A Human–AI Compass for Sustainable Art Museums: Navigating Opportunities and Challenges in Operations, Collections Management, and Visitor Engagement

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper proposes a clear, human-centric approach to AI implementations that helps museums integrate AI across operations, collections management, and visitor engagement while keeping human critical judgment, ethical norms and sustainability at the core. Grounded in a broad literature synthesis and real-world cases, it balances enthusiasm with sober attention to organizational, technical, environmental, ethical, and operational risks. The framework is actionable and well-aligned with current GLAM needs.

In operations, the paper maps practical uses -  from audience forecasting and resource allocation to conservation, marketing, and curatorial workflows - delivering efficiency gains without reducing curatorial intent. In collections management, it makes a convincing case for automated metadata, improved search, and pattern recognition that modernize description, analysis, and interpretation while fulfilling stewardship and governance. For visitor engagement, it highlights chatbots, audio guides, personalization, recommendations, and co-creation as realistic paths to deeper, more inclusive experiences with clear ethical guardrails.

General recommendation: Accept with minor changes.

Single suggestion for improvement:

In the key resource of the article - Table 1: “Artificial intelligence in museums — advantages, challenges and impact on sustainability,” add a new column (name it for example: Tags, Keywords, or features or similar) to aid users in scanning, reading and navigating across such a large table. This will help readers quickly locate rows by aspect (e.g., “DL,” “metadata automation,” “personalization,” “LMM”). See TableModificationSuggestion.docx (attached) for example entries for the first four lines.

Comments for author File: Comments.pdf

Author Response

Thank you for your valuable suggestion. To help readers navigate the table more easily, I have added two new columns that provide specific keywords and features. These columns now detail the Museum Functions Enhanced and the AI Techniques/ Tools Used, which will help readers quickly locate information on topics like “Recourse Management” or "DL" "metadata automation," and "NLP".

 

Reviewer 2 Report

Comments and Suggestions for Authors

Here is my comment

- The manuscript lacks a clear definition of the Human–AI compass.
- The author structures the use of AI into three main thematic domains, each containing multiple case studies.
- Although Table 1 summarizes the sources, it feels too dense with information. I suggest dividing this into three separate tables—one for each theme—to summarize the case studies more clearly and make it easier for readers to follow.
- While Section 7 discusses the challenges of AI, I recommend adding a “benefits + challenges” sub-section for each thematic domain. This could be placed either within each thematic section or consolidated in Section 7.
- In Section 6, the author divides the content into two categories: “AI-powered Chatbots” and “Other AI-driven Visitor Experiences.” However, I suggest expanding this to four categories:
- AI-powered Chatbots & Virtual Assistants – Conversational tools for Q&A, multilingual support, personalized itineraries. Virtual assistants in museums could be discussed as an example.
- AI-based Recommendation & Personalization Systems – Systems that adapt exhibit suggestions, tour routes, and content based on visitor interests, demographics, or behavior patterns.
- AI-driven Immersive & Interactive Experiences – AR/VR/MR storytelling, adaptive narratives, location-based interactions.
- AI-enhanced Accessibility & Inclusion Tools – Real-time captioning, sign-language avatars, audio descriptions, and haptic/tactile feedback for visually impaired visitors
- Strengthen the methodological section by providing transparent inclusion/exclusion criteria and details of the search strategy.
- Provide a clear conceptual diagram of the Human–AI compass, showing its components, directional guidance, and relationship to sustainability pillars.
- even the author mention in ethical in section 7 but i suggest to include sub section more discussion in section discussion on ethical concerns related to AI—particularly in the case of Generative AI, since GenAI may produce fake data and AI hallucinations. How to tacking with AI how to prevent and protect how to implement. 

minor issue 
- Define all abbreviations (e.g., CV for computer vision, NLP for natural language processing) at first use.
- Some AI background explanations in early sections (e.g., general AI capabilities) are repeated in later sections—consider condensing to avoid duplication.
- When referencing figures or screenshots, provide a brief explanatory sentence in the main text to connect them to the argument, rather than leaving them as stand-alone visuals.
- Ensure that key terms (e.g., “Human–AI compass,” “sustainability pillars,” “Generative AI”) are spelled and capitalized consistently throughout the text.

Overall, this is a well-organized review on AI in museums, structured around three thematic domains—operations, collections, and visitor experience—supported by diverse case studies and a balanced discussion of opportunities and risks. However, the central Human–AI compass concept is insufficiently defined and lacks visual representation, the methodology section needs clearer inclusion/exclusion criteria and source evaluation, and several sections are overly descriptive with limited comparative analysis. Table 1 is too dense, the visitor experience section could be reorganized into four clearer categories, and benefits–challenges should be discussed for each theme. Ethical issues, especially around Generative AI, require deeper treatment. Strengthening these areas would enhance both the scholarly contribution and practical value of the paper.

Author Response

Comment 1: The manuscript lacks a clear definition of the Human–AI compass.

Response 1: Thank you for pointing this out. We agree with this comment. We have addressed this by adding a clear definition of the Human-AI Compass at the end of Section 7, "Navigating challenges with a human-AI compass" (lines 865-881). References to the compass have also been integrated throughout the Results-Discussion (lines 967-970) and Conclusions (lines 1017-1019) to strengthen the manuscript's argumentative power and overall coherence. We have also included a new conceptual diagram of the framework (Figure 12).

 

Comment 2:  Although Table 1 summarizes the sources, it feels too dense with information. I suggest dividing this into three separate tables—one for each theme—to summarize the case studies more clearly and make it easier for readers to follow.

Response 2: We have accordingly restructured Table 1, and, based on a suggestion from another reviewer, have added new columns with keywords to improve scannability. We also classified each entry by its thematic domain or domains (Operations Efficiency, Collection Management, and Visitor Experience) and identified the enhanced functions and risk categories.

While a strict separation of the original table into three distinct domain-specific tables wasn't practical—as many sources cover multiple domains, which would have required duplicating entries—we instead created separate benefit-focused tables for each domain (Tables 2-4), based on the processed data from the original Table 1 and the narrative of the corresponding thematic chapter. In a few instances, we also added new entries that were previously omitted but proved to be highly valuable.

Conversely, give the similarity and general nature of the risks referenced across the sources, we decided to consolidate these elements into a single, unified table (Table 5). This approach allows us to effectively present the benefits of AI for each domain while centralizing the challenges for the focused discussion in Section 7.

Although Table 1 is densely packed with information, it remains a crucial data repository organized by source. Its readability is facilitated by tags and keywords, which helps to identify information that leads to significant findings. By providing a simultaneous view of the source description and the emerging results, it thus maintains the transparency and verifiability of the entire research process.

The new tables are now in the Appendix, with their content summarized within each thematic chapter (lines 328-334, 424-429, 666-671) and at the beginning of Chapter 7 (677-685). We have also revised the manuscript's content to ensure conceptual unity and coherence, with references integrated throughout the Abstract (lines 30-35), Methods (206-210), Results-Discussion (lines 938-943), and Conclusions (lines 1000-1003). While the content of Chapter 7 was largely aligned with the new Table 5, we have now restructured it for greater clarity. Furthermore, the description of the Human-AI compass was placed at the end of this chapter to serve as the culminating response to the challenges discussed.

 

Comment 3:  While Section 7 discusses the challenges of AI, I recommend adding a “benefits + challenges” sub-section for each thematic domain. This could be placed either within each thematic section or consolidated in Section 7.

Response 3: Thank you for your valuable feedback. Please see our detailed response to Comment 2, which fully covers this point.

As a result of this suggestion, an extensive and meticulous in-depth analysis of the data from Table 1 was conducted through a multi-step process involving preliminary tables and diagrams. The goal was to identify general categories of museum operations enhanced by AI and specific functions that fall within or are served by them. This data was then correlated and cross-referenced with the presented case studies and applications from each thematic section to ensure all important information was included.

This rigorous process resulted in categories and subcategories of benefits for each thematic domain. The challenges, on the other hand, were selected to be presented collectively for all domains to provide a more holistic overview and for the sake of clarity and conciseness. These results are now visible and presented in Tables 2-5, with a summary description of their content included in the corresponding chapters (4-7).

 

Comment 4: In Section 6, the author divides the content into two categories: “AI-powered Chatbots” and “Other AI-driven Visitor Experiences.” However, I suggest expanding this to four categories: - AI-powered Chatbots & Virtual Assistants – Conversational tools for Q&A, multilingual support, personalized itineraries. Virtual assistants in museums could be discussed as an example.
- AI-based Recommendation & Personalization Systems – Systems that adapt exhibit suggestions, tour routes, and content based on visitor interests, demographics, or behavior patterns.
- AI-driven Immersive & Interactive Experiences – AR/VR/MR storytelling, adaptive narratives, location-based interactions.
- AI-enhanced Accessibility & Inclusion Tools – Real-time captioning, sign-language avatars, audio descriptions, and haptic/tactile feedback for visually impaired visitors

Response 4: Thank you for your valuable feedback. We have restructured Section 6, dividing it into the four categories you suggested. This has provided greater clarity and accuracy in our analysis of the case studies. This change also led to corresponding updates in other parts of the manuscript, including the Methods section (lines 195-198), Table 1 (entries 51-53) and the references.

 

Comment 5:  Strengthen the methodological section by providing transparent inclusion/exclusion criteria and details of the search strategy.

Response 5: This section already provides a detailed account of our search strategy and inclusion/exclusion criteria, including the specific keyword combinations, timeline, and justification for our chronological limits. We also previously addressed the supplementary role of non-peer-reviewed literature and the reliability criterion for its selection. However, to further enhance its clarity and address your comment, we have added a more detailed explanation of how the reliability of non-peer-reviewed sources was ensured (lines 160-167). In the same section, we have included descriptions of the new tables and elements that have been added to the paper, such as the expansion of the Visitor Experience categories from two to four and the addition of four new tables detailing the benefits and challenges of AI use in museums. This more granular approach strengthens our overall methodology by providing a clearer, more systematic basis for our analysis, transforming the qualitative review into a more structured, data-driven framework that enhances the transparency and rigor of our findings.

 

Comment 6:  Provide a clear conceptual diagram of the Human–AI compass, showing its components, directional guidance, and relationship to sustainability pillars.

Response 6: We agree with your suggestion and have addressed this in our revision. As part of our revisions, we've added a clear definition of the Human-AI Compass (see also Response 1) and included a new conceptual diagram of the framework (Figure 12). This diagram illustrates the model's structure and logic, showing how its components work: the compass needle represents AI, which serves as an essential tool for navigation, while the cardinal points—museum mission, values, and ethics—represent human intelligence (the user), which guides the direction. The compass's true north points to the ideal sustainable museum of the future, symbolically fusing the three pillars of sustainability within the cultural space.

 

Comment 7:  even the author mention in ethical in section 7 but i suggest to include sub section more discussion in section discussion on ethical concerns related to AI—particularly in the case of Generative AI, since GenAI may produce fake data and AI hallucinations. How to tacking with AI how to prevent and protect how to implement.

Response 7: The ethical challenges of AI are already presented in Table 5 and discussed in detail in Section 7 (lines 719-799), while Section 8 focuses on sustainability and the contribution of existing strategic approaches and frameworks. We believe that a dedicated discussion on ethical concerns here would complicate the chapter's structure and undermine its clarity. However, as part of our revisions, we have added a reference to our proposed conceptual model—the Human-AI compass—within Section 8 (lines 967-970). This model, which is based on addressing ethical challenges, serves to integrate and reinforce the study's contribution within a broader, global human-centered conceptual framework for action and self-governance in a complex, unregulated space.

Regarding the discussion on how to tackle GenAI issues, particularly fake data and AI hallucinations, we have the following points. While hallucinations are noted in Table 5 and further discussed in Chapter 7 as a technical challenge (lines 70-70), the issue of fake data was not identified within the literature reviewed for this study. We believe the specific nature of GenAI requires a dedicated approach and discussion that falls outside the scope of this study. Nevertheless, to address this valuable point, we have added a brief reference to the ethical and philosophical challenges of this type of AI (e.g., copyright, ownership of AI-generated content, responsibility, and human creativity) in the subsection on ethical challenges in Section 7. We have also included a reference to specialized literature that examines these issues in more depth (lines 72-72).

 

Comment 8: Minor issues: - Define all abbreviations (e.g., CV for computer vision, NLP for natural language processing) at first use.
- Some AI background explanations in early sections (e.g., general AI capabilities) are repeated in later sections—consider condensing to avoid duplication.
- When referencing figures or screenshots, provide a brief explanatory sentence in the main text to connect them to the argument, rather than leaving them as stand-alone visuals.
- Ensure that key terms (e.g., “Human–AI compass,” “sustainability pillars,” “Generative AI”) are spelled and capitalized consistently throughout the text.

Response 8: Thank you for these helpful suggestions. We have addressed these  points in the revised manuscript. We checked and corrected all abbreviations as needed, and we ensured that all figures and images are explained in the text and properly connected to the argument, having also added two images that were inadvertently omitted. All key terms have been checked for consistent spelling and capitalization throughout the text. Regarding the general background information on AI in the introductory section, we have reviewed the content and do not believe there is a repetition, as we feel this information provides a smooth and necessary foundation upon which the more specific details in later sections are built.

 

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for addressing all of my concerns. I believe that this paper could be beneficial to the museum field.