Harnessing Machine Learning to Unveil Emotional Responses to Hateful Content on Social Media

Round 1

Reviewer 1 Report

Authors claim that they use attack detection technologies, psycholinguistic feature extraction, and sentiment analysis algorithms to the issue of emotional responses to hateful content on social media. Their findings suggest that personal attacks and trolling exert a significant dampening effect on user engagement, exacerbating the polarization of the online environment. Overall, I consider that this paper has an interesting and important issue to some extents for the field of exploring social media problems; however, this paper still has some paper structure remarks. Here I pose some major and minor drawbacks, as follows.

1.      Really, in my carefully reviewing, this paper has some contributions for the related meaningful issue of measuring social media, but it is fully necessary that the Abstract part should be revised in a refined statement with the main methods used and focused on main results of research with empirical results. Now, it is not enough and unclear.

2.      The paper seems some messy with AI-assisted technological methods. It should be reorganized and restructured with a good order.

3.      The Section 2 is now unclear for the main issue of literature review; particularly, this section has cited 60 papers; thus, I suggest that the authors can add a table to completely and clearly outline various problems and the corresponding studies of them or the main issues and methods used. That is specifically useful and helpful for improving the readability of the paper.

4.      For Section 3: Contributions, I suggest that it can be moved to Conclusions part, and authors can add a traditional Methodology section to define the study methods used, and it is better to draw the research flowchart of this paper. It will be useful for improving the readability of the paper.

5.      For Section 5, it is also unclear now, please make a clarification for further addressing the data analysis flow and main methods used.

6.      Please check all literature format in References part and the text, they should be a consistent format for citations and should follow the style of this leading Computers journal; Now, it has some errors and exist different formats, and they should be revised carefully.

7.      Also, authors should make a more clear strength for their study when compared to past studies.

8.      Please also elaborate on adding a comprehensive clarification of the study for the research results and application-oriented values. 

9.      Please also explore the research limitations for the study.

10.  I also consider that it is lack there are few explanations of the rationale for this study. Please give the comments.

11.  More potential application discussions for practical applicable values on such a machine learning and social media issue would enhance the paper.

Minor editing of English language required.

Author Response

Special thanks to the reviewer for the time and effort spent to evaluate the article. We have revised and improved the article according to the reviewer's comments. Modifications are highlighted in blue in the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

In the article titled "Harnessing Machine Learning to Unveil Emotional Responses to Hateful Content on Social Media," the authors aim to investigate the impact of derogatory and harmful responses on user activity after such attacks. By gathering data from various sources and subjecting it to rigorous analysis, they identified a correlation between online attacks, such as hostile comments, and a decline in user engagement. Specifically, they observed an average decrease of 5% in user activity after 1-2 attacks, 15% after 3-5 attacks, and 25% after 6-10 attacks. These findings underscore the necessity for innovative approaches leveraging artificial intelligence and swift moderator intervention to curb the dissemination of hateful content.

Improvement Points

Some previous studies have already tackled the issue of user activity in response to hateful content on social media. Authors need to compare their study to previous studies and explain how their findings are better or are different from existing works.

n/a

Author Response

Special thanks to the reviewer for the time and effort spent to evaluate the article. We have revised and improved the article according to the reviewer's comments. Modifications are highlighted in blue in the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper aims to demonstrate the usefulness of machine learning in revealing emotional responses to hateful content on social media. The paper is interesting as it addresses a real and current issue in cybersecurity. Hate speech is a complex phenomenon that goes beyond mere words. It embodies prejudices, discrimination, and hostility, often targeting specific groups based on race, ethnicity, religion, gender, sexual orientation, or other protected characteristics. Analyzing hate speech involves navigating a range of legal, ethical, and technological considerations. Hate speech manifests through language, exploiting its power to harm, incite violence, and perpetuate stereotypes. Linguists study the vocabulary, syntax, and semantics of hate speech to identify patterns and distinguish it from legitimate discourse. In the fight against hate speech, AI emerges as a powerful ally, thanks to its machine learning algorithms enabling the rapid and accurate identification and reporting of harmful content. By analyzing large amounts of data, AI models can learn to recognize linguistic patterns and nuances associated with hate speech, enabling them to categorize and effectively respond to offensive content.

The paper is well-organized and enjoyable to read. A comprehensive literature review has been conducted. The methodology is straightforward. The results appear coherent and compelling. The figures and tables are clear and of good quality. However, I would like to offer some comments:

1. 1- The abstract seems too generic and does not reflect the specific content adequately.
2. 2- Separate the Results section from the Discussion section.
3. 3- Specify the limitations in the application of machine learning to understand emotional responses to hateful content.
4. 4- Specify how the results of this study could be used to improve content moderation on social media platforms.
5. 5- The conclusion is too generic. It should be strictly related to the content of the paper.

Author Response

Special thanks to the reviewer for the time and effort spent to evaluate the article. We have revised and improved the article according to the reviewer's comments. Modifications are highlighted in blue in the revised manuscript.

Author Response File: Author Response.pdf