly attractive for students seeking quick solutions to academic challenges.

In a similar vein, Alnasib and Alharbi [37] highlights that the integration of AI tools like Gemini into the teaching of English as a foreign language generated student interest, although several limitations were identified, such as repetitive responses and issues with accuracy. While these tools were shown to enhance motivation, their practical effectiveness in fostering language learning remains limited. This suggests that, despite their potential, improvements in AI response quality are necessary to optimize their role in educational contexts.

The study by Zichar and Papp [38] explored students’ expectations when using AI tools for 3D modeling and programming tasks. While these tools offer valuable initial support in code generation, students are not yet able to fully delegate these tasks to AI. Torres-Peña et al. [39], in their paper, discussed how AI tools such as ChatGPT and Wolfram Alpha have raised students’ expectations in mathematics, particularly in calculus, by helping them solve derivative problems with greater accuracy and conceptual understanding. Conversely, Yoseph et al. [40] revealed that while spine surgery patients prioritize clear answers, their learning expectations also focus on a comprehensive understanding of complex medical procedures, such as those facilitated by AI. This is evidence of how patients can benefit from the clarity and conciseness that AI can provide when explaining intricate topics.

Similarly, Almassaad et al. [41] describes how the use of GenAI tools in higher education in Saudi Arabia has raised expectations of greater efficiency and academic support. However, students also express concerns about the accuracy of the information provided by these tools. Taken together, these studies shed light on how AI tools are shaping expectations in various domains, including education and healthcare, while emphasizing challenges related to the reliability and practical use of these technologies.

2.1. A Blended Constructivist and Project-Based Learning (PBL) Approach with the Use of Artificial Intelligence Tools in Education

Constructivist learning theories, articulated by Bruner, Vygotsky, and Piaget, emphasize that knowledge is actively constructed in the mind of the learner through interactive and meaningful experiences. These principles are exemplified in recent studies that integrate AI tools into educational practices. Rossettini et al. [34], for example, investigated the use of AI chatbots for exam assessments, framing it as an educational project in which students interact with technology to assess their performance in specific contexts. This approach allows students to reflect on the problem-solving capabilities of AI, in line with the constructivist view that interaction with tools and feedback are essential for knowledge construction.

Similarly, Alnasib and Alharbi [37] explored how the AI tool Gemini supports English language learners, highlighting how students actively engage with linguistic challenges. By providing immediate feedback and facilitating practice, Gemini helps learners respond to existing uncertainties, encouraging critical reflection, an integral component of constructivism. Zichar and Papp [38] extends this idea to complex tasks such as 3D modeling and programming, where students use AI tools to address challenges in coding and design. In these tasks, AI serves as a resource and collaborator, promoting deeper engagement with the technology as students integrate it into problem-solving processes. This aligns with the principles of project-based learning (PBL), where students address real-world problems by applying and reflecting on their knowledge.

In the field of mathematics, Torres-Peña et al. [39] examined how AI tools, such as ChatGPT and Wolfram Alpha, improve the teaching of calculus. These tools enable students to solve derived problems more accurately, providing immediate feedback that encourages autonomous learning. The use of AI in this context supports the development of mathematical understanding, and reinforces constructivist values by allowing students to actively interact with the material, verify solutions, and reflect on their learning process. Similarly, Yoseph et al. [40] demonstrates the role of AI in medical education, specifically in providing clear and understandable explanations of complex procedures, such as cervical discectomy surgeries. By interacting with AI-generated responses, patients and students develop a more interactive understanding of medical concepts, which encourages informed decision-making and reflective learning.

Almassaad et al. [41] highlighted the integration of AI tools in higher education in Saudi Arabia, focusing on language learning. This study highlights the use of AI as a facilitator of project-based learning, where students work on specific tasks, such as improving language proficiency, while reflecting on their progress and adjusting their learning strategies. Incorporating AI tools into these projects allows students to solve problems in practical, real-world contexts, reflecting key features of project-based learning and constructivist learning environments.

The use of AI in education allows students to face authentic and challenging tasks that require an integration of cognitive and technical skills, fostering autonomous and collaborative learning, which are characteristics of a blended (constructivist and PBL) framework. In addition, students interact with AI technologies, allowing them to build knowledge through practice, problem solving, and continuous feedback (see

Table 2. Work related to the areas of study.

Scope	References
Comprehension of the content	Rane et al. [6], Imran and Almusharraf [7], Solano Hilario et al. [28], Rossettini et al. [34], Fabijan et al. [35]
Interest level	Edwards et al. [8], Hernández González et al. [26], Padilla Piernas and Martín-García [27], Schei et al. [36], Alnasib and Alharbi [37]
Learning expectations	Vitale and Dello Iacono [33], Zichar and Papp [38], Torres-Peña et al. [39], Yoseph et al. [40], Almassaad et al. [41]

).

2.2. A Comparative Analysis of Gemini’s Effectiveness Against Other Generative AIs

In recent years, artificial intelligence (AI) models have evolved from being simple support tools to become powerful enablers of learning, especially in educational and medical contexts. Among the most prominent models is Gemini, which has proven to be an effective tool in generating understandable and accessible content for a wide range of users. Unlike other models such as ChatGPT, which is distinguished by its depth of understanding and its ability to generate accurate and detailed answers in specialized fields, Gemini stands out for its clarity and readability, making it particularly valuable in scenarios where readability and accessibility are crucial. This was evidenced in studies such as that of Hanci et al. [42], who observed that Gemini produced easier to understand answers compared to other models in the context of palliative care, although improvements were required in the quality and depth of the content.

Comparative analysis between Gemini and other AI models, such as ChatGPT and Bard, has yielded interesting results in terms of their effectiveness in content comprehension. In studies conducted by Karaca [43] and Gomez-Cabello et al. [44], it was observed that Gemini is able to generate content suitable for various educational levels, showing an outstanding ability to adapt its language and complexity according to the user’s needs. However, in terms of depth and accuracy in technical areas, ChatGPT outperformed Gemini, as seen in research on interpreting medical results and decision-making in plastic surgery. Despite these differences, both models show significant potential for improving the accessibility of knowledge, especially in contexts where users do not have a deep understanding of the subject matter.

The study of Gemini’s effectiveness also highlights its potential as a key tool for improving learning expectations. In its comparison with ChatGPT-4 and other models in educational tasks, Gemini showed superior performance in terms of content appropriateness at lower or simple educational levels. This is particularly relevant when analyzing the quality of the generated texts in terms of readability. According to Karaca [43], AI models such as Gemini are able to adapt the difficulty level of texts to the needs of the audience, making this AI suitable for a diverse range of users, from primary school students to university students. In the context of education, this ability to generate accessible content facilitates learning and improves understanding of complex concepts, positioning Gemini as a potential ally in the teaching of disciplines that require a clear and direct pedagogical approach.

However, Gemini’s effectiveness as a learning enhancer must also be evaluated in terms of its performance in solving specialized questions and the ability to generate accurate answers in complex areas. While Gemini is effective at generating readable and understandable answers, its performance on more complex tasks, such as interpreting medical results or solving advanced chemistry problems, often falls short of the level of accuracy and depth demonstrated by ChatGPT-4. This point was highlighted by Kharchenko and Babenko [45], who observed that, while Gemini is proficient in tasks that do not require deep logical reasoning, in complex tasks related to science and medicine, ChatGPT showed a greater ability to handle abstract concepts and generate more complete solutions. In this sense, although Gemini represents a significant advance in the generation of accessible content, its role as an educational and support tool remains complementary to more specialized models, such as ChatGPT.

Gemini has an advantage in the readability of answers, especially compared to ChatGPT-4, which tends to generate more complex answers. However, in terms of accuracy and quality of answers, especially in specialized contexts such as medicine or chemistry, Gemini was outperformed by ChatGPT in most of the studies analyzed (see

Table 3. Works relating Gemini to the fields of study.

Scope	Relationship	Reference
Comprehension of the content	Gemini shows significant differences in quality, readability, clarity, and precision in specific fields, i.e., it is more concise and excels in the readability of texts.	Hanci et al. [42], Karaca [43], Sonmezoglu and Sonmezoglu [46]
Interest level	Gemini shows competence in providing quick and direct answers that capture the user’s attention, but still lacks the ability to generate deeper interest in more complex topics.	Kharchenko and Babenko [45], Meyer et al. [47], Durmaz Engin et al. [48]
Learning expectations	Gemini performs competitively, shows a better fit in terms of readability at simpler educational levels and limitations in tasks requiring deep logical reasoning.	Gomez-Cabello et al. [44], Farghal and Haider [49], Is and Menekseoglu [50]

).

In summary, the use of generative artificial intelligence, specifically Gemini in Google Colab, influences university students’ understanding and interest in learning programming, from basic to advanced levels, which implicitly aligns with these approaches by promoting active learning, interaction with technology, and problem solving.

3. Materials and Methods

This study implements a research methodology with a quantitative approach and a quasi-experimental design, which falls within an empirical-analytical framework. Seven courses of programming, a subject with high enrollment and belonging to the first level of studies in engineering degrees of a polytechnic university in the city of Guayaquil, Ecuador, took part. A work strategy was applied based on laboratory activities in six phases in each two-hour laboratory day, twice a week, for five months. In addition, the survey technique is applied to a population of 250 students at the end of the five months, at the end of the academic period. With a confidence level of 95% and a margin of error of 5%, a set of six structured questions with Likert scale are established, and focus on important aspects that allow us to evaluate how the GenAI tool, such as Gemini in Google Colab, affect the learning experience of students.

The six phases of work as shown in Figure 1 are established on the basis of activities planned by the team of professors, known as teaching staff, who run these courses [51]. The cloister of professors are professionals, professors, and researchers who share in their teaching activity the development of the subject of programming [52,53]. This group of professors is responsible for designing, developing, implementing, and evaluating the activities of the subject syllabus [54,55]. From an instructional design applied in a basic classroom [56], professors integrated research-based activities in an educational environment, based on their didactic experience and techno-pedagogical skills [57], as well as curricular adaptations based on the results of the evaluations of the course activities [11,58,59,60].

3.1. Working Phase 1

In this phase, the course’s professor presents a problem to be solved algorithmically, including the use of simple and compound variables, as well as the control, selection, and repetition structures found in the syllabus, as an introduction to specific topics. With this, he begins the subject by presenting a problem, applies the technique of algorithms by means of pseudocode and its corresponding flowchart, and then its implementation in Python. Thus, the explanation, dialogue, and discourse are directed to its corresponding analysis from the accompaniment of schemes, such as pseudocode and its corresponding flowchart, to finally implement it [37]. In this phase, the use of programs such as PseInt (available at https://pseint.sourceforge.net/, accessed on 30 November 2024), among others, allows students to become familiar with a process where algorithmic reflection and analysis are required for translation into programming language (see Figure 1a).

3.2. Working Phase 2

In this phase, the course professor requests the use of the Google Colab (Available at https://colab.research.google.com/, accessed on 30 November 2024) development environment for a variety of factors. Among these factors is the use of Gmail accounts which, in the context of this study, all students use on their Android mobile devices. With this, under this particular context, it encourages mobility, quick access, automatic synchronization, the use of default tools, and an optimized interface, allowing students to collaborate in real time, an environment that facilitates their adaptability, operability, access to the files created, and their backup and saving in the Google Drive cloud, which allows for a convenient and efficient experience. They are also introduced to Gemini’s built-in assistant [34] (see Figure 1b).

3.3. Working Phase 3

In this phase, the students use the teams for individual and personal work on the development of a solution. In other words, the analysis developed is implemented in the Google Colab environment. With this, reflection, reasoning, logic, memorization, and recall of what has been learned from the theoretical aspect converge in this space [41]. The resources used by the students are those documents produced by the tutor professors, and hosted in the virtual environment of the course, in a Moodle (available at https://moodle.org/, access on 30 November 2024) work environment. In other words, having resources in a virtual classroom that integrates external technologies, such as Google applications, allows for an ideal working ecosystem for the subject of programming (see Figure 1c).

3.4. Working Phase 4

In this phase, students establish a collaborative workspace with the professor. The professor shares the link of his proposal with the students for a collaborative access work. With this, students can develop two situations. The first situation is to work together with the professor in their environment, propose changes to their code, exchange lines of instruction, and test on the professor’s proposal combined with any student [51]. The second situation is that the student can improve their proposal from the collaborative work of all, in the company of the professor and their fellow peers, knowing and applying good programming practices (see Figure 1d).

3.5. Working Phase 5

In this phase, students share and present their working prototype so that the professor and their peers can improve the lines of instruction. This collaborative work is developed among assigned peers, so that each code can be reviewed and improved. In this way, students become peer reviewers in an environment where everyone can contribute to the changes [54]. The importance of having additional projection equipment, as well as remote internet access using desktop computers and mobile devices, allows the professor to have movement within the work lab, generating a close and, at the same time, a mobile space in this scenario that combines the traditional with the experience of cloud-based tools (see Figure 1e).

3.6. Working Phase 6

In this phase, the professor and the students of the course begin their interaction with GenAI Gemini, starting with small queries [38]. These queries are first carried out by the professor and projected to the whole course. In this way, the student begins to recognize the style of `Prompt’ to design or perform in order to train or guide Gemini during the refinement of their proposal. This space becomes a question and answer experience with the conviction to test and improve. In many cases, the professor tests the answers, and instructs Gemini to rewrite the code, or not to use complexity but simplicity in the lines. In this scenario, the possibilities of obtaining short lines, use of unstudied functions, complexity in syntax for first year students may provoke a preoccupation with learning what the AI knows rather than a motivation to learn to program [60]. Finally, the solution to be obtained, from the initial approach, is chosen by the group after establishing the conditions of the requested problem in the elaboration of the code, which is a simple code for its understanding, avoiding codes with complex syntax (see Figure 1f).

4. Results

The survey applied electronically, using the Google Forms tool, included six questions that addressed different sub-fields of study, and were structured on a five-level Likert scale of ‘Strongly Disagree’, ‘Disagree’, ‘Neither Agree nor Disagree’, ‘Agree’, and ‘Strongly Agree’.

Table 4. Relationships between variables and questions in the evaluation of the impact of the use of GenAI Gemini in the subject of programming.

Sub-Area	Variable	Questionnaire Question
Understanding programming concepts	Level of understanding of basic concepts (e.g., arrays, control structures) using GenAI.	Q1: has the use of GenAI tools such as Gemini in Google Colab improved my understanding of programming concepts?
Ease of learning complex issues	Perceived ease of understanding complex issues when using GenAI tools	Q2: do I find it easier to understand complex topics, such as matrices, when I use GenAI in programming classes?
Interest in the subject matter	Level of interest in learning programming using GenAI compared to traditional methods	Q3: has the use of GenAI in programming classes increased my interest in learning more about the subject?
Motivation for learning programming	Perceived level of motivation to use GenAI tools	Q4: do I feel more motivated to learn programming when using GenAI tools compared to traditional methods?
Fulfilling expectations	Satisfaction regarding the fulfillment of learning expectations with GenAI	Q5: does the integration of GenAI in my programming classes meet my expectations of understanding the concepts effectively?
Exceeding expectations	Perception of support received for learning programming through GenAI	Q6: do I feel that the use of GenAI has exceeded my expectations in terms of support for learning programming?

presents the relationship between the variables determined and the questions of the questionnaire in the evaluation of the impact of the use of generative artificial intelligence, Gemini, in the subject of programming in Google Colab.

Table 4 offers an approach that integrates the relationship between the assessed sub-domains, the relevant variables, and questionnaire questions in the context of the use of GenAI tools in programming education [61]. It addresses the understanding of programming concepts, where the level of understanding of basic notions, syntax, control structures, and data storage with the use of arrays and matrices is evaluated, facilitated by the experience with Gemini GenAI. Through the question formulated in the first sub-area, we seek to determine whether tools such as the Gemini GenAI in Google Colab really contribute to improving the understanding of these fundamental concepts [62]. The level of understanding is fundamental to determine a solid foundation in programming fundamentals and is essential for academic success. Another relevant aspect is the ease of learning complex topics, which focuses on students’ perception of their ability to understand advanced content with the support of the GenAI. The associated question invites participants to reflect on whether the GenAI facilitates the understanding of difficult topics, such as the use of vectors and matrices for data storage, during lectures. This highlights the potential of GenAI to simplify the learning of complex content [14].

Interest in the subject of programming is also assessed, focusing on the level of interest students show in learning programming through the use and support of a GenAI compared to traditional methods. This question shows how the integration of GenAI can foster greater interest in learning, which is a key means of academic engagement. In relation to motivation to learn programming, we seek to understand whether students feel more motivated when using GenAI tools. The question explores whether this technology, such as the use of GenAI tools, improves motivation compared to conventional teaching approaches. Motivation is a decisive factor influencing dedication and commitment to learning.

It also examines the fulfillment of expectations by analyzing students’ satisfaction with how the GenAI meets their learning expectations. The question posed focuses on whether the integration of GenAI in programming classes, in conjunction with the professor, enables students to understand the concepts effectively. Satisfaction in the learning process is important for assessing the quality of teaching. Exceeding expectations is considered by assessing how students perceive the support received for their learning through the GenAI. This question asks whether the GenAI has exceeded expectations in terms of educational support. This aspect assesses the fulfillment of expectations, and highlights the additional impact the GenAI can have on the educational experience.

4.1. Analysis of the Survey Responses

The results generally show a positive trend towards the integration of GenAI tools for learning the subject of programming. Most students report significant improvements in their understanding, motivation, and interest, highlighting the potential of GenAI to transform programming education. These findings highlight the need to continue to explore and apply GenAI technologies in educational settings to foster more dynamic and effective learning (see Figure 2).

For question Q1 of the survey, 55% of the students strongly agreed that the use of GenAI has improved their understanding of basic concepts, 32% agreed, and 12% neither agreed nor disagreed, while only 1% disagreed. This indicates that 87% of the participants agree, indicating a strong acceptance of the effectiveness of the GenAI Gemini in Google Colab in learning programming fundamentals (see Figure 3).

For question Q2 of the survey, 47% of students said they strongly agreed that GenAI has made it easier for them to understand complex topics, with a further 42% saying they agreed. A further 9% neither agreed nor disagreed, while a further 2% disagreed. This result shows that the GenAI Gemini in Google Colab not only helps with basic concepts, but is also useful for more advanced topics (see Figure 4).

For question Q3 of the survey, 51% of students strongly agree that the use of GenAI Gemini in Google Colab has increased their interest in learning more about programming. A further 40% agreed. A further 8% neither agreed nor disagreed, while 1% disagreed. This finding is critical, as greater interest is often correlated with better participation and engagement in learning (see Figure 5).

For question Q4 of the survey, 46% of students strongly agreed that they feel more motivated to learn programming because of the GenAI Gemini tools in Google Colab. A further 40% agreed. A further 10% neither agreed nor disagreed, while 4% disagreed. This increase in motivation is essential to maintain interest and perseverance in learning (see Figure 6).

For survey question Q5, 46% of students strongly agreed that the integration of the GenAI Gemini with Google Colab meets their learning expectations, indicating that, overall, expectations are being met. A further 44% agreed, and 9% neither agreed nor disagreed, while 1% disagreed, suggesting that there is still room for improvement in satisfaction (see Figure 7).

For question Q6 of the survey, 49% of students felt that they strongly agreed that the use of GenAI has exceeded their expectations in terms of learning support. This is encouraging, as it implies that the GenAI Gemini in Google Colab is not only meeting expectations, but is also providing additional value in the educational process. An additional 42% agreed. A further 8% neither agreed nor disagreed, while 1% disagreed (see Figure 8).

The results show a positive trend towards the integration of GenAI tools in programming education. Most students report significant improvements in their understanding, motivation, and interest, highlighting the potential of GenAI to transform programming education. These findings point to the need for further research in exploring and applying GenAI technologies in specific educational settings to foster more dynamic and effective learning.

Finally, the results of the number of students who passed the subject in percentages of the period from October 2023 to March 2024 compared to the period from April to September 2024, show an improvement of 1%, using strategies that involve intelligent tools, such as the integration of the GenAI Gemini, which benefit students in their learning process and professors in their continuous monitoring or follow-up. That is, for the study period where workspaces for code generation were integrated, the perception of students has been of greater importance to have an intelligent assistant in conjunction with the professor, establishing a way of working in the classroom for analysis and implementation supported with a GenAI tool [58,59,60] (see Figure 9).

4.2. Statistical Analysis

The results of the correlation analysis reveal significant relationships between the variables. Pearson’s and Spearman’s correlation coefficients, together with their respective p-values, are presented in

Table 5. Correlation table.

Variable		Q1	Q2	Q3	Q4	Q5	Q6
Q1	Pearson’s r	–
	p-value	–
	Spearman’s rho	–
	p-value	–
	Kendall’s Tau B	–
	p-value	–
Q2	Pearson’s r	0.598	–
	p-value	<0.001	–
	Spearman’s rho	0.633	–
	p-value	<0.001	–
	Kendall’s Tau B	0.601	–
	p-value	<0.001	–
Q3	Pearson’s r	0.579	0.619	–
	p-value	<0.001	<0.001	–
	Spearman’s rho	0.574	0.628	–
	p-value	<0.001	<0.001	–
	Kendall’s Tau B	0.540	0.595	–
	p-value	<0.001	<0.001	–
Q4	Pearson’s r	0.616	0.653	0.672	–
	p-value	<0.001	<0.001	<0.001	–
	Spearman’s rho	0.611	0.693	0.696	–
	p-value	<0.001	<0.001	<0.001	–
	Kendall’s Tau B	0.580	0.657	0.660	–
	p-value	<0.001	<0.001	<0.001	–
Q5	Pearson’s r	0.532	0.457	0.664	0.596	–
	p-value	<0.001	<0.001	<0.001	<0.001	–
	Spearman’s rho	0.581	0.542	0.691	0.659	–
	p-value	<0.001	<0.001	<0.001	<0.001	–
	Kendall’s Tau B	0.546	0.509	0.665	0.616	–
	p-value	<0.001	<0.001	<0.001	<0.001	–
Q6	Pearson’s r	0.602	0.586	0.681	0.766	0.614	–
	p-value	<0.001	<0.001	<0.001	<0.001	<0.001	–
	Spearman’s rho	0.604	0.641	0.700	0.782	0.699	–
	p-value	<0.001	<0.001	<0.001	<0.001	<0.001	–
	Kendall’s Tau B	0.575	0.604	0.669	0.748	0.664	–
	p-value	<0.001	<0.001	<0.001	<0.001	<0.001	–

.

In the context of using generative artificial intelligence tools in teaching programming, a high and positive Pearson’s r (e.g., 0.6 or higher) between variables, such as concept understanding and motivation, is evidence that as students feel a greater understanding of programming concepts, they also tend to feel more motivated to learn.

A low p-value associated with Pearson’s r (typically less than 0.05) indicates that there is a high probability that the observed correlation is not due to chance, suggesting a significant relationship between the variables evaluated. In the context of correlation analysis, the p-value allows us to determine whether the relationship between the variables is statistically significant or whether it could be attributable to chance. For example, if the p-value associated with a Pearson correlation is less than 0.001, as observed in the data presented, this is evidence of a strong relationship between the variables in a significant way. The p-value complements the Pearson correlation coefficient, providing information on the reliability of the observed relationship.

A low p-value, together with a considerable correlation coefficient, supports the claim that the use of generative artificial intelligence tools in teaching programming has a positive impact on aspects such as concept understanding, motivation, and student interest [63,64].

Pearson’s r value for the relationship between question Q2 (perceived ease of understanding complex topics) and question Q1 (level of understanding of basic concepts) is 0.598. This determines a moderate positive correlation; that is, students who report a better understanding of basic concepts also tend to find it easier to understand complex topics. This is further evidence that the use of GenAI tools such as Gemini in Google Colab can facilitate the understanding of basic concepts and the transition to the cognitive development of more advanced concepts.

Pearson’s r coefficient between question Q3 (interest in learning programming) and question Q1 (level of understanding of basic concepts) is 0.579, which also indicates a moderate positive correlation. This is evidence that those students who feel that they have improved their understanding tend to show a greater interest in programming. This further implies that improving understanding can have a positive effect on motivation and interest in the subject.

The relationship between question Q4 (motivation when using GenAI tools) and question Q1 (level of understanding of basic concepts) shows a coefficient of 0.616, indicating a moderate-high positive correlation. Students who feel more motivated to learn programming are those who also report a better understanding of the concepts. This is evidence that motivation and understanding are intimately connected, which is a direct result of using GenAI like Gemini in Google Colab in learning [13,57].

The analysis of the relationship between question Q5 (satisfaction with fulfillment of expectations) and question Q1 (level of understanding of basic concepts) shows a Pearson’s r value of 0.532. Although this value is lower than in other cases, it still indicates a significant positive correlation. This is evidence that students who feel that their learning expectations have been met are those who also have a better understanding of basic concepts. This reflects that the effectiveness of the GenAI Gemini with Google Colab in the classroom is aligned with students’ expectations.

The relationship between question Q6 (perception of support received) and question Q1 (level of understanding of basic concepts) has a coefficient of 0.602. This value indicates a significant positive correlation, suggesting that students who feel that the GenAI Gemini with Google Colab has exceeded their expectations in terms of support also tend to have a better understanding of programming concepts. This reinforces the idea that the implementation of GenAI tools meets expectations, and offers additional support in learning.

The use of Spearman’s rho is a non-parametric measure that assesses the correlation between two variables, considering their ordinal relationship. In the analysis conducted, the results showed significant Spearman’s rho values, which ranged from 0.574 to 0.782, indicating moderate to strong positive correlations between the variables evaluated. These results show that as students experience a greater understanding of programming concepts, and an increase in their motivation, they also tend to report greater interest and satisfaction with the use of generative artificial intelligence tools. The associated p-values, all less than 0.001, reinforce the statistical significance of these correlations, implying that it is highly unlikely that these observed relationships are due to chance. Taken together, these findings highlight the positive impact of generative artificial intelligence on programming instruction, supporting the idea that its integration can facilitate more effective and motivating learning.

In the results obtained, Kendall’s Tau B values ranged from 0.540 to 0.748, evidencing a moderate to strong positive correlation between the variables assessed, similar to that observed with Spearman’s rho. The corresponding p-values, all less than 0.001, indicate that these correlations are statistically significant, reinforcing confidence that the findings reflect real relationships, and are not the product of chance. This implies that, as with the other correlation measures, there is a clear association between the use of generative artificial intelligence tools and improvements in students’ understanding, motivation and interest in the subject, emphasizing the value of GenAI in education.

The use of GenAI tools in teaching programming is linked to significant improvements in student understanding, motivation, and satisfaction. In addition to facilitating the assimilation of concepts, it also increases interest and motivation, meeting and even exceeding student expectations. This reality is of great importance when integrating GenAI Gemini with Google Colab in the teaching of programming, thus promoting more effective and stimulating learning.

5. Discussion

The findings of this study highlight the significant impact of generative artificial intelligence tools in teaching programming. Through an exhaustive analysis, it became evident that the use of GenAI not only improves the comprehension of fundamental concepts, but also boosts students’ motivation and interest [37]. The correlations observed, both Pearson’s r and Spearman’s rho and Kendall’s Tau B, indicate that as students interact with technologies such as Google Colab and Gemini, they experience greater satisfaction and a higher fulfillment of expectations in their learning process [38]. This positive effect suggests that GenAI can play an important and significant role in transforming programming education by offering a more dynamic and accessible approach to tackling complex topics [27].

However, it is essential to consider the implications of these results for curriculum design and professor training. The integration of GenAI in the classroom should not be merely mechanical or technical; it is critical that educators understand how to use these tools effectively to maximize their educational potential. In addition, more research should be conducted to explore how different contexts and student populations may influence the effectiveness of these tools [26]. As technology advances, it is critical to adapt teaching strategies to ensure that all students benefit from innovations in learning, thereby fostering a more inclusive, relevant, and meaningful education.

Indeed, this study corroborates the works [6,7,8], revealing that generative artificial intelligence tools, such as Gemini at Google Colab, have a significant effect on the understanding of fundamental concepts. Students report an improvement in their assimilation of basic topics, suggesting that GenAI facilitates the teaching of essential fundamentals such as vectors and arrays, as well as the design of control structures, selection, and repetition [34,37]. This understanding is fundamental for students, as academic success depends on it, and serves as a solid foundation for tackling more advanced concepts, thus highlighting the importance of integrating these tools into the programming curriculum while avoiding unethical patterns of behavior [11].

In terms of ease of learning complex topics, GenAI proves to be a valuable resource. The results indicate that students find learning difficult content more accessible compared to traditional methods. This finding suggests that GenAI acts as a support and transforms the way students interact with the material, promoting more dynamic and comprehensible professor-supervised learning. In addition, the integration of GenAI in the classroom seems to spark a greater interest in programming, which is a key motivating factor in the educational process [59].

Finally, when examining learning expectations, most students feel that the GenAI not only meets, but often exceeds, their expectations. This positive perception extends to the support they perceive they receive from these tools in their learning process. Satisfaction with the use of GenAI in teaching programming demonstrates the effectiveness of these technologies in improving both student motivation and engagement which, in turn, has a lasting impact on their academic and professional development in programming [23].

6. Conclusions

The results of this study conclusively show that the incorporation of generative artificial intelligence tools in the teaching of programming has a significant positive impact on student learning, motivation, and interest. The observed correlations suggest that the use of technologies such as Gemini in Google Colab not only facilitates the understanding of fundamental concepts, but also fosters greater satisfaction in the learning process. This finding highlights the importance of integrating GenAI as an essential component in programming curricula, which can transform the educational experience by making it more engaging and effective.

However, it is essential to recognize that the successful implementation of these tools requires adequate training for educators, ethical training for students and users, an open attitude towards their possibilities, and a curriculum design capable of taking full advantage of GenAI capabilities. By continuing to research and adjust innovative methodologies in course content, it is ensured that programming education not only remains relevant in an ever-evolving technological world, but also prepares students effectively to meet the challenges of the future.

The use of these tools has been found to significantly improve students’ understanding of fundamental programming concepts. This improvement translates into greater assimilation of key content which, in turn, establishes a solid foundation for learning more complex topics. Also, by analyzing how generative artificial intelligence facilitates the learning of difficult topics, it has been shown that students feel better able to tackle advanced content compared to traditional methods.

The integration of GenAI has also significantly influenced students’ interest and motivation towards programming, which has been reflected in their greater willingness to participate in related academic activities. Finally, it has been evidenced that students’ learning expectations are not only met, but in many cases exceeded, highlighting the added value that GenAI brings to the educational experience. Taken together, these findings emphasize the importance of incorporating advanced technologies into programming instruction to improve student engagement and satisfaction.

In conclusion, this work has made it possible to evaluate the impact of generative artificial intelligence in the teaching of programming, revealing that tools such as Gemini in Google Colab have positive effects on various aspects of learning. The authors’ intention, in addition to adding to the existing literature on the field of generative AI in education, is to open spaces for meaningful learning activities that enhance the classroom experience with the use of novel tools. In addition, future research focuses on exploring best practices for integrating these technologies ethically, and how they can be adapted to diverse populations of students with different professional profiles.