The Detachment of Function and the Return to Essence: Exploring the Public’s Emotional Attitudes Towards Gamified Education

Abstract

Gamified education, as an emerging educational model, is gradually transforming traditional learning methods and has sparked widespread public discussion about its effectiveness and potential. According to connectivism, thinking and learning occur through the connections and interactions among a large number of units. Gamified education can serve as a form of connection, facilitating learners’ links and knowledge construction across different units through interactions, tasks, and feedback. This study aims to explore the public’s emotional attitudes toward gamified education, particularly analyzing the phenomenon of detachment and return regarding its educational function and essence. Through sentiment analysis and LDA topic modeling, three main themes were identified: the gamification and effectiveness of language learning, programmatic learning under temporal and spatial flexibility, and interactive entertainment and social learning. The study found that the public’s emotional attitude toward gamified education is diverse, reflecting the recognition of its potential in providing flexible learning and enhancing interactive experiences, as well as concerns regarding the entertainment-focused nature of educational functions. Additionally, based on the conclusions drawn, the study offers recommendations for educators and designers of gamified education to address issues such as distraction and excessive entertainment during the promotion of gamified education, helping them gain a deeper understanding of its evolution and challenges.

1. Introduction

In recent years, the application and practice of gamification have received widespread attention, particularly with significant changes in public emotional attitudes. This shift has attracted considerable interest across various industries (Lampropoulos et al., 2022) as well as within academia (Loijas, 2024; Ourdas & Ponis, 2023). Some scholars have defined gamification as the use of game design elements in non-game contexts (Deterding et al., 2011). With the continuous development of information technology, the scope of gamification has expanded to multiple industries, such as education (Perez-Aranda et al., 2024), business (Sharma et al., 2024), and healthcare (Damaševičius et al., 2023).

In the field of education, the design, development, and application of educational games began to receive increasing attention in the 1990s. During the COVID-19 pandemic, global educational systems faced unprecedented challenges, and educational models underwent an unprecedented transformation. Online education and flipped classroom models quickly became the norm in higher education, serving as key measures to ensure the continuity of education during the pandemic (Nieto-Escamez & Roldán-Tapia, 2021). Gamification-based learning has gradually transformed traditional teaching methods and interaction modes, with digital teaching skills and student support becoming core focuses of online education (Deacon et al., 2023), enriching teaching methods and improving teaching effectiveness. This teaching model has not only been applied in traditional classrooms but has also rapidly expanded to online education (Khaldi et al., 2023), special education (Hussein et al., 2023; Ramos Aguiar et al., 2023), and higher education (Huseinović, 2024; Murillo-Zamorano et al., 2023), among other areas.

With the ongoing advancement of digital learning transformation, gamified learning in educational games is considered an important way to enhance digital learning skills (Barragán-Pulido et al., 2023). Connectionism theory posits that thinking and learning are achieved through the connections and interactions of numerous units (Hanson & Burr, 1990). Gamified education promotes the connection and knowledge construction between different units through interaction, tasks, and feedback. It encourages learners to develop knowledge and connect with sustainability concepts through digital facilitators, discussions, and social networks (Dziubaniuk et al., 2023). For example, students promote knowledge development through social networks, and the interactions and connections similar to neural networks gradually systematize the influence between elements (Duan et al., 2023). Some scholars suggest that incorporating gamified elements such as challenges, badges, points, and leaderboards into the teaching process can enhance students’ learning motivation and engagement (Mohd et al., 2023). For instance, using gamified platforms in programming courses can improve students’ academic performance, participation, and motivation (Abu-Hammad & Hamtini, 2023). The implementation of gamification varies significantly across different educational levels, and there is currently a lack of a systematic design process. Many gamified applications do not follow theoretical frameworks, and the effects of game elements fail to meet the expected outcomes (Kalogiannakis et al., 2021). Some scholars also suggest combining gamification techniques with learning strategies, particularly introducing augmented reality technology, which can enhance users’ sense of immersion and interactivity during the learning process (Haller & Ast, 2024). For example, combining scenario-based game learning with experiential learning strategies has a significant impact on improving medical students’ learning performance and six core competencies during training (Chang et al., 2024). However, some scholars have suggested that while gamification strategies have a positive effect on students’ learning motivation, in the short term, they enhance motivation through novelty effects and external rewards. However, as learning progresses, motivation may gradually diminish (Ratinho & Martins, 2023).

Gamification as an educational approach has gradually gained recognition in various educational practices, but its limitations must also be considered (Pasqualotto et al., 2023). Based on previous experiments, researchers have started exploring whether adjusting game elements can effectively enhance learners’ motivation and engagement. Adaptive gamification, as a personalized gamification method, aims to enhance learners’ motivation and sense of participation by dynamically adjusting game elements according to participants’ needs and personalities (Bennani et al., 2022). The impact of this adjustment is closely related to individual factors, including players’ personal characteristics and the initial motivation for the learning task (Machmud et al., 2023). Existing studies have shown that adaptive gamification assessment within a blended learning framework can significantly enhance learners’ motivation and learning outcomes, demonstrating its effectiveness as an educational approach (Zhang & Huang, 2024).

This phenomenon highlights the urgent need to explore public emotional responses to the divergence of gamified education from its educational essence and the movement toward its core values. The primary goal of this study is to examine public sentiment toward gamified education through sentiment analysis, investigating its advantages and disadvantages. This paper contributes to enriching and expanding research in the field of user information behavior and offers valuable insights for the development and adaptation of game-based education.

2. Research Design

2.1. Sentiment Recognition

Sentiment analysis (Nandwani & Verma, 2021) is a text analysis technique used to detect emotions within a text, aimed at understanding and identifying the emotional responses and underlying intentions of individuals or groups in specific contexts. The classifiers used in sentiment analysis can generally be divided into three categories: machine learning, deep learning, and ensemble learning (Tan et al., 2023). Machine learning trains data using mathematical models and algorithms to predict emotions, with common methods including logistic regression (Shobayo et al., 2024), Naive Bayes, and a Support Vector Machine (SVM) (Danyal et al., 2023). Deep learning predicts emotions by constructing multi-layer artificial neural networks, which can effectively capture complex features and nonlinear relationships. Common models include Recurrent Neural Networks (RNNs) (Durga & Godavarthi, 2023) and Long Short-Term Memory networks (LSTMs) (Mahadevaswamy & Swathi, 2023). Ensemble learning enhances sentiment analysis performance and accuracy by combining the outputs of multiple classifiers, effectively reducing the bias and variance of a single model, thereby improving predictive performance. Common methods include Random Forest (Alghazzawi et al., 2023) and Boosting (Mewada & Dewang, 2023). Sentiment analysis is a highly complex research topic, and as a result, researchers have invested significant effort and technology into various experiments, such as facial recognition (J. Lee et al., 2023), voice recognition (Hamsa et al., 2023), and body recognition (Blythe et al., 2023).

In this study, the VADER sentiment analysis tool was used. A sentiment score greater than or equal to 0.6 is classified as a positive sentiment, while a score less than or equal to 0.4 is classified as a negative sentiment. If the sentiment score falls between 0.4 and 0.6, it is considered a transition zone, indicating ambiguous or weak emotional expression, and is categorized as a neutral sentiment. Since VADER does not explicitly define thresholds for emotional classification, text-based sentiment analysis applications should be universally applicable, and their scoring system needs to maintain consistency. The VADER scoring system was mapped to the scoring systems of other common sentiment analysis tools to determine this emotional threshold, ensuring comparability for subsequent research. The following linear mapping formula was used for the conversion.

$$new\_score={\displaystyle \frac{vader\_score+1}{2}}$$

2.2. LDA Topic Modeling

Latent Dirichlet Allocation (Yu & Xiang, 2023) is an automated statistical modeling method widely used in large-scale text corpora to mine latent semantic structures. The topic selection process of the LDA model is based on a probabilistic generative mechanism. The model assumes that the topic distribution of each document is generated by a Dirichlet distribution. A topic is selected probabilistically, and then, based on the word probability distribution of the selected topic, a specific word is drawn. This process effectively captures the latent thematic structure within documents. Many scholars have used the LDA model for experimental analysis. For example, by applying topic modeling to a large body of the literature, they systematically extracted and identified the core research directions and development trends in the field (Kukreja, 2023, 2024).

We use the Latent Dirichlet Allocation (LDA) model, which employs a three-layer Bayesian probabilistic model to uncover latent thematic information in large-scale documents. Naive Bayes effectively overcomes the issue of insufficient vocabulary (Saura et al., 2022). The results of LDA can be refined into two types of probability distributions: the first is the document–topic distribution, which represents the probability weights of each topic in a document; the second is the topic–word distribution, which represents the importance weights of each word within a topic. These two distributions can be used not only for tasks such as document topic classification and similarity analysis but also for generating text summaries or supporting information retrieval systems.

Each document d samples a topic distribution from a Dirichlet distribution.

Θd~Dirichlet(α)

Each topic k samples a word distribution from a Dirichlet distribution.

ʊk~Dirichlet(β)

2.3. Research Platform Selection

Data was collected from social media platforms such as YouTube and Facebook, but issues such as significant topic drift were identified during the data collection process, leading to ineffective data and insignificant analysis results. After filtering, the Google Play Store and App Store were selected as the sources for obtaining review data. In 2024, the Google Play Store had over 2 billion monthly active users, with high levels of daily active users; meanwhile, since its launch, the App Store has become a major global platform for digital app distribution. Therefore, using these two platforms as data sources is highly persuasive and scientifically grounded. However, it must be acknowledged that these data may contain certain biases, such as differences in user experience caused by the types of devices used, platform preferences, and variations in user needs.

2.4. Research Framework

Firstly, the review data from the Google Play Store and the App Store is collected using Appbot (a tool for gathering user feedback). Secondly, the data undergoes preprocessing, including deduplication, cleaning, stopword removal, and tokenization. Then, sentiment is evaluated by combining a sentiment lexicon and rating rules to score the sentiment of the text. Sentiment is classified into three categories: positive sentiment, neutral sentiment, and negative sentiment, with the sentiment score used to determine the sentiment tendency. In order to delve deeper into the underlying meaning of the text, the TF-IDF feature extraction method is applied to the reviews to obtain feature words and their corresponding term frequency vectors. Then, LDA is used for topic modeling to uncover the topics within the text and determine the sentiment tendency of each topic. Finally, conclusions are drawn from the analysis of the topics and sentiment tendencies, as referenced in Figure 1.

3. Methodology

3.1. User Group and Sample Data

In the data collection process of this study, the core principle is to gather users’ genuine experiences, ensuring the authenticity and reliability of the data. The subjects of this study are users of gamified software, and the analysis of users’ emotional attitudes is based on reviews left by users after using the gamified software. Two currently popular gamified educational software, Kahoot and Duolingo, were selected for this study. Both share certain commonalities in the application of gamification strategies, as both enhance user motivation and engagement by increasing the fun and interactivity of learning. This is why they were chosen as the subjects for comment collection. During the data collection process, we strictly filtered the data based on the research topic to ensure a high degree of relevance to the theme. The comments primarily focused on users’ feedback and usage experiences, including aspects such as learning experience, interface design, learning outcomes, and suggestions for improvement. Below are several extracted data, As shown in

Table 1. Data presentation table.

Country	Time	Author	Comments
UK	27 November 2024	Santosh Khandare	This app is very good for learning English and many other languages.
USA	20 September 2024	AIRHORN99	wow! so fun! so amazing! just great!
UK	27 May 2023	_Kid_Dragon_	It gets hard
USA	21 January 2022	&@278gar	I love this game it really helps me
France	22 February 2022	Dilys6191616	Awesome app

.

3.2. Tools Used

This study primarily utilizes PyCharm2021 for all analyses, with the string library for string manipulation, the OS library for file path handling and operations, and the pandas library for data analysis and tabular data processing. The Langid Python library is used for language identification. For LDA topic modeling and sentiment analysis, Gensim is employed for LDA model construction and text vectorization, NLTK is used for text processing and sentiment analysis, pyLDAvis is used for visualizing LDA model results, and vaderSentiment focuses on the sentiment analysis of social media texts, as shown in Figure 2.

3.3. Program Details

3.3.1. Data Preprocessing

Data from January 2022 to November 2024 were initially collected from the two platforms of the Google Play Store and App Store, totaling approximately 30,000 user reviews. To enhance the validity of the data, multiple rounds of screening and comparison were conducted to ensure that the data on which the final analysis was based had high representativeness and accuracy. After the initial data collection, the data were denoised and preprocessed. The process of denoising mainly includes removing advertising content, insulting language and non-English text, etc. During this process, a program written in Python 3.8.0 was used, and non-English texts were eliminated through the langid library using machine learning algorithms. Preprocessing mainly utilizes functions such as duplicate marking and replacement in Excel to conduct a preliminary cleaning of the dataset, removing irrelevant information such as null values, duplications, emojis, and web pages. Then, approximately 24,600 valid data were ultimately retained. We wrote a program in python to remove stop words. Simply put, it uses the NLTK word segmentation tool to remove irrelevant words (including common function words, personal pronouns, possessive pronouns, high-frequency prepositions, conjunctions, abbreviations and affixes, numbers and symbols, etc.) to improve data accuracy and quality.

3.3.2. Comment Sentiment Analysis

Due to the large amount of data in this study, sentiment analysis was conducted by using programming methods. First of all, a program was written in Python to extract the emotional words in the comments. The comment data was imported through the pandas library, and the emotion-related words were extracted using functions and sentiment analysis was conducted. Then, a program was written to extract and count the keywords in the comments. The TF-IDF method (a statistical method used to evaluate the importance of words in a text) was employed to extract keywords. After calculating the features through a tool that converts text data into digital features, the keywords in the comments were extracted using a function, and their occurrence frequencies were counted. Finally, the Vader library was used to conduct sentiment analysis on all comments, and the sentiment score of each comment and the overall sentiment attitude were obtained.

3.3.3. Potential Topic Analysis

The potential topic exploration in this study utilized the Latent Dirichlet Allocation (LDA) model. The primary module from the gensim library (used for natural language processing tasks) was imported, which provides core functionalities and algorithms for loading and training topic models. In the preliminary stage, multiple tests were conducted to evaluate the coherence scores for different numbers of topics, in order to assess the coherence measure and determine the optimal number of topics. The experimental results show that the number of topics was analyzed from 2 to 6. When the number of topics was set to 3, the coherence score fluctuated around 0.63, indicating good model performance with low topic overlap. Setting the number of topics to 3 provided higher stability and interpretability. Text data was processed using the core module of the gensim library and the gensim.corpora submodule to create a dictionary and train the topic model. In the main code of the LDA model, as shown in Figure 3, the document dataset was input, the number of topics was set to 3, a dictionary mapping of words to indices was specified, the corpus was iterated 10 times, and the model was trained using two threads in parallel.

3.3.4. Theme Emotion

In this study, the Dirichlet Topic Model (LDA) was used to identify three potential topics, and then VADER was used to perform sentiment analysis on these topics. First, we filter out relevant comments based on the keywords of each topic and assign them corresponding topics; then, initialize the VADER sentiment analyzer and perform sentiment analysis on the comments of each topic. The specific method is to create a dictionary to store the information of each topic, comments, and perform sentiment analysis by topic classification; finally, we display the results visually. Figure 4 is the core code for filtering relevant comments and assigning topics.

4. Research Results

4.1. Sentiment Analysis

4.1.1. Emotional Words and Keyword Analysis

In the initial sentiment lexicon statistics of the data, positive words (21,060) were significantly more numerous than negative words (7573). However, after data processing, the number of positive words decreased to 654, while negative words increased to 720. The following

Table 2. Emotional words table.

Num	Word	Frequency
1	fun	2969
2	love	2314
3	best	1239
4	easy	1072
5	help	814
6	free	786
7	better	542
8	recommend	474
9	hearts	409
10	friends	369

consolidates the positive and negative words along with their frequencies. As shown in Table 2, entries 2.9, and 3.7, in particular, show high frequencies of occurrence and share similar meanings in certain contexts. The results indicate that the initial positive lexicon contained a large number of synonyms or repetitive expressions, which were significantly reduced after deduplication. This suggests that positive emotional expressions tend to be simplified and standardized, typically conveyed through fixed ways of expressing emotions or opinions. In contrast, the deduplicated negative lexicon exhibits greater diversity in expression, becoming more nuanced and varied.

The analysis of keywords indicates that the initial extraction yielded 15,444 keywords, which belong to a medium-sized dataset. After removing data with a frequency of two or less, 4150 valid keywords were obtained, ensuring a higher level of data validity.

Table 3. Keyword table.

Num	Word	Frequency
1	app	5972
2	learn	2917
3	fun	2750
4	language	2359
5	love	2122
6	game	1449
7	best	1153
8	new	1131
9	languages	1015
10	easy	1010

below displays the top 10 keywords and their frequencies, sorted in descending order by frequency.

In the word cloud, the font size of each word visually reflects its frequency of occurrence. The larger the font, the higher the frequency of the word, which helps to reveal the thematic tendencies of the comments and the focal points of the reviews. As shown in Figure 5, positive emotion words such as “fun,” “love,” and “best” reflect some users’ positive experiences and satisfaction with the use of the gamified educational software. On the other hand, negative emotion words like “wrong,” “annoying,” and “hard” highlight concerns, dissatisfaction, and frustration that users may have experienced during use. As shown in Figure 6, these keywords reveal that the content related to education, technology, and emotional experience dominates the dataset. Particularly in the learning process, the application of technology and emotional experience have become key factors in enhancing learning outcomes. The frequency of “app” is the highest, reflecting the trend in the field of education that increasingly relies on technological support and applications to promote learning. The emergence of the keywords “learn”, “language” and “languages” indicates that gamified learning still takes learning as its core goal. Meanwhile, the frequent appearance of “fun” and “love” highlights the crucial role of emotional factors in stimulating learning motivation.

4.1.2. Sentiment Score Analysis

The depth of the color directly reflects the average score of the emotion. The darker the color, the higher the average score of the emotion, which can better highlight the emotional tendency of the comment. As shown in Figure 7, the distribution of the average monthly emotional scores from 2022 to 2024 clearly indicates that the overall trend is towards positive emotions. Since 2023, the average emotional score has steadily fluctuated around 0.7, reaching its peak in June 2024 and showing its lowest point in December 2022. This indicates that as time goes by, the emotional scores of users gradually increase. The average value of the emotion score we obtained was 0.6957, indicating that most of the emotion scores belonged to positive emotions. The median was 0.7464, further confirming the concentration of the emotion distribution, indicating that most of the emotion data were concentrated in the higher score range and tended to be positive emotions. However, the standard deviation is 0.2174, indicating a certain degree of dispersion in the emotional scores. This suggests that in addition to positive emotions, the data contains some neutral and negative emotional scores, showing certain fluctuations. The phenomenon where the median is higher than the average suggests that the data has a skewed distribution, and the density of the higher emotion score interval is higher, while that of the low emotion score is lower.

4.1.3. Emotional Attitude Analysis

Users generally hold a relatively positive emotional tendency towards the topic or application under discussion. As shown in Figure 8 and Figure 9, there are a total of 17,193 comments with positive emotions, accounting for 69.7% of the total number of comments. There were 4955 comments holding a neutral sentiment, accounting for 20.1%. The number of comments holding negative emotions was 22,508, accounting for 10.1%. From December 2022 to January 2023, the fluctuations in comment data were particularly significant. This change might be related to the increase in the volume of research data collected at the time of collection, or it could be associated with the platform. The reasons for users’ emotional fluctuations may be rather complex. Among them, new features, bugs or performance issues brought about by application updates, policy changes such as adjustments to privacy policies and changes in payment mechanisms, as well as external social events and public opinions, may all become direct triggers for emotional fluctuations. In addition, overly frequent marketing activities and advertising pushes, or adjustments to the reward system, can bring in traffic while also potentially intensifying the accumulation of negative emotions. Overall, users’ emotional fluctuations are influenced by multiple factors.

4.2. Potential Theme Analysis

After analyzing the comment data through the LDA model, these three themes were obtained, mainly because they frequently appear and are representative in the comments, covering the diverse functions of the learning application and user needs. The key terms and comments of each topic demonstrate users’ demands and feedback for different functions. Therefore, these three topics can comprehensively present the core value and application scenarios of the learning application. The topics are as follows:

Topic 1: Programmatic learning under temporal and spatial flexibility. The key terms in the discussion included time, lessons, make, family, day, and work. It explored aspects such as daily learning, course arrangement, time management and planning, and family life, emphasizing the multiple roles and influences of learning applications in achieving a balance between learning and life. For example, the review from a Google Play user, “The family plan is a lot of fun,” and the review from an App Store user, “It’s a family and a class game so you better download it!” both reflect the value of learning applications in family and educational settings, highlighting their effectiveness and applicability in achieving a balance between learning and life. Comments such as “great little app to waste time at work under the guise of training” reflect that users may be skeptical about the actual educational value of the application, thinking that it lacks sufficient depth and instead wastes time. This indicates that although gamified education can enhance the balance between family and study life, its limitations in in-depth education may cause some users to question its effectiveness. Figure 10 is the most salient terms of Topic 1.

Topic 2: The fun and effectiveness of gamified language learning. The key terms in the discussion include fun, learn, easy, language, Spanish, makes, and use, emphasizing the learning of new languages in an interesting way and using interactive and game-like methods to make language learning more enjoyable. For example, reviews like “I learned more Spanish in a week than I did a different language that I took for 3 semesters in college.10 Stars.” reflect how the application may stimulate learning motivation and improve learning efficiency through more interactive and engaging methods. Other reviews, such as “When I do the voice thing, everything says it’s wrong,” and “It helps me so much, but it gets it wrong for spelling,” highlight issues with voice input and spelling. The theme reveals that language learning can become more interesting and efficient through interactive and gamified methods, and also reflects that there are still deficiencies in technical implementation. Figure 11 is the most salient terms of Topic 2.

Topic 3: Interactive entertainment and social learning. The key terms in the discussion include game, app, love, play, quiz, and friends. The discussion mainly focuses on gamified learning applications, interactive games and social applications, exploring their important roles in the user experience. Based on reviews such as “This game is so amazing!”, “Very good game,” and “The best game in the world,” users have provided highly positive feedback about the game. In an environment where gaming and learning are combined, it effectively enhances participants’ interest and motivation to learn. Reviews like “Duo texts me more than my own friends” and “This helped me to speak to my dad and some friends” reflect that gamified learning applications not only assist with language learning but also promote real-life communication among users. This theme demonstrates that by designing gamified educational applications that are both entertaining and social, learning motivation can be stimulated and social interaction can be promoted. Figure 12 is the most salient terms of Topic 3.

Figure 13 is a multi-dimensional visual view of the topic model, showing the overall topic model. Its visual interface consists of two main panels: The left panel shows the relative proximity between topics. The size of each circle indicates the importance and popularity of the topic, while the position reflects the degree of semantic separation of the topic. The right panel displays the key terms of each topic and presents the representative vocabulary of each topic through a bar chart. LDAvis emphasized the clear separation between topics. If each topic is significantly separated in space, it indicates that the topic has a unique semantic definition, the model structure is good, and it can effectively distinguish different topics.

4.3. Topic Sentiment Analysis

Through the latent topic analysis of LDA, under different topics and different learning experiences, the emotional responses and learning effects also vary due to their respective characteristics. The following is the thematic sentiment analysis, as shown in Figure 14.

Topic 1: Programmatic learning under temporal and spatial flexibility. The total number of comments related to the theme was 11,430, among which 2581 were positive emotions, accounting for approximately 22.6%. Neutral sentiment: There were 2903 items, accounting for approximately 25.4%. Negative emotions: There were 5946 items, accounting for approximately 52.0%. Compared to other themes, the proportion of negative emotions is relatively high. This may be due to the fact that programmatic learning, which progresses through fixed schedules and patterns, ensures the systematic nature of education but lacks personalized adjustments. As a result, it may affect learners’ motivation and emotions, with learners experiencing challenges or setbacks during the process, such as difficulty in learning or slow progress.

Topic 2: The fun and effectiveness of gamified language learning. The total number of comments related to the theme was 7502, among which 4090 were positive emotions, accounting for approximately 54.5%. There were 1758 neutral emotions, accounting for approximately 23.4%. There were 1654 negative emotions, accounting for approximately 22.1%. The gamification and effectiveness of language learning can enhance learners’ engagement and satisfaction to some extent. However, due to constraints such as learning outcomes, content depth, and technological implementation, some users’ learning experiences may not have met their expected goals.

Topic 3: Interactive entertainment and social learning. The total number of comments related to the theme was 5724, among which 3169 were positive emotions, accounting for approximately 55.4%. There were 978 neutral emotions, accounting for approximately 17.1%. There were 1577 negative emotion words, accounting for approximately 27.5%. Interactive entertainment and social learning offer a high level of engagement and entertainment, which can stimulate positive emotions. However, they also have certain technical barriers or design flaws, which may cause some learners to feel disappointed or uncomfortable, leading to negative emotions.

5. Discussion and Insights

This study utilizes user review data from both the Google and App Store platforms to explore the phenomena of the detachment and return of gamified education features, as well as the public’s emotional attitudes toward gamification in education. The insights gained from this research greatly expand our understanding of the complex interactions between educational methods, learning motivation, technological applications, and participants’ emotional responses.

5.1. Discussion Based on Topic Modeling and Sentiment Analysis

5.1.1. The Public’s Emotional Attitudes Toward Gamified Education Are Diverse

The public’s emotional attitudes toward gamified education exhibit a simplification of positive emotions and a nuanced diversity of negative emotions. Under the framework of connectionism, positive emotions can promote active interaction among learners in social networks, while negative emotions may lead to decreased engagement (Yang et al., 2024). The simple expression potential of positive emotions holds an optimistic attitude (Mee et al., 2021), significantly enhancing learners’ sense of participation and motivation, thereby fostering the achievement of learning outcomes. For example, applying an emotional design to scaffolding can enhance the educational value of gamified learning environments, increasing players’ situational interest and self-efficacy (Koskinen et al., 2023). Negative emotions can affect learners’ cognitive and emotional states, leading to long-term learning difficulties and psychological issues (Johnson, 2024). This emotional state can undermine motivation and concentration, thereby impacting learning outcomes. Therefore, learners tend to express themselves more delicately during psychologically sensitive periods to better understand information and alleviate learning pressure (Mohamed et al., 2023).

This emotional and expressive diversity not only reveals the multiple expectations people have towards gamified education but also reflects the contradictions and challenges it faces during its implementation (Tomé Klock et al., 2024). It is precisely the source of these emotional fluctuations that highlights the uncertainty and multifaceted impact of gamified education in practice.

5.1.2. Separation of Gamification Education Function

The social interaction and knowledge sharing emphasized in connectionism theory may be suppressed due to the interference of negative emotions (Chow & Mercado, 2020). Through the sentiment analysis of the theme “Programmatic learning under temporal and spatial flexibility”, it is found that the public holds more negative emotions toward gamified education. “Spatiotemporal flexibility” originally aims to provide flexible learning arrangements, but for learners who lack self-discipline or have weak time management skills, this flexibility can become an additional burden, increasing the risk of procrastination and slow learning progress (Asikainen & Katajavuori, 2023). Programmed learning ensures systematization through fixed progress but lacks personalized adjustments. This is particularly challenging for users with family and work responsibilities, as a fixed schedule is difficult to align with these responsibilities, thereby increasing stress and anxiety (Tan et al., 2024). Scholars have pointed out various challenges, such as increased workload, unfavorable learning environments, and poor communication (Gelles et al., 2020). Additionally, in interactive entertainment and social learning, some have suggested that the implementation of gamification in learning may decrease students’ motivation, affecting their attention, confidence, satisfaction, as well as community support, peer collaboration, and problem-solving abilities (Phung, 2020). At the same time, the excessive focus on entertainment and interactivity in gamified education can lead to distraction, preventing deep learning (Hadi Mogavi et al., 2022). For example, overly elaborate visual and sound designs, or frequent task rewards and interactive elements, may cause attention to shift from the learning content to the game mechanics (Gejandran & Abdullah, 2024).

In this context, gamified education seems to deviate from the core functions of traditional education, raising public concerns about the quality of education. If gamification leads to unsustainable learning motivation (Sofiadin, 2021), emotions such as anxiety and frustration may weaken learners’ engagement when they face setbacks or challenges, thereby hindering knowledge sharing and connections (Wang et al., 2024), while neglecting the depth and rigor of education (Dah et al., 2024). Additionally, the application of gamification in the classroom faces challenges such as negative competition, technical issues, a lack of resources, addiction risks, and mental health problems (Demirbilek et al., 2022).

5.1.3. Return to the Essence of Gamification Education

Learning is continuously formed and reconstructed through interactions and feedback between individuals, with connectionism theory emphasizing the dynamic construction of information flow and interactive networks in the learning process (Downes, 2022). Learning within the context of temporal and spatial flexibility provides learners with a decentralized and adaptable learning environment, allowing them to adjust learning time and pace according to their own needs (Rahardja et al., 2021). This adaptive mechanism enables gamified education to adjust based on students’ immediate needs, thereby allowing them to engage in personalized learning according to their individual learning styles (Zourmpakis et al., 2023). At the same time, in education applications aimed at language learning, there has been a further promotion of changes in students’ emotional attitudes and deeper cognitive construction. For example, the design of task-driven activities, reward mechanisms, and interactive contexts can effectively enhance learners’ knowledge construction abilities and sense of achievement (Zhan et al., 2024). Additionally, through the use of progress tracking, game competition mechanisms, and other features, the enjoyment and intrinsic motivation in the learning process are enhanced, thereby promoting improvements in learning outcomes (Chen & Huang, 2024). In interactive entertainment and social learning, by stimulating learning motivation, increasing engagement, and enhancing memory and comprehension, knowledge internalization is promoted, thereby effectively improving learning outcomes (Uz Bilgin & Gul, 2020). For example, when learners experience enjoyment and a sense of achievement in gamified education, they are more willing to share experiences and knowledge, which strengthens network connections, drives knowledge flow and innovation, increases learner engagement and motivation, and fosters cooperation and discussion (Lou & Xu, 2022).

In this context, gamified education returns to the essence of education. Gamified education provides a flexible and engaging experience for the learning process, but it still needs to integrate with the rigor, depth, and knowledge systems of traditional education (Lampropoulos & Sidiropoulos, 2024) in order to ensure the comprehensiveness, long-term effectiveness, and sustainability of knowledge internalization.

5.2. Research Contributions and Insights

Gamified education demonstrates immense innovative potential in future educational practices, offering more support for individualized learning and having a positive impact across a broader range of educational fields. The main contribution of this study lies in the following:

(1)

It expands the explanatory framework of connectionism theory. Connectionism emphasizes the influence of proximity and integrative factors, and the functions embodied by gamified education may be shaped under this trend. On the one hand, gamified education tends to focus on intuitive and simple teaching tasks, such as language and programming education; on the other hand, it may lead the public to focus on the entertainment value brought by the game itself, resulting in a deviation from its functional purpose. Therefore, the development of gamified education should pay focus on a moderate balance between the tool carriers and the educational subjects. The gamified tools themselves must closely align with the educational goals, while avoiding deviations between the game content and the teaching tasks.

(2)

It proposes the dynamic influence of emotions and the learning process. The article points out that the state of emotions may have an impact on the learning process and outcomes. Positive emotions can enhance learners’ concentration, cognitive processing ability and self-efficacy, thereby strengthening learning motivation and efficiency. This helps learners to enhance the internalization of knowledge, build a knowledge system, and ultimately optimize learning outcomes. Negative emotions have a negative effect on learners’ cognitive and emotional states during the learning process, thereby affecting the learning outcome. These negative emotions can interfere with learners’ concentration and information processing ability, and reduce learning efficiency.

(3)

It has been clarified that the ultimate goal of gamified education is to return to the essence of education. Gamification serves as a bridge between thinking and learning, and its core theme has always been education. Gamified education has undergone multiple evolutions and integrations. It is not only about integrating game elements into learning, but also about innovating in educational concepts and teaching methods. Through gamified design, education is transformed into a more flexible, interactive and decentralized learning environment, enabling students to explore themselves and construct knowledge through active participation. In such an environment, not only can the mastery of subject knowledge be enhanced, but also deeper exploration and development can be achieved in terms of emotions and cognition. The advantages of gamification have been brought into play, ultimately achieving a return to the essence of education.

6. Conclusions

This study, through the analysis of user review data from two platforms, Google and the App Store, reveals the public’s diverse emotional attitudes towards gamified education, points out its advantages in promoting learning motivation and social interaction, and also reflects the “dissociation” phenomenon that gamified education may lead to superficial educational content and entertaining educational processes. Despite the challenges, gamified education has significant potential in enhancing learning efficiency, interactivity and personalization, and can effectively return to the core goals of education. Therefore, we put forward the following suggestions.

(1)

Optimize the balance between entertainment elements and deep learning to enhance learning motivation and effectiveness. Gamification design should be closely integrated with the depth of disciplinary knowledge (Oliveira et al., 2023) to ensure that entertainment elements do not distract learners from the academic depth. To solve the problem of students’ limited attention, the interactivity and interest of learning can be enhanced by reducing cognitive load (Yuldasheva, 2025) and optimizing multimedia resources and multi-level tasks (M. Lee, 2023), thereby strengthening students’ concentration and learning motivation. Meanwhile, combined with the immediate feedback and reflection mechanism (Christopoulos & Mystakidis, 2023), educators can help students deepen the learning content in the game by setting up additional assessment and reflection sections after task completion, ensuring that the entertainment elements do not weaken the depth of knowledge mastery.

(2)

Introduce an adaptive learning system to enhance the flexibility of gamified education. Although gamified education has broken the temporal and spatial limitations of traditional education, the overly fixed and standardized curriculum design has restricted the individualized needs of learners. Adaptive learning adjusts the teaching content and learning path in real time according to the individual characteristics of learners, such as learning progress, interests and comprehension ability (Sikora et al., 2024). Through artificial intelligence and data analysis technologies, the system can dynamically optimize the learning experience (Song et al., 2024), monitor the learning status and progress of learners in real time (Aly, 2024), and provide personalized course recommendations and feedback, thereby achieving more efficient learning and meeting their individualized needs.

(3)

Pay attention to the relationship between emotional factors and learning motivation. To further enhance learners’ sense of achievement, challenge and social identity, educators can strengthen emotion-driven learning design and emotional support in gamification (Mou et al., 2024). These emotional factors can be enhanced through means such as motivational feedback (Fong & Schallert, 2023) and mobile social interaction (Zhao & McClure, 2024). This can also be achieved by improving technical means.

However, it must be admitted that there are some deficiencies in this study. The research results may only represent the emotional trends of some users, because there are significant differences among user groups on different platforms. The user characteristics, interest preferences, cultural backgrounds, etc., of each platform are different, and these factors may affect their emotional expressions and attitudes. Looking ahead, it is expected that through artificial intelligence and data analysis technologies, larger-scale data can be obtained for sentiment analysis, thereby enhancing the representativeness and accuracy of research. Personalized learning support in gamified education will become more precise and efficient (Rodrigues et al., 2021). Current research mainly relies on text analysis and fails to fully incorporate multiple research methods and technical means. Future research can draw on qualitative methods such as in-depth interviews and focus group discussions (Urhan, 2024). It could gain an in-depth understanding of the participants’ emotional responses, viewpoints, and the underlying background factors. Furthermore, with the development of technology, tools such as face recognition have gradually been used to analyze users’ emotional responses during the gamified learning process (Gupta et al., 2023), and to deeply explore the practical effects and multi-dimensional influences of gamified education. However, ethical issues have become an important challenge in its application. It is hoped that gamified education in the future will pay more attention to stimulating learners’ autonomous learning motivation and promoting in-depth understanding and long-term mastery of knowledge.