Deep Learning and Reinforcement Learning for Assessing and Enhancing Academic Performance in University Students: A Scoping Review

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

A scoping review was conducted on the use of AI-based strategies of DL and RL for assessing and enhancing academic performance. A literature overview of the newest empirical studies was carried out. However, there are still some issues that need to be addressed.

1. It is recommended to add specific descriptions of the research subjects and research scope to the abstract to enhance the integrity of the abstract content. In the results section, it only states that "Data were encouraging and promising", without mentioning specific data, key findings, or research outcomes. Although the conclusions section indicates that the strategies are effective and mentions the limitations of the research and future research directions, it is overly brief.

2. The authors are advised to clearly elaborate on the unique contributions and value of this review in the introduction section, highlighting its unique innovative aspects compared to other similar studies.

3. More research should be considered for related work. e.g. “Asymmetric light-aware progressive decoding network for RGB-thermal salient object detection”；“An Efficient Perceptual Video Compression Scheme Based on Deep Learning-Assisted Video Saliency and Just Noticeable Distortion”；“Adaptive Differentiation Siamese Fusion Network for Remote Sensing Change Detection”；“Full-Scale Feature Aggregation and Grouping Feature Reconstruction-Based UAV Image Target Detection” ；“Cfanet: Efficient detection of uav image based on cross-layer feature aggregation ”；“VSS-Net: Visual Semantic Self-Mining Network for Video Summarization”and “Multi-Scale Spatiotemporal Feature  Fusion Network for Video Saliency Prediction”. The concepts of these papers share similarities with your methodology.

4. The authors are recommended to provide detailed information on the training of judges during the literature screening process and the specific handling methods for divergent literature, so as to enhance the credibility of the research methods. In addition, searching only the Scopus database may lead to literature omissions, affecting the comprehensiveness of the research results. It is suggested to add searches of other relevant databases.

5. The listing of various studies is relatively detailed, but there is a lack of integration and in-depth analysis of the research results. The internal connections and common laws among different studies have not been fully explored, and there is a lack of systematic comparison and comprehensive evaluation of the research results, making it difficult for readers to quickly grasp the overall trends and key points of research in this field.

Author Response

Comments 1: [It is recommended to add specific descriptions of the research subjects and research scope to the abstract to enhance the integrity of the abstract content. In the results section, it only states that "Data were encouraging and promising", without mentioning specific data, key findings, or research outcomes. Although the conclusions section indicates that the strategies are effective and mentions the limitations of the research and future research directions, it is overly brief].

Response 1: [The abstract section has been improved]

 

Comments 2: [The authors are advised to clearly elaborate on the unique contributions and value of this review in the introduction section, highlighting its unique innovative aspects compared to other similar studies. More research should be considered for related work. e.g. “Asymmetric light-aware progressive decoding network for RGB-thermal salient object detection”；“An Efficient Perceptual Video Compression Scheme Based on Deep Learning-Assisted Video Saliency and Just Noticeable Distortion”；“Adaptive Differentiation Siamese Fusion Network for Remote Sensing Change Detection”；“Full-Scale Feature Aggregation and Grouping Feature Reconstruction-Based UAV Image Target Detection” ；“Cfanet: Efficient detection of uav image based on cross-layer feature aggregation ”；“VSS-Net: Visual Semantic Self-Mining Network for Video Summarization”and “Multi-Scale Spatiotemporal Feature  Fusion Network for Video Saliency Prediction”. The concepts of these papers share similarities with your methodology].

Response 2: [The “Related Studies” section has been introduced to emphasize the innovative contribution of the work].

 

Comments 3: [The authors are recommended to provide detailed information on the training of judges during the literature screening process and the specific handling methods for divergent literature, so as to enhance the credibility of the research methods].

Response 3: [Additional information on judge training and study screening and selection processes has been included].

 

Comments 4: [In addition, searching only the Scopus database may lead to literature omissions, affecting the comprehensiveness of the research results. It is suggested to add searches of other relevant databases].

Response 4: [To ensure rigor in identifying and selecting relevant studies, this review exclusively utilized the Scopus database for literature retrieval. Scopus was chosen due to its comprehensive coverage of high-quality, peer-reviewed journals and conference proceedings across multiple disciplines. It is widely recognized for indexing authoritative sources and maintaining rigorous inclusion criteria, ensuring that the selected studies meet high standards of academic integrity and relevance. Additionally, Scopus provides advanced search functionalities, enabling precise filtering based on keywords, study types, and publication periods, which aligns with the structured approach of this review.

While alternative databases such as Web of Science, IEEE Xplore, and Google Scholar also contain relevant literature, Scopus was deemed sufficient to achieve the research objectives due to its extensive indexing of AI, education, and cognitive science journals. Furthermore, cross-referencing within Scopus ensures that seminal and widely cited works are captured, mitigating the risk of literature omissions. Given these advantages, the use of Scopus as the primary source enhances the reliability, consistency, and replicability of this review's findings while maintaining a focused and methodologically sound approach].

 

Comments 5: [The listing of various studies is relatively detailed, but there is a lack of integration and in-depth analysis of the research results. The internal connections and common laws among different studies have not been fully explored, and there is a lack of systematic comparison and comprehensive evaluation of the research results, making it difficult for readers to quickly grasp the overall trends and key points of research in this field].

Response 5: [The internal connections between the various studies have been emphasized. A systematic comparison of the research results has been proposed, analyzing similarities and differences].

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript provides a comprehensive scoping review of the application of Deep Learning (DL) and Reinforcement Learning (RL) in evaluating and enhancing university students' academic performance. The study covers key areas such as predictive analytics for identifying dropout risks, personalized learning systems, intelligent tutoring applications, and language proficiency tools. While the manuscript is well-written and contributes meaningfully to the literature, I have a few concerns and suggestions for improvement:

1. The manuscript predominantly relies on specific datasets, which lack diversity in student demographics. This reliance raises concerns about the broad applicability of the findings, particularly across varied educational contexts and populations. Additionally, the manuscript does not sufficiently address non-STEM disciplines or holistic educational metrics, such as emotional intelligence and creativity. It is recommended that the authors consider how their findings might be generalized or adapted for diverse demographic and disciplinary contexts.

2. The review primarily focuses on short-term impacts of DL and RL applications, with limited exploration of their long-term effects on students' academic trajectories and personal development. Future iterations of the study could benefit from incorporating longitudinal research to evaluate sustained impacts and the development of transferable skills over time.

3. Critical issues related to data privacy, algorithmic bias, and transparency are insufficiently addressed in the manuscript. Given the increasing reliance on AI in educational settings, it is essential to prioritize ethical considerations. The authors are encouraged to discuss frameworks for ethical AI usage, emphasizing the importance of privacy, fairness, and transparency in deploying AI-driven solutions.

 

Author Response

Comments 1: [The manuscript predominantly relies on specific datasets, which lack diversity in student demographics. This reliance raises concerns about the broad applicability of the findings, particularly across varied educational contexts and populations. Additionally, the manuscript does not sufficiently address non-STEM disciplines or holistic educational metrics, such as emotional intelligence and creativity. It is recommended that the authors consider how their findings might be generalized or adapted for diverse demographic and disciplinary contexts].

Response 1: [The available studies are limited to specific geographic areas, which may impact the generalizability of our findings. As stated in Section 7, future research should include samples from diverse geographic, cultural, and linguistic backgrounds to assess the effectiveness of the proposed solutions in heterogeneous educational settings. As reported, future studies should expand the analysis to educational dimensions beyond STEM fields, such as emotional intelligence and creativity, as well as examine the long-term retention of the observed results. These aspects represent promising future research directions that could contribute to a broader and more applicable understanding of the adopted methodologies.].

 

Comments 2: [The review primarily focuses on short-term impacts of DL and RL applications, with limited exploration of their long-term effects on students' academic trajectories and personal development. Future iterations of the study could benefit from incorporating longitudinal research to evaluate sustained impacts and the development of transferable skills over time].

Response 2: [We agree with the reviewer's analysis. As described in the "Conclusion" section, longitudinal studies are needed to assess the impact of AI-based interventions on learning outcomes. While many studies focus on immediate improvements in performance and engagement, the long-term effects of these technologies on critical thinking, adaptability, and collaboration remain underexplored].

 

Comments 3: [Critical issues related to data privacy, algorithmic bias, and transparency are insufficiently addressed in the manuscript. Given the increasing reliance on AI in educational settings, it is essential to prioritize ethical considerations. The authors are encouraged to discuss frameworks for ethical AI usage, emphasizing the importance of privacy, fairness, and transparency in deploying AI-driven solutions].

Response 3: [The section “Ethical Considerations in AI-Based Education” has been introduced].

Reviewer 3 Report

Comments and Suggestions for Authors

The authors present a review evaluating the effect of deep learning (DL) and reinforcement learning (RL) strategies on assessing and enhancing academic performance in university education. They analyzed 28 papers published between 2014 and 2024, following PRISMA guidelines for systematic reviews.

 

The paper has different strengths:

• It synthesizes current research in an emerging and relevant field of educational technology.
• The review uses the structured methodology following PRISMA guidelines, which adds rigor to the review process.
• The included studies cover different AI applications across various educational contexts.

 

However, the paper has different limitations:

• While the abstract outlines the study's aims and methods, it lacks concrete results and conclusions, which are essential for readers to understand key findings and implications without reading the full text.
• Section 2, titled "theoretical framework," presents background literature rather than developing an actual theoretical foundation for the review.
• The search methodology described in section 3 needs refinement. In particular, the search expression requires explicit Boolean operators' precedence rules. The exclusion criteria would benefit from a nested hierarchical structure to better illustrate the relationship between different criteria.
• The results section presents a series of brief paper summaries that lack proper classification and analysis. The Riedel et al. paper on ChatGPT falls outside the review's scope, as it only tests ChatGPT's question-answering capabilities rather than examining its role in academic performance assessment or enhancement. The summary table (incorrectly labeled as a figure) needs revision. Its "authors" column is more like a paper citation column. The table fails to distinguish between papers focusing on performance assessment and those addressing performance enhancement. To improve the presentation of the results, I suggest the creation of a visual classification of the studied papers by using a hierarchical clustering diagram to show relationships between different studies and their approaches. This diagram would help readers better understand patterns in the literature and identify gaps in current research.
• The discussion section should synthesize findings across studies, identify patterns, and draw meaningful conclusions.
• The paper would benefit from a dedicated conclusion section synthesizing key findings and providing clear directions for future research.
• In addition, the paper requires careful proofreading to address missing spaces between words.

 

Given these limitations, I suggest reconsidering the paper after major revisions. While the paper addresses an important topic and demonstrates methodological understanding, the authors should focus on deepening their analysis, improving organization, maintaining consistent classification schemes, and developing meaningful conclusions. With these revisions, this work could contribute to understanding how AI technologies can improve university education.

Author Response

Comments1: [While the abstract outlines the study's aims and methods, it lacks concrete results and conclusions, which are essential for readers to understand key findings and implications without reading the full text.
Response1: [The results and conclusions have been further clarified in the abstract].

Comments 2: [Section 2, titled "theoretical framework," presents background literature rather than developing an actual theoretical foundation for the review].
Response 2: [The second section primarily reviews the background literature rather than developing a theoretical foundation. The purpose of the section is to provide an analysis of existing studies on DL and RL-based solutions, focusing on their applications and limitations. This analysis serves as a basis for the subsequent assessment of potential applications of these technologies in the university context. To better reflect the content and purpose of this section, it has been renamed «Background Literature». This revised title accurately conveys its role in synthesizing relevant studies and identifying gaps in the current research landscape. This approach ensures a logical progression from the literature review to the analysis of applications and implications in the following sections. This revision aligns the section with the overall objectives of the review, clarifying its scope and improving its coherence].

Comments 3: [The search methodology described in section 3 needs refinement. In particular, the search expression requires explicit Boolean operators' precedence rules. The exclusion criteria would benefit from a nested hierarchical structure to better illustrate the relationship between different criteria].
Response 3: [The search methodology has been refined to include explicit Boolean operators' precedence rules for clarity. The exclusion criteria have been reorganized into a hierarchical structure, illustrating relationships between criteria more effectively. These revisions address the review's rigor and transparency in study selection and exclusion processes].

Comments 4: [The results section presents a series of brief paper summaries that lack proper classification and analysis].
Response 4: [The results section has been revised to provide a more structured analysis, grouping studies into clear categories: assessment and enhancement of academic performance. Each category now includes a detailed synthesis of findings, highlighting methodologies, outcomes, and their implications. This approach ensures better classification and a deeper analytical perspective, addressing the comment regarding the need for improved structure and clarity].

Comments 5: [The Riedel et al. paper on ChatGPT falls outside the review's scope, as it only tests ChatGPT's question-answering capabilities rather than examining its role in academic performance assessment or enhancement].
Response 5: [The Riedel et al. paper has been removed from the analyzed studies, as its focus on testing ChatGPT’s question-answering capabilities does not align with the research question regarding the assessment and enhancement of academic performance through DL and RL solutions].

Comments 6: [The summary table (incorrectly labeled as a figure) needs revision. Its "authors" column is more like a paper citation column. The table fails to distinguish between papers focusing on performance assessment and those addressing performance enhancement].
Response 6: [The summary table has been revised to align with the reviewer's suggestions. The "authors" column has been changed from citations to the articles to the names of the authors of the studies. The table now includes distinct sections to clearly separate studies that focus on performance assessment from those that address performance improvement. This ensures a more accurate representation of the research reviewed].

Comments 7: [To improve the presentation of the results, I suggest the creation of a visual classification of the studied papers by using a hierarchical clustering diagram to show relationships between different studies and their approaches. This diagram would help readers better understand patterns in the literature and identify gaps in current research].
Response 7: [A hierarchical clustering diagram was added to illustrate relationships between studies and methodologies. This visual representation highlights patterns in the literature, aiding readers in identifying trends and research gaps, thereby enhancing the clarity and comprehensiveness of the presented results].

Comments 8: [The discussion section should synthesize findings across studies, identify patterns, and draw meaningful conclusions].
Response 8: [The discussion has been revised, in order to summarize the results of the studies, highlighting key patterns in DL and RL applications to assess and promote learning in the university setting. Finally, as suggested, significant conclusions have been drawn].

Comments 9: [The paper would benefit from a dedicated conclusion section synthesizing key findings and providing clear directions for future research].
Response 9: [Directions for future research have already been defined in the section dedicated to the analysis of the analyzed studies. The reviewer’s suggestion was positively received. A concluding section has been drafted, summarizing the main findings and providing directions for future research].

Comments 10: [In addition, the paper requires careful proofreading to address missing spaces between words].
Response 10: [The document has been reviewed].

Comments 11: [Given these limitations, I suggest reconsidering the paper after major revisions. While the paper addresses an important topic and demonstrates methodological understanding, the authors should focus on deepening their analysis, improving organization, maintaining consistent classification schemes, and developing meaningful conclusions. With these revisions, this work could contribute to understanding how AI technologies can improve university education].
Response 11: [The suggestions have been implemented. Changes based on the reviewer's comments are highlighted in green.].

Reviewer 4 Report

Comments and Suggestions for Authors

 

The paper presents a survey of works of deep learning and reinforcement learning as means to enhance academic learning. The survey might be useful but the paper requires substantial improvement. Here are some initial pointers.

The structured abstract can be change for a regular abstract summarizing the work contribution.

The introduction describes the challenges of SRL, but it fail to described the work presented in the paper and its motivation. It must summarize the motivation of the survey, the contribution and justify the scope. In fact, the reason why only Deep Learning (DL) and Reinforcement Learning (RL) are choose when many AI algorithms has been apply to this problem need to be clearly stated.

Instead of adding the URL of the protocol (https://osf.io/5ausc/view_only=2093e5f348d94abca8883bcf87d76b56) consider to include a brief description, so that the paper is self-contained.

The selection of the keywords is not justify. “University Students” is a very generic key-phrase when the scope of the survey is much more limited (only to SRL).

Most of the paper text is just an enumeration of works, without an in-depth analysis or comparison. The only comparison that can be made is in the table. The explanations of the works and methods should be interweave, describing the advantages and limitations of them in comparison with others. Otherwise, there is no real analysis and the summary results quite useless.

Author Response

Comments 1: [The structured abstract can be change for a regular abstract summarizing the work contribution].
Response 1: [The structure of the abstract has been improved and reworked].

Comments 2: [The introduction describes the challenges of SRL, but it fail to described the work presented in the paper and its motivation. It must summarize the motivation of the survey, the contribution and justify the scope. In fact, the reason why only Deep Learning (DL) and Reinforcement Learning (RL) are choose when many AI algorithms has been apply to this problem need to be clearly stated].
Response 2: [The introduction has been revised to clearly summarize the survey's motivation, contributions, and scope, justifying the focus on DL and RL].

Comments 3: [Instead of adding the URL of the protocol (https://osf.io/5ausc/view_only=2093e5f348d94abca8883bcf87d76b56) consider to include a brief description, so that the paper is self-contained].
Response 3: [The research protocol was described in the method section].

Comments 4: [The selection of the keywords is not justify].
Response 4: [The selection of the keywords was justified].

Comments 5: [“University Students” is a very generic key-phrase when the scope of the survey is much more limited (only to SRL)].
Response 5: [The scope of the investigation is the promotion and assessment of learning in university students through DL and RL based solutions. The scope of the research has been better specified].

Comments 6: [Most of the paper text is just an enumeration of works, without an in-depth analysis or comparison. The only comparison that can be made is in the table].
Response 6: [The results and discussion section has been improved, through in-depth comparisons of the models used and the results of the studies].

Comment 7: [The explanations of the works and methods should be interweave, describing the advantages and limitations of them in comparison with others. Otherwise, there is no real analysis and the summary results quite useless].
Response 7: [The suggestion has been implemented. Explanations of methods are now interwoven with their advantages and limitations, providing a cohesive and comparative analysis throughout the text. Changes based on the reviewer's comments are highlighted in light blue].

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have carefully revised their manuscript according to my comments and suggestions. I have no other problems.

Comments on the Quality of English Language

The English could be improved to more clearly express the research.

Author Response

Comments 1: [The authors have carefully revised their manuscript according to my comments and suggestions. I have no other problems].
Response 1: [We are grateful for the constructive feedback, which has been very helpful in improving the quality of our work].

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

 

Although improved, section 5 is still a enumeration of works, the text could be a little more fluid. Table 1 is too big to enable a comparative analysis, maybe if works in each subsection are compared separately with a smaller table at the end of the section (and such comparison explained) it would be more useful The dendogram need to be explained in the text, how the distances were calculated, they have some meaning? Many techniques are mentioned, but the discussion is not focus on which have resulted better or advantages/disadvantages.

 

Author Response

Comments 1: [Although improved, section 5 is still a enumeration of works, the text could be a little more fluid].
Response 1: [The text of Section 5 has been revised to make it more fluid].

Comments 2: [Table 1 is too big to enable a comparative analysis, maybe if works in each subsection are compared separately with a smaller table at the end of the section (and such comparison explained) it would be more useful].
Response 4: [Table 1 was split into two smaller tables, which were then compared].

Comments 3: [The dendrogram need to be explained in the text, how the distances were calculated, they have some meaning?]
Response 3: [The calculation and meaning of the distances present in the dendrogram have been better explained].

Comments 4: [Many techniques are mentioned, but the discussion is not focus on which have resulted better or advantages/disadvantages].
Response 4: [The discussion was enriched by highlighting the advantages and disadvantages of AI-based solutions].

Author Response File: Author Response.pdf