The Impact of Social Media on Sexual Self-Concept: A Qualitative Analysis of Instagram in Mexico

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Title: The Impact of Social Media on Sexual Self-Concept: A Quali-tative Analysis of Instagram in Mexico

The purpose of this manuscript was to evaluate examine how machine learning techniques can be used to explore women´s self-representation and detrimental effects on Instagram. This manuscript has several strengths, such as the use of new technology within the study methodology and the relevance of the topic, which is highly pertinent in current discussions.

However, there are a few revisions to consider for improvement, as outlined below:

Introduction

Overall, the introduction is well-organized and focused and the references are current.

To clarify the objective, you pose two research questions, but could you please clarify what you mean by "detrimental effects"?

Method

How are the detrimental effects evaluated? Could you provide more clarity on this in the methodology section?

Results

The graphs presented in the results section are excellent

However, I’m not sure if I’m missing information or if I haven’t fully understood it due to a lack of knowledge. How was the relationship between objectification and the detrimental consequences analyzed in the evaluated population?

Dicussion

In the introduction, you refer to gender stereotypes. It would be interesting to analyze the results in relation to the theory of objectification and the gender perspective. Could you mention this in the discussion? Do the results confirm this theory?

Additionally, I noticed that the limitations of your study have not been addressed. Could you include them at the end of the discussion?

 

Author Response

Comments1: Overall, the introduction is well-organized and focused and the references are current. To clarify the objective, you pose two research questions, but could you please clarify what you mean by "detrimental effects"?

Response1:

We appreciate the reviewer's positive feedback on the organization and relevance of our introduction. Regarding the clarification of "detrimental effects," we have added the following lines on the Introduction section with the aim to clarify the meaning of "detrimental effects":

"In this research, we are considering as detrimental effects to the negative consequences that hypersexualized self-representation on Instagram may have on women and society. These effects include increased body dissatisfaction, reinforcement of harmful gender stereotypes, self-objectification, and potential mental health risks, such as anxiety and low self-esteem. Additionally, our study considers how algorithmic curation may amplify these detrimental effects by promoting content that aligns with certain beauty standards and sexualized imagery."

Similarly, we have expanded our research questions with the aim to specify our main contribution:

RQ1: It is possible to identify women´s self-representation and detrimental effects on Instagram images using the Google Vision API in Mexico?

• RQ2: Is there a relationship between racy content, body parts, and clothing?
  • 1: Is there a significant association between the presence of specific body parts in images and the likelihood of racy content classification by Google Vision API?
  • 2: How does the type of clothing influence the classification of images as racy content?

 

Comments2: How are the detrimental effects evaluated? Could you provide more clarity on this in the methodology section?

Response2:

We appreciate the reviewer’s request for greater clarity regarding how detrimental effects are evaluated in our study. To enhance transparency, we have revised the methodology section to explicitly state the scope and limitations of our approach in evaluating detrimental effects.

Therefore we have included the following lines on section 2. Materials and Methods to clarify this situation:

"It is important to note that our study does not directly measure psychological or social consequences but rather identifies patterns of self-representation via the Google Vision API and classifies images through its SafeSearch mode."

 

Comments3: The graphs presented in the results section are excellent. However, I’m not sure if I’m missing information or if I haven’t fully understood it due to a lack of knowledge. How was the relationship between objectification and the detrimental consequences analyzed in the evaluated population?

Response3: 

We appreciate the reviewer’s positive feedback on the graphs presented in the results section and the request for further clarification regarding the analysis of the relationship between objectification and detrimental consequences in the evaluated population. Our study primarily focuses on detecting patterns of self-representation in Instagram images using machine learning techniques rather than directly measuring the psychological or social consequences of objectification.

We clarify this point in the results section by adding the following lines:

"In consequence, this classification aligns with the definition of hypersexualized content, which includes partial nudity, revealing clothing, and provocative poses. This suggests that the algorithm is sensitive to visual elements commonly associated with hypersexualization. It should be noted that our study identifies visual patterns associated with self-objectification but does not directly measure individual experiences, motivations, or psychological impacts."      

 

Comments4: In the introduction, you refer to gender stereotypes. It would be interesting to analyze the results in relation to the theory of objectification and the gender perspective. Could you mention this in the discussion? Do the results confirm this theory? Additionally, I noticed that the limitations of your study have not been addressed. Could you include them at the end of the discussion?

Response4: 

We thank the reviewer for this observation to enhance our manuscript. We have now added the following text into the discussion section to try to explain how our findings are related to Objectification Theory and Gender Perspective:

"Our results indicate that Instagram images classified as racy content by Google Vision API predominantly feature specific body parts, reinforcing the selective focus on female physical attributes. This reflects broader societal patterns where self-representation becomes shaped by external validation, particularly within algorithm-driven social media platforms that amplify content based on engagement metrics and making it more difficult for women to engage in self-representation without conforming to established gender norms. Such mechanisms contribute to the reinforcement of hypersexualized beauty standards, increasing the likelihood of self-objectification."

 

We appreciate the reviewer’s observation regarding the need to explicitly address the limitations of our study. In response, we have incorporated the following text in the Discussion section:

"This study presents some limitations that should be considered when interpreting the findings. First, data collection was constrained by Instagram’s platform restrictions, as only publicly available images were analyzed. Similarly, the Google Vision API's automated classification system presents inherent constraints. As a proprietary black-box model, its decision-making process remains opaque, preventing us from directly assessing potential biases in the labeling of racy or adult content. The algorithm’s classification is based purely on visual elements, which means contextual factors such as cultural norms, camera angles, lighting, and user intent are not accounted for, potentially leading to misclassification or oversimplification of self-representation practices. In addition, the absence of qualitative user perspectives further restricts the interpretation of self-representation motives, as this research does not capture women’s subjective experiences or motivations behind their content choices."

 

 

 

 

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have developed an application that can automatically detect racy images on Instagram using Google Vision API. They also said that this application reveals the relationship between women's clothes and their bodies. It is thought to be a very useful study. It would be beneficial if a few deficiencies were addressed.

1-It has been determined that there is a relationship between body parts and clothing in racy content detection. This part needs to be expanded further. Is this approach correct for every racy image?

2-It is understood that the organs such as waist and legs detected in the visuals examined in the study indicate the lower body parts of women in striking poses. The sociological and social harms of this situation should be emphasized.

3-Why do women living in Mexico share more sexually explicit images than women living outside of Mexico?

4-Why do women upload images to social media that conform to gender norms and male perspectives?

5-Should the discussion section of the study be expanded? A critical approach should be taken. If there are any limitations of the study, they should be mentioned.

 

Author Response

Comments1: 1-It has been determined that there is a relationship between body parts and clothing in racy content detection. This part needs to be expanded further. Is this approach correct for every racy image?

Response1:

We appreciate the reviewer’s insightful comment regarding the need to further expand the analysis of the relationship between body parts, clothing, and racy content detection. We acknowledge that while our findings suggest a strong correlation between these elements, it is important to clarify whether this approach applies universally to every image classified as racy.

We have expanded our discussion section and we have added the following lines:

"Thus, the most frequently identified body parts by the Google Vision API were the thigh (803 occurrences), waist (781), leg (685), and shoulder (674). These body parts were predominantly exposed in provocative postures, reinforcing their correlation with racy content detection. The findings suggest that Instagram posts highlighting certain body parts are more likely to be classified as racy by the algorithm. Furthermore, clothing plays a crucial role in determining whether an image is classified as racy content. The Google Vision API identified a high correlation between revealing clothing items such as swimwear, shorts, skirts, dresses, and brassieres and racy classification. Statistical analysis showed that images featuring these types of attire had a higher likelihood of being labeled as racy. This suggests that clothing choices significantly influence the perception and algorithmic classification of images as sexually suggestive or explicit.

It is important to highlight that not every racy image necessarily follows this pattern. While our results indicate a strong statistical association, there are exceptions where racy content is identified without explicit visibility of revealing clothing or specific body parts. This could be influenced by factors such as pose, lighting, background context, or algorithmic biases in Google Vision API’s detection process."

 

Comments2: 2-It is understood that the organs such as waist and legs detected in the visuals examined in the study indicate the lower body parts of women in striking poses. The sociological and social harms of this situation should be emphasized.

Response2:

We sincerely appreciate the reviewer’s suggestion to further emphasize the sociological and social harms associated with the visual patterns identified in our study. In response, we have expanded the discussion to better contextualize the implications of hypersexualized self-representation on social media platforms such as Instagram.

We have expanded the discussion section to explicitly address the sociological and psychological risks associated with hypersexualized digital self-representation:

"Our findings indicate that images frequently classified as racy by the Google Vision API tend to highlight lower body parts, such as thighs, waist, and legs, often in provocative or suggestive poses. In this regard, the repetition and normalization of such imagery can contribute to the hypersexualized portrayal of women on social media fosters self-objectification, where social validation becomes tied to conforming beauty standards. This exposure can negatively impact mental health, leading to body dissatisfaction, anxiety, and low self-esteem. Additionally, it reinforces harmful gender dynamics, normalizing unrealistic beauty ideals and increasing the risk of online harassment and discrimination."

 

Comments3: 3-Why do women living in Mexico share more sexually explicit images than women living outside of Mexico?

Response3:

We appreciate the reviewer’s question regarding why women in Mexico might share more sexually explicit images compared to women in other countries. However, we would like to clarify that our study does not include a comparative analysis between Mexico and other countries, nor do we claim that women in Mexico share more hypersexualized images than those in other regions.

Our study focuses exclusively on Instagram users in Mexico, analyzing self-representation patterns through the Google Vision API. The findings highlight trends within Mexico but do not provide a basis for cross-country comparisons. A broader, multi-country study would be required to make such claims, incorporating diverse cultural, regulatory, and algorithmic factors that influence digital self-presentation.

To avoid any misinterpretation, we have revised section 3.4 to clearly state that our findings are limited to the Mexican context and should not be generalized beyond this scope. 

 

Comments4: 4-Why do women upload images to social media that conform to gender norms and male perspectives?

Response4:

We appreciate the reviewer’s question regarding why women upload images that conform to gender norms and male perspectives. Our study, as discussed in the Introduction and Discussion sections, highlights several factors that contribute to this phenomenon. 

Social and Cultural Expectations: The paper discusses how gender stereotypes are deeply embedded in society, shaping women’s self-representation on social media. Traditional femininity standards emphasize attractiveness, delicacy, and conformity to the male gaze, influencing how women present themselves online (Haines et al., 2016; Baek & Choo, 2018).

Engagement and Validation: Instagram’s platform dynamics encourage content that attracts likes, followers, and engagement, reinforcing the use of hypersexualized imagery. Studies cited in the paper indicate that women who share such content often receive more interactions, reinforcing the idea that sexualized self-representation is a path to visibility and social approval (Sheldon & Bryant, 2016; Drenten et al., 2020).

Algorithmic Influence: The platform’s algorithmic curation plays a role in amplifying certain content types, particularly those that align with popular engagement metrics. The study suggests that Instagram’s visual culture and algorithm-driven exposure contribute to the reproduction of hypersexualized imagery, as users adapt to the type of content that gains more visibility (Guizzo et al., 2021).

In this context we have added the following lines in the discussion section with the aim to clarify your observation:

"This study does not claim that all women intentionally conform to male perspectives but rather that sociocultural norms, engagement incentives, and algorithmic amplification create a digital environment where hypersexualized self-representation is prevalent."

 

Comments5: 5-Should the discussion section of the study be expanded? A critical approach should be taken. If there are any limitations of the study, they should be mentioned.

Response5:

We appreciate the reviewer’s suggestion to incorporate a more critical approach by explicitly addressing the limitations of our study. While our discussion section already has been expanded (text in red color), we recognize the importance of acknowledging technical, methodological, and contextual constraints. In this context, we have added the following lines to indicate some limitations:

"This study presents some limitations that should be considered when interpreting the findings. First, data collection was constrained by Instagram’s platform restrictions, as only publicly available images were analyzed. Similarly, the Google Vision API's automated classification system presents inherent constraints. As a proprietary black-box model, its decision-making process remains opaque, preventing us from directly assessing potential biases in the labeling of racy or adult content. The algorithm’s classification is based purely on visual elements, which means contextual factors such as cultural norms, camera angles, lighting, and user intent are not accounted for, potentially leading to misclassification or oversimplification of self-representation practices. In addition, the absence of qualitative user perspectives further restricts the interpretation of self-representation motives, as this research does not capture women’s subjective experiences or motivations behind their content choices."

 

 

 

 

 

 

 

 

 

 

Reviewer 3 Report

Comments and Suggestions for Authors

 

The paper presents an study to explore women's self-representation measured through the portrayal of their bodies using Instagram images in the region of Mexico. Although it may be interesting, the contribution of the work is limited, it is a very shallow analysis of an image sample. It lacks of rigurosity in the analysis. Here are some comments to improve the paper.

The research questions, particularly, RQ2 are too broad. I think that RQ2 can be decompose in sub-questions and demonstrate each hypothesis separately.

The labeling of images in Section 3.1 is completely unclear. It should be clarified whether the labels are obtained from Google Vision and how the probabilities of Table 2 are obtained.

The answer to RQ1 “We may respond to Q1 and indicate that it is highly likely to detect self representation with hypersexualized content on Instagram using the Google Vision API across Mexico.” is completely unsupported. How authors reach such conclusion? The fact that Google Vision can detect some labels do not related with the hypothesis.

The answer to RQ2 seems to be also obtained from observing some plots instead of a rigorous hypothesis test. The conclusion “We identify a strong association between clothing and body parts in these images, suggesting that women often share photos in revealing attire.” is too weak;y supported by the results.

The analysis of the regions of Mexico images can from lacks of purpose and also conclusions are vague and quite unsupported.

Author Response

Comments1: The paper presents an study to explore women's self-representation measured through the portrayal of their bodies using Instagram images in the region of Mexico. Although it may be interesting, the contribution of the work is limited, it is a very shallow analysis of an image sample. It lacks of rigurosity in the analysis. Here are some comments to improve the paper. The research questions, particularly, RQ2 are too broad. I think that RQ2 can be decompose in sub-questions and demonstrate each hypothesis separately.

Response1: 

We appreciate the reviewer’s insightful comment regarding the breadth of RQ2. We recognize that breaking down this research question into sub-questions can provide a clearer structure for our analysis and enhance the demonstration of our hypotheses.

In response, we have refined RQ2 in the paper as follows:

RQ2.1: Is there a significant association between the presence of specific body parts in images and the likelihood of racy content classification by Google Vision API?

RQ2.2: How does the type of clothing influence the classification of images as racy content?

Moreover, we have included in the discussion section the following lines to intend to respond your observation: 

"...Thus, the most frequently identified body parts by the Google Vision API were the thigh (803 occurrences), waist (781), leg (685), and shoulder (674). These body parts were predominantly exposed in provocative postures, reinforcing their correlation with racy content detection. The findings suggest that Instagram posts highlighting certain body parts are more likely to be classified as racy by the algorithm. Furthermore, clothing plays a crucial role in determining whether an image is classified as racy content. The Google Vision API identified a high correlation between revealing clothing items such as swimwear, shorts, skirts, dresses, and brassieres and racy classification. Statistical analysis showed that images featuring these types of attire had a higher likelihood of being labeled as racy. This suggests that clothing choices significantly influence the perception and algorithmic classification of images as sexually suggestive or explicit."

 

Comments2: The labeling of images in Section 3.1 is completely unclear. It should be clarified whether the labels are obtained from Google Vision and how the probabilities of Table 2 are obtained.

Response2: 

We appreciate the reviewer’s observation regarding the clarity of image labeling in Section 3.1. To address this concern, we have revised the section to explicitly state that the labels are obtained directly from the Google Vision API's SafeSearch mode (section 1.2), which categorizes images into five explicit content types: adult, racy, medical, spoof, and violence. In this regard, the Table 2 title has been changed.

 

Comments3: The answer to RQ1 “We may respond to Q1 and indicate that it is highly likely to detect self representation with hypersexualized content on Instagram using the Google Vision API across Mexico.” is completely unsupported. How authors reach such conclusion? The fact that Google Vision can detect some labels do not related with the hypothesis.

Response3: 

We appreciate the reviewer’s concern regarding the justification of our conclusion for RQ1. We acknowledge that our statement needed further clarification. To strengthen our conclusion, we have revised the discussion of RQ1 by explicitly connecting our findings with the research question. Specifically:

To strengthen our conclusion, we have revised the discussion of RQ1 by explicitly connecting our findings with the research question. Specifically:

1. Detection of Hypersexualized Content: Our study found that 81.6% of the analyzed images were classified as "racy" and 12.3% as "adult" by the Google Vision API (Table 2). This classification aligns with the definition of hypersexualized content, which includes partial nudity, revealing clothing, and provocative poses.

2. Correlation Between Image Elements and Classification: The analysis of body parts and clothing revealed that images containing exposed thighs, waists, legs, and shoulders were more frequently labeled as "racy." Similarly, images featuring swimwear, brassieres, skirts, and tight dresses showed a strong correlation with racy classification. This suggests that the algorithm is sensitive to visual elements commonly associated with hypersexualization.

We have added a summarized explanation on section 3 Results: 

"In consequence, this classification aligns with the definition of hypersexualized content, which includes partial nudity, revealing clothing, and provocative poses. This suggests that the algorithm is sensitive to visual elements commonly associated with hypersexualization."      

 

Comments4: The answer to RQ2 seems to be also obtained from observing some plots instead of a rigorous hypothesis test. The conclusion “We identify a strong association between clothing and body parts in these images, suggesting that women often share photos in revealing attire.” is too weak;y supported by the results.

Response4: 

We appreciate the reviewer’s concern regarding the methodological rigor in answering RQ2 and the need for stronger statistical support. We acknowledge that our original conclusion was based primarily on observational trends from plots, rather than a formal hypothesis testing framework.

To address this concern, we have taken the following steps:

We have explicitly described our statistical methodology in the revised manuscript. While the visualizations (e.g., Figures 5 and 6) highlight trends, we also conducted a quantitative analysis of the relationships between body parts, clothing items, and the classification of racy content. To ensure a more rigorous evaluation, we calculated the correlation coefficients between clothing items (e.g., swimwear, skirts, brassieres) and body parts (e.g., thighs, waists, shoulders) as identified by Google Vision. Our findings confirm a statistically significant relationship (correlation coefficients ranging from 0.5 to 0.97) between certain clothing types and body parts in images classified as racy.

We have revised our conclusion and we have added the following lines:

"The contribution of this study is to present quantitative evidence supporting a correlation between exposed body parts and revealing attire in hypersexualized self-representation." 

We also acknowledge the limitations of automated detection , therefore we added the following paraghap to the conclusion section: 

"... It is important to note that there is inherent limitations of automated detection methods, particularly those relying solely on machine learning models such as Google Vision API. While these algorithms are highly effective at identifying visual patterns and categorizing images, they lack contextual awareness and may not fully capture the nuances of self-representation, cultural variations, and individual intent. Therefore, a hybrid approach, combining computational techniques with human evaluation, would enhance the accuracy and interpretability of findings, these studies could provide deeper insights into the complex interplay between clothing, body representation, and social engagement on digital platforms. "

Comments5: The analysis of the regions of Mexico images can from lacks of purpose and also conclusions are vague and quite unsupported.

Response5: 

We appreciate the reviewer’s comments regarding the regional analysis of images from Mexico and the need for clearer purpose and stronger conclusions. To address this, we have refined section 3.4 to explicitly state why this analysis was conducted and how it contributes to answering our research questions. We have added the following lines to section 3.4 Georeferencing hypersexualized imagery in Mexico:

"Although the greatest concentration of hypersexualized images was found in central Mexico, this does not necessarily indicate causation. Rather, the apparent patterns may be influenced by socioeconomic conditions, cultural tendencies, and population density. It should be noted that the visibility of certain regional content may be biased by the dynamic changes in platform algorithms, which are beyond our sampling control."

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The paper has been improved regarding the previous versions, addressing some of my previous concerns. Although some results has been statistically justified, there is still some unsupported hypothesis. Even when there seems to be a relation among body parts and categories it would be important to clarify the relation with hypersexualization. Moreover, this require a stronger theoretical framework. Another issue that need to be discussed is that both body parts and categories are extracted from Google Vision, i.e. both come from the same source, the correlation can be encoded within the Google algorithm, not necessarily observed in real world. Finally, the importance of analyzing the different country regions is still unclear to me.

Author Response

Comments1: The paper has been improved regarding the previous versions, addressing some of my previous concerns. Although some results has been statistically justified, there is still some unsupported hypothesis. Even when there seems to be a relation among body parts and categories it would be important to clarify the relation with hypersexualization. Moreover, this require a stronger theoretical framework.

Response1:

We sincerely appreciate the reviewer’s acknowledgment of the improvements made in this revision and their valuable feedback on strengthening the relation between body parts, classification categories, and hypersexualization.

• Our results demonstrate a strong association between certain body parts (e.g., thighs, waists, and legs) and racy classification by the Google Vision API, with correlation coefficients ranging between 0.5 and 0.97.
• These findings align with previous research (Introduction section) on hypersexualized self-representation, which suggests that women’s lower body visibility in digital imagery is often framed within sexualized contexts to attract engagement (e.g., likes, comments, and followers).
• The concept of hypersexualization extends beyond individual clothing choices, encompassing how digital platforms amplify and reinforce specific beauty and desirability standards.
• Studies indicate that algorithmic curation on Instagram prioritizes sexually suggestive images, increasing their visibility and reinforcing self-objectification practices. Our findings suggest that the prevalence of certain body parts in racy-classified images aligns with these platform-driven dynamics.
• We have explicitly defined hypersexualization in relation to digital media, ensuring that our findings are clearly framed within this conceptual understanding.

Moreover, it should be noted that the research findings mentioned earlier suggest a link between women's lower body visibility in digital imagery and sexualized contexts. This means that when women's lower bodies are more visible in photos or videos, imagery is more likely to be presented in a way that emphasizes their sexual appeal. This can contribute to the hypersexualization of women in digital spaces, as it reinforces the idea that their value lies primarily in their physical appearance and sexual desirability.However, it is important to remember that hypersexualization is a complex issue with diverse perspectives and potential consequences. While some argue that it can empower or liberate women to express their sexuality, others highlight the potential for objectification, exploitation, and negative impacts on mental health and body image.

In this regard, we have highlighted in section 2.2 that:

"It is important to note that this study does not directly measure psychological or social consequences but rather identifies patterns of self-representation via the Google Vision API and classifies images through its SafeSearch mode."

Similarly in section 3.1 we have added the following lines to clarify this point:

"In consequence, this classification aligns with the definition of hypersexualized content, which includes partial nudity, revealing clothing, and provocative poses. This suggests that the algorithm is sensitive to visual elements commonly associated with hypersexualization. It should be noted that our study identifies visual patterns associated with self-objectification but does not directly measure individual experiences, motivations, or psychological impacts. "     

 

Comments2: Another issue that need to be discussed is that both body parts and categories are extracted from Google Vision, i.e. both come from the same source, the correlation can be encoded within the Google algorithm, not necessarily observed in real world.

Response2: We appreciate the reviewer’s critical observation regarding the potential limitations of our findings due to the use of Google Vision API for both body part detection and category classification. We acknowledge that the observed correlation between body parts and racy classification may be influenced by the algorithmic design of Google Vision rather than purely reflecting real-world user behavior.

Since both body part detection and racy/adult content classification are generated by the same Google Vision model, it is possible that the correlation between certain body parts and racy classification is a result of predefined patterns within the algorithm rather than an unbiased reflection of actual social media practices.

Computer vision models are trained on datasets that encode societal biases, meaning that pre-existing stereotypes about hypersexualization may be embedded in the model’s decision-making process.

The Google Vision API operates as a black-box model, meaning that its classification criteria are not fully transparent, making it difficult to determine whether it objectively detects hypersexualized content or simply replicates learned biases from its training data.

This limitation highlights the need for human validation or complementary qualitative analysis to confirm whether the algorithm’s classifications align with real-world user intent and cultural contexts. However this part is out of the scope of this paper.

To address these concerns, we have highlighted the Discussion section to explicitly indicate that:

"This research presents some limitations that should be considered when interpreting the findings. First, data collection was constrained by Instagram’s platform restrictions, as only publicly available images were analyzed. Similarly, the Google Vision API's automated classification system presents inherent constraints. As a proprietary black-box model, its decision-making process remains opaque, preventing us from directly assessing potential biases in the labeling of racy or adult content. The algorithm’s classification is based purely on visual elements, which means contextual factors such as cultural norms, camera angles, lighting, and user intent are not accounted for, potentially leading to misclassification or oversimplification of self-representation practices. In addition, the absence of qualitative user perspectives further restricts the interpretation of self-representation motives, as this research does not capture women’s subjective experiences or motivations behind their content choices."  

Similarly we highlighted on the Conclusions section the following lines:

"It is important to note that there are inherent limitations of automated detection methods, particularly those relying solely on machine learning models such as Google Vision API. While these algorithms are highly effective at identifying visual patterns and categorizing images, they lack contextual awareness and may not fully capture specific details of self-representation, cultural variations, and individual intent. Therefore, a hybrid approach, combining computational techniques with human evaluation, would enhance the accuracy and interpretability of findings, these studies could provide deeper insights into the complex interplay between clothing, body representation, and social engagement on digital platforms. "

 

Comments3: Finally, the importance of analyzing the different country regions is still unclear to me.

Response3:

We appreciate the reviewer’s request for further clarification regarding the importance of analyzing different regions within Mexico. In this regard, the inclusion of a geographical perspective in our study serves for:

Understanding the Digital Representation of Women Across Different Cultural and Social Contexts: Mexico is a socially and culturally diverse country, with regional variations in attitudes toward gender norms, media consumption, and digital self-representation. In this context we have added the folloing lines on Section 3.4:

"Our findings indicate that certain highly populated regions such as central Mexico display a higher prevalence of racy-classified content, which may be linked to urban digital culture and greater social media influence."

We sincerely appreciate this insightful feedback, as it has helped us improve the clarity and significance of this aspect of our research.