A Systematic Review of Library Service Accesibility using Artificial Intelligence: Features, Challenges, and Library Roles

Libraries are information centers that play an important role in providing access to scientific resources for the wider community. However, accessibility in libraries is often a challenge for users with disabilities. According to an ILO report, in 2013 approximately 15% of the world's population (more than 1 billion people) were persons with disabilities (International Labour Organization, 2013). Meanwhile, according to a report by the Central Statistics Agency of Indonesia, of the total population aged 5 years and above, nearly 2 million people are people with disabilities (Directorate of Statistical Analysis and Development, 2023). Research in countries such as India, Kenya, the United States, Sweden, Japan, and South Africa shows limited access to resources, inadequate internet connectivity, and the need for special assistance from skilled library professionals (Onchagwa & Wambiri, 2022, pp. 120—121; Salauddin, 2022, p. 47; Yoon, 2020, p. 1). Efforts to address these issues have been made, including the implementation of disability policies in public libraries (Lee, 2022, p. 1) and the transformation of academic libraries to improve accessibility through the use of new technologies, website modifications, and inclusive content (Longmeier & Foster, 2022, p. 823). However, there is still room for improvement in making libraries more accessible to users with disabilities. This explains why library accessibility continues to remain one of the problems that libraries need to address. Artificial Intelligence (AI) technology is emerging as apotential solution to improve library accessibility in the digital age. Research shows that Al can enhance library activities, support decision making, and provide a personalized, in-depth, and accessible user experience (Amzat & Adewojo, 2023, pp. 17-18; Mahmud, 2024, p. 5; Subaveerapandiyan, 2024, pp. 1—2). The application of Al in libraries can result in operational efficiency, increased research visibility, quick access to information sources, space savings, maximization of staff efforts, and improved user experience (Ikenwe & Udem, 2022, pp. 12; 15; 18; 21; 24; 27). Al technologies such as chatbots and text-to-speech can also be leveraged to serve more users, including individuals wityh visual impairments (Manoharan et al., 2024, p. 2; Sanji et al., 2022, pp. 1—2). However, the integration of Al in libraries also faces challenges such as funding, librarians’ attitudes, technocal skills (Hussain, 2023, p. 15), as well as ethical considerations related to bias and discrimination in Al systems (Saeidnia, 2023, pp. 1—3; Smith, 2021). Although previous research has explored the potential and challenges of Al in library services in general, there are still significant research gaps regarding the specific use of Al to improve accessibility for users with disabilites. Accessibility in libraries is an important issue that encompasses various aspects, both physical and digital, to ensure that all users, including those with disabilities, can access library services and resources. The following discussion explains some of the research findings related to accessibility problems in libraries. Axel Schmetzke (2001, pp. 36—37, 2003, p. 145) emphasized the importance of accessible web design to ensure everyone, including those with print disabilities, could access web-based information. Research shows that many university library webpages do not meet accessibility standards, reflecting a lack of awareness about this issue among library professionals. Catherine J. Carter (2017, pp. 14—17) highlights that academic libraries should focus on three key areas to meet the needs of students with disabilities: bibliographic instruction, web page design, and staff training. Improving library services can provide benefits for all students by improving access to information and the quality of library services. Camilla Fulton (2012, pp. 34—37) discusses web accessibility in the context of United States federal law, such as the ADA and Section 508 of the Rehabilitation Act. The lack of awareness about accessibility issues among electronic journal service providers demonstrates the need for increased awareness and proactive action from librarians to ensure all web content is accessible to people with disabilities. Clayton A. Copeland (2011, pp. 226; 231; 235) explores library services from the perspective of users of different abilities. The study revealed that many libraries are not yet fully prepared to meet the needs of users with disabilities, pointing to the importance of staff training and policy improvements to improve accessibility. Sung Jae Park (2012, pp. 150—152) highlights physical accessibility issues in public libraries, including the spatial relationship between libraries and other facilities, travel distances, modes of transportation, and the temporal perspective of library access. The study recommends the concept of co-sponsorship with other facilities, the placement of material drop boxes on main routes and a lunchtime library programme to improve accesibility. The findings suggest that improving accessibility in libraries requires a holistic approach that includes web design, staff training, legal compliance, and active engagement with the user community. Given the potential and challenges that exist, as well as the existence of research gaps, this research aims to explore the use of Al technology in improving accessibility in libraries, with a focus on users with disabilities. The main objectives of this study are identifying key features of Al technology that works well with people suffering from disabilities using libraries, examining challenges related to incorporating tools meant for access into public libraries’, and recognizing how such institutions engage in offering training and assistance regarding Al technology use. This research aims to fill a gap in existing literature by exploring ways through which artificial intelligence can be effectively utilized in order to enhance library accessibility for users with disabilities. The results of this research are expected to contribute to the development of more effective Al systems, overcome implementation challenges, and increase the role of libraries in supporting the use of accessibility technology so that library accessibility for users with disabilities can be significantly improved. Thus, this research has important implications in realizing a more inclusive and accessible A Systematic Review of Service Accessibility with Artificial Intelligence 4 library for all users, regardless of their physical abilities or limitations. The following research questions will be the focus of this study: 1) What are the key features of Al technologies that can improve the accessibility of library services? 2) What are the challenges face in implementing Al technologies to improve their services? 3) What is the role of libraries in the development and the use of Al technologies?

This study used Systematic Literature Review (SLR) as research methodology (Hidayat et al., 2024, p. 37289). This approach seeks to find, assess, and summarize the findings of different studies concerning the application of Al technology-enhanced library accessibility for disabled persons. The first step of this research is to construct a PICOC (Population, Intervention, Comparison, Outcome, Context) framework that clarifies an explicit and effectual focus (Hidayat et al., 2024, pp. 37289—37290). Disabled individuals form the target population that was intervened upon through the application of artificial intelligence. There were no direct comparisons in this study, but it was expected that library services would become more accessible. This research’s framework entails a library service employing Al technology to support navigation and user interaction with disabilities. The procedure for searching the literature was conducted using ScienceDirect digital database and designed search strings including a variety of related terms such as "disability", "handicap”, "impairment", "artificial intelligence, "AI", and "machine learning”. This study only considered studies published between 2014 and 2024 written in English according to very strict criteria for inclusion and exclusion. This way, only the most recent and relevant literature is included in research works. Moreover, to maintain consistency and relevance of the findings, studies whose language is different or outside this range were excluded. Each selected study was subjected to quality assessment using a checklist that contained several key questions such as clarity of research objectives, methodology and whether the research questions were answered. This means that only top-notch studies will be included at the end. A data extraction form was created and used to gather information like the author’s name, year of publication, type of study, sample size, methods employed, objectives, characteristics, challenges faced, and the role of library. By using this approach researchers are able to have an overall view on how artificial intelligence can be used in libraries in order to make their services more accessible for persons with disabilities while at the same time noting down what might stand on their way and how they will need help from such institutions. The article selection process shown as PRISMA flowchart is depicted in Figure 1, which outlines how articles were selected. The diagram illustrates how the movement of information changes as a consequence of systematic review of literature. It also indicates how many records were identified, those that were entered but not included, and provides reasons for their exclusion (PRISMA Statement, 2020, pp. 1—2). o -§ ® ':U E & = %0 £ E o} g 1] £} 5 = g = Source: author's documentation Records removed before screening: Duplicate records removed (n = 0) . . Records identified from: ScienceDirect (n = 28) Records screened (n = 28) - Reports sought for retrieval (n=12) PP Relfl"m assessed for eligibility (n=11) v Studies included in review (n=10) Reports of included studies (n=10) > Records marked as ineligible by automation tools (n = 0) Records removed for other reasons (n=0) — | Records excluded (n = 16) . — | Reports not retrieved (n = 1) Reports excluded: ___, Not passing the eligibility assessment (n=1) Figure 1. PRISMA flowchart of the article identification and selection process

The selection and analysis process resulted in a total of ten articles. The selected studies are spread from 2015 to 2024 with an uneven distribution. While there was only one article per year for the years 2015, 2018, 2019, and 2021; the number increases from 2022 to 2024 (two, three, two articles per year, consecutively). The increase in the number of studies in the later years suggests a growing interest in the topic. This may reflect an increased urgency to adopt new technologies and improve accessibility in libraries. In terms of research location, there was a predominance of studies conducted in the United Stated. Six of the ten studies were located in the United States, indicating a strong focus on the US library context. One study was conducted in Pakistan, one in Indonesia, and one study covered both Canada and the United States. This geographical distribution suggests gaps in research, with limited representation from developing countries and other regions of the world. On the other hand, this result is understandable, due to the limitations of the database used as the source for this study. The temporal analysis shows the evolution of the research focus over time. Early studies (Yi, 2015) focused on compliance with accessibility standards in public libraries, research in subsequent years began to explore the use of advanced technologies such as predictive analytics (Litsey & Mauldin, 2018) and services for users with special needs (Shea & Derry, 2019). The most recent studies (2023-2024) focused more on Al implementation and the associated ethical challenges, indicating a shift towards more advanced technologies. Table 1. Literature review data extraction results No. 1 Author(s) Year Ke; Featu);'es Challenge The Role of Libraries Investigating applications of Artificial Intelligence in university libraries of Pakistan: An empirical study Muhammad Asim, Muhammad Arif, Muhammad Rafiq, dan Rafiq Ahmad 2, 2023 | Text-to- speech, Integrated infrastructure speech-to-text, | needs, lack of Google Assistant, RFID, intelligent data analysis. budget, high costs, lack of staff expertise. Adopt and implement AT applications for library services. Unintended consequences of humanoid service robots: A case study of public service organizations Yazhu Maggie | 2024 | Ability to Unexpected . Location Pakistan As a public service | Australia ‘Wang, Sabine Matook, dan Alan R. Dennis understand, impact on imitate speech, | customers and and interact physically. service organizations. organization that adopts humanoid robots. No. 3. 4. Author(s) Year Key Features Challenge The Role of Libraries An initial interpretation of the U.S. Department of Education's Al report: Implications and recommendations for academic libraries Leo S. Lo 2023 | It was not specifically mentioned, focusing on the Library staff and | Promote AT users’ resistance; ensure transparency literacy, engage in | States Al implementation, develop AT implications of | and involvement | guidelines, AT policies in of educatorsin | prepare for AT general. the use of AL issues, and collaborate with stakeholders. Knowing What the Patron Wants: Using Predictive Analytics to Transform Library Decision Making Ryan Litsey 2018 Machine . Location United Ithasnotbeen | Usinglibrary data | United dan Weston Mauldin 5. learning explained in algorithms and | detail in the predictive analytics. article. to develop predictive analytics systems. Transforming academic librarianship through AT reskilling: Insights from the GPT-4 exploration program Leo S. Lo 2024 | Advanced Al Technical and States Integrating Alin | United technologies like GPT-4 6. ethical library services challenges of AI. | and operations. States Perceptions of artificial intelligence: A survey of academic librarians in Canada and the United States Sandy Hervieux dan Amanda Wheatley 7. 2021 Virtual Lack of assistants such | understanding as Google Integrate Al to improve library and training of | services and Assistant, Siri, | librarians Alexa, and Cortana. related to Al workflows. Canada; United States Students with disabilities and library services: Blending accommodation and universal design Casey 2022 | - Alexander Roberson, Trisha Barefield, dan Eric Griffith 8. Not mentioned | Provide inclusive | United services for students with disabilities. States Academic librarian perceptions of facilities and services for persons with visual impairments Syahrani Nur | 2023 | - Lack of Azizah dan Provide adequate | Indonesia accessibility and | facilities and Rahmi library policies services for users for people with | with visual visual disabilities. impairments. No. 9. 10. Author(s) Year Key Features Challenge The Role of Libraries Academic Libraries and Autism Spectrum Disorder: What Do We Know? Gerard Shea 2019 - . Location Lack of training | Identify initiatives | United dan Sebastian Derry and awareness that academic for librarians, as | libraries can well as limited consider to help resources, services, and students with ASD. States space. Compliance of Section 508 in public library systems with the largest percentage of underserved populations Yong Jeong Yi | 2015 - Source: author's documentation Lack of awareness of As a mediator to United support digitally | States web accessibility | excluded user in public interactions. libraries. Despite the dominance of studies from the United States, research from other countries provides valuable perspectives. The study from Pakistan provides insights into the application of Al in university libraries in developing countries (Asim et al., 2023). Research from Australia explored the unintended consequences of humanoid robots in public service organizations (Wang et al., 2024). The study from Indonesia focuses on accesibility for visually impaired users, adding an important dimension to the global discussion on inclusivity in libraries (Azizah & Rahmi, 2023). This temporal and geographical distribution suggests some areas that require further research. There is a need for more studies from different countries and regions, especially from developing countries and Europe, which are underrepresented in this sample. Also, while there has been an increased focus on Al and advanced technologies in recent years, there is still room for further research on the long-term impact of these technologies on library services and their users. Comparative research between countries could also provide valuable insights into differences in technology adoption and accesibility practices across different cultural and economic contexts. RQ.1 Analysis Results: Key Features of AI Technology for Accessibility in Libraries Based on the results of the analysis of the articles obtained, there are several features that are commonly used in library services, including speech-to-text, text-tospeech, RFID (Radio Frequency Identification), intelligent data analysis (Asim et al., 2023, p. 4); machine learning algorithms and predictive analytics (Litsey & Mauldin, 2018, pp. 142—143); as well as advanced Al technologies such as GPT-4 (Lo, 2024, pp. 1—2). Text-to-speech and speech-to-text, particularly, have the largest percentage of usage (44.5%) as reported by Asim et al., “The findings (Table 2) shows that the majority of the respondents were using text-to-speech and speech-to-text technologies to market their library resources among disabled library users (53; 44.5 %), Google Assistant to search for information by voice command (44; 37.0 %), RFID system for sel-fcheckout and check-in and/or security purposes (38; 31.93 %), intelligent data analysis for collection management (35; 29.4 %), patron recognition by QR Code, Bar Code (31; 26.1 %), and language translation to access multilingual resources (31; 26.1 %).” (Asim et al., 2023, p. 4) On the other hand, Ryan Litsey and Weston Mauldin particularly mentioned The Automated Library Information Exchange Network (ALIEN) system in the following explanation. “The Automated Library Information Exchange Network (ALIEN) is the first library developed machine learning/predictive analytic tool that can analyze a variety of disparate library data points, combine those points into a holistic picture of library behavior, as well as recommend and predict services and actions that can help improve the overall function of the library.” (Litsey & Mauldin, 2018, p. 140) Two articles specifically mention the use of the Google Assistant as virtual assistant, namely the research of Asim et al. (2023, p. 4) and Hervieux & Wheatley (2021, p. 4). In addition to Google Assistant, Hervieux & Wheatley also mentioned the use of Siri, Alexa, and Cortana by academic librarians in the United States and Canada as stated below, “Further emphasis has been placed on virtual assistants (VA) also known as voice assistants, digital assistants or intelligent personal assistants, mainly the mass-market applications Google Assistant, Siri, Alexa, and Cortana.” (Hervieux & Wheatley, 2021, p. 4) One of the papers that discusses the use of robots in public libraries in Australia mentions the ability to understand, imitate speech, and interact physically as its main features, shown in the following quotation. “Drawing on prior literature on robot capabilities (...), we identify the following five capabilities (...): (1) ability to understand human speech, (2) ability to mimic human speech, (3) physical embodiment, (4) sensory ability, and (5) human-like appearance.” (Wang et al., 2024, p. 10) Unfortunately, research on the specific use of Al for services for users with disabilities is not explicitly mentioned in some of the articles analyzed. Table 2 presents a summary of the results of the identification of the main features of Al technology extracted from various articles that have been analyzed. Table 2. Identification of key features of AI technology FEATURES Google Assistant Siri rtual Assistant applications Alexa Cortana Comprehension Mimic speech Physical interaction Text-to-speech Speech-to-text Machine learning algorithms Al Technology and Algorithms [@¥esHEnyEN BN e Advanced Al technologies like GPT-4 Intelligent data analysis RFID (Radio Frequency Identification) Source: author's documentation from the results of the analysis of the reviewed article The key features of Al technologies as displayed in Table 2 are divided into four main categories. The first category is virtual assistant applications, which includes four popular examples: Google Assistant, Siri, Alexa, and Cortana. This illustrates the range of Al assistants available to consumers from leading technology companies. The second category is interoperability, which includes five important capabilities: comprehension, speech imitation, physical interaction, text-to-speech, and speech-totext. These features demonstrate Al’s ability to interact with humans through various modalities, enabling more natural and flexible communication. The third category is Al technology and algorithms, which includes three elements: machine learning algorithms, predictive analytics, and advanced AI technologies such as GPT-4. These describe the core of Al processing and reasoning capabilities. The last category is miscellaneous, which includes intelligent data analytics and RFID, showing additional applications of Al technologies in data analysis and object tracking. Overall, Table 2 highlights a comprehensive overview of the various aspects of Al technology for accessibility in libraries, from user interfaces to core processing capabilities, demonstrating the breadth of applications and potential of Al in various domains RQ.2 Analysis Results Challenges Faced by Libraries in the Implementation of AI for Services In general, it was identified that the challenges in Al implementation are mostly related to human resources. Three studies cited a lack of staff expertise (Asim et al., 2023, pp. 6—7), a lack of understanding and training of librarians regarding Al (Hervieux & Wheatley, 2021, p. 8), as well as staff and users’ resistance, transparency, and educator involvement in the use of AI (Lo, 2023, p. 3). The following quotations represent these aspects. “Another participant highlighted that university librarians are not equipped with AI knowledge or expertise to implement such technologies in the libraries, and the university libraries may struggle to access professionals with the necessary skills. The lack of AI expertise may cause library staff to be hesitant to implement AI applications.” (Asim et al., 2023, p. 6) “The survey responses highlight the fact that academic librarians require more training with regard to artificial intelligence and its potential applications in libraries.” (Hervieux & Wheatley, 2021, p. 1) “The report emphasizes the importance of transparency and understanding in Al systems used in education. It suggests that AI should not be a “black box”, but rather, its decision-making process should be clear to all stakeholders, including teachers and students.” (Lo, 2023, p. 1) “One of the primary challenges could be resistance to change from both library staff and users. This resistance often stems from a lack of understanding about AI and its benefits, fear of job displacement, or discomfort with new technology. To mitigate this, libraries can adopt a transparent communication strategy that clearly articulates the benefits of AL how it will enhance library services, and how it will impact staff and users.” (Lo, 2023, p. 3) More specifically related to disability, lack of training and awareness for librarians is the main challenge reported by Shea & Derry in the implementation of academic library services for students with Autism Spectrum Disorder (ASD) in the United States, as stated below. “Findings point to lack of adequate training and awareness for librarians and staff, resources, services and spaces are universal challenges.” (Shea & Derry, 2019, p. 326) Other challenges are diverse, including the complexity of integration implementation (Asim et al., 2023, p. 7), budget constraints (Asim et al., 2023, p. 6), lack of accessibility, and library policies for people with visual impairments (Shea & Derry, 2019, pp. 329-330); technical challenges in the form of technology maintenance and updates; and ethical challenges related to the security and privacy of sensitive personal data (Lo, 2024, p. 6). Some other articles do not explicitly state the challenges faced in the application of Al but rather focus on services for users with disabilities in general. The following Figure 2 shows the concept of mapping the challenges faced by libraries in the analyzed articles. Treemap shown in Figure 2 is divided into three main categories: human resources, technical, and others. The human resources category has the largest area, suggesting that this may be the most significant challenge. Within this, there are references to several studies, including Asim et al. (2023), Lo (2023), Hervieux & Wheatley (2021), as well as Shea & Derry (2019). The technical and other categories have a smaller area, with each listing two references. There are also two additional references which did not mentioned specifically the challenges of Al implementation in the libraries. Figure 2 provides a visual representation of the challenges libraries face in implementing Al for everyday services. The prevalence of the human resources category indicates that factors related to personnel, training, or management may be a major obstacle to Al adoption in libraries. The presence of the technical category suggests challenges related to technology or infrastructure, while the “other” and “not mentioned" categories indicate that there are additional factors to consider. RQ.3 Analysis Results of the Role of Libraries in the Application of AI Technology Libraries have a central role in providing training and support for users to make optimal use of Al technology. One of the important roles of libraries is to integrate and/or adopt Al in daily services and operations. Four articles clearly emphasize integration/adoption including Sandy Hervieux and Amanda Wheatley (2021, p. 7); Muhammad Asim et al. (2023, p. 7); Yazhu Maggie Wang et al. (2024, p. 1) and Leo S. Lo (2023, 2024) as highlighted in the following statements. “Librarians’ involvement in AI implementation can manifest in several ways. One of the key areas of involvement is in the selection of Al tools. Librarians, with their deep understanding of the library's needs, can contribute significantly to choosing the most suitable AI tools. Another crucial area is the integration of Al tools with existing library systems and services.” (Lo, 2023, p. 2) “The GPT-4 Exploration Program serves as a blueprint, offering valuable lessons on integrating Al into library services and operations, ultimately ensuring that libraries continue to be indispensable centers of knowledge, learning, and innovation in the digital age.” (Lo, 2024, p.7) Furthermore, Leo S. Lo (2023, pp. 2—3) added the promotion of Al literacy; the development of guidelines for the use of Al preparation for Al-related issues; and collaboration with stakeholders as the task of libraries in the application of AI. Another article that focuses more on users with disabilities underlines the role of libraries in providing inclusive facilities and services (Roberson et al., 2022, p. 6) and adequate (Azizah & Rahmi, 2023, pp. 1, 13). Beyond these points, Yong Jeong Yi identifies the role of libraries as mediators in supporting the interaction of all users without exception digitally as quoted in the following statement. “One of the main roles of public libraries is to mitigate information disparity or narrow the gap between haves and have-nots (American Library Association, 2012).” (Yi, 2015, p. 1) Yi's writing is the result of an analytical study on compliance with the Rehabilitation Act Section 508 rules issued by the United States government to encourage libraries in providing information technology products that can be accessed by all parties, including users with disabilities (2015, p. 1). On the other hand, libraries can also be a source of data for the development of analytically intelligent systems (Litsey & Mauldin, 2018, p. 142)), in addition to identifying initiations that can be considered to help users with disabilities (Shea & Derry, 2019, pp. 329—330). However, the two articles do not directly mention the role of libraries in Al adaptation. The result highlights the critical role of libraries in integrating Al technology into their services and operations. Several articles focus on this integration, with an emphasis on promoting Al literacy, developing policies, preparing for Al-related issues, and collaborating with stakeholders. There is also a special focus on libraries providing inclusive facilities and services for users with disabilities. The role of libraries as facilitators of digital interaction for all users and as a potential source of data for analytically intelligent systems is also highlighted. While some articles provide valuable insights, there is a gap in directly addressing the role of libraries in Al adaptation.

Based on the literature studies that have been discussed, there are several important aspects in the application of Al technology for library services, especially in relation to accessibility for users with disabilities. Key Features of Effective AI Technology for People with Disabilities in Libraries Effective Al technology must have a variety of features that can help people with disabilities in libraries. Speech recognition is one of the key features, which allows textto-speech conversion and provides a virtual assistant to help users search for information or books (Impana & Manjula, 2022, p. 1). Screen reader technology is also important because it can read digital content loudly and aid in website navigation for those with visual impairments (Kirboyun, 2018, p. 1). As discussed in the Results section, text-to-speech and speech-to-text features as well as virtual assistants have been commonly implemented in various libraries. Image recognition, including OCR (Optical Character Recognition), allows for automatic description of images and text reading of images, which is especially helpful for users with visual impairments (Pundlik et al., 2019, p. 1). Touch-based interactions, such as devices with haptic feedback and accessible touchscreens, enhance user experiences with physical limitations (Luthra & Ghosh, 2015, p. 26; McGookin et al., 2008, p. 19). In addition, chatbots and virtual assistants can provide real-time book search assistance and library service information (Okunlaya et al., 2022, p. 1882; Panda & Chakravarty, 2022, pp. 4-5). Augmented reality (AR) and virtual reality (VR) technologies can also be used to create interactive learning environments and provide access to more immersive digital collections (Ebinger et al., 2022, p. 29; Fu, 2021, p. 6). Challenges of Implementing AI Technology for Accessibility in Libraries The implementation of AI technology for accessibility in libraries faces a number of challenges. One of the main challenges is the cost of implementation (Hussain, 2023, p. 16) which includes the procurement of hardware and software, as well as maintenance and upgrade costs. Infrastructure limitations are also an obstacle, such as the need for a stable internet connection and equipment that supports Al technology (Ajani et al., 2022, p. 225). The complexity of Al technology requires training of library staff so that they can operate and assist users with this technology, as well as adaptation from the users themselves (Upshall, 2022a, p. 14). Data security and privacy are critical issues, given that the protection of user data must be guaranteed and libraries must comply with existing regulations (O’Brien et al., 2018, DPP-737; 743)- The limitations of Alin terms of speech and image recognition accuracy, as well as the ability to understand different languages and cultures (Divekar et al., 2018, p. 1; Prabhakaran et al., 2022, pp. 1—2) are also challenges that need to be overcome. Finally, resistance to change from both staff and users can slow down the adoption process of this new technology, so strategies are needed to overcome this resistance (Lund et al., 2020, p. 875; Nakhoda & Tajik, 2017, pp. 528-530; J. Yoon et al., 2022, p. 1906). Challenges related to staff and users (in other words, human resources) are in fact often faced in the application of Al technology in libraries. The Role of Libraries in the Implementation of AI Technology in Libraries Libraries have an important role in the implementation of AI technology to improve accessibility. One of the main roles is the provision of the necessary infrastructure, including supporting facilities and technology as well as adequate internet networks (Gul & Bano, 2019, p. 767). Given that one of the main challenges in the implementation of AI technology is Human Resources, libraries are also responsible for training and education, both for library staff so that they can assist users with Al technology, and for users themselves so that they can make optimal use of this technology (Haffenden et al., 2023, pp. 45—46; Ridley & Pawlick-Potts, 2021, p. 6). Collaboration with third parties, such as technology providers and disabilityfocused organizations, can speed up the implementation process and ensure that the technology used fits the needs of users (Clark, 2020, p. 273; Onchagwa & Wambiri, 2022, p. 118). Continuous evaluation and development is essential to ensure that the Al technology used remains effective and relevant, with monitoring and feedback from users used as the basis for further development (Haffenden et al., 2023, pp. 38—39; Upshall, 2022b, p. 14). Promotion and dissemination of new services and awarenessraising campaigns on accessibility are also the duty of libraries to ensure that all users are aware of and avail of these services (Abdullah et al., 2015, p. 45). Additionally, libraries should implement clear accessibility policies and ensure compliance with local and international regulations regarding accessibility and use of Al technology (Papadakis et al., 2024, p. 3). The following Figure 3 presents a summary of the concept of the role of libraries in the implementation of Al technology, especially as a support for users with disabilities The structure of the diagram in Figure 3 showcases the various aspects and responsibilities associated with the library’s role in adapting and implementing Al technologies. Training and education aspects directly connected to staff and user, indicating that libraries have instrumental role in training their staff as well as educating users about Al technologies. This may include workshops, seminars, or special training programs to ensure that all parties involved have a good understanding of these new technologies. This training is crucial to ensure successful implementation and effective use of Al technologies in the library environment. Provision of infrastructures is another aspect highlighted in the diagram. It includes facilities, technology, and internet network. This indicates that libraries must ensure that they have the necessary physical and technological infrastructure to support Al implementation. This may involve upgrading hardware, improving internet connectivity, or even renovating physical spaces to accommodate new technologies. The diagram also features collaboration as an important component, which is connected to technology providers and disability organizations. It reveals that libraries need to collaborate with various external stakeholders in the AI implementation process. Collaboration with technology providers can help in getting the right Al solution, while cooperration with disability organizations can ensure that the technology implementation considers the need of all users, including those with disabilities. Policy and compliance aspects of implementing Al technologies in libraries includes two important elements: clarity of accessibility and compliance with regulations. Libraries should ensure that they have clear policies on how AI technologies will be accessed and be used by staff and users. They should also ensure that Al implementations comply with all applicable regulations, including data privacy and accessibility laws. These policies may need to include ethical guidelines on the use of Als, as well as procedures for handling potential issues or complaints. Clarity in policy and compliance with regulations not only protects the library from legal risk, but also build trust with users and other stakeholders. Another critical component in the implementation of Al technology in libraries is evaluation and development, which consists of two main elements: monitoring and feedback. A continuous evaluation process is essential to ascertain that the implemented Al technology is functioning as expected and delivering the desired benefits. Monitoring involves constant oversight of the AI system’s performance, including its accuracy, efficiency, and impact on library services. Feedback, on the other hand, involves collecting and analysing input from staff and users about their experiences with Al technologies. The information gathered through monitoring and feedback can then be used for further development, system improvements, or even to inform decisions about future Al technology implementation. Effective evaluation and development processes allow libraries to continuously improve their services and to make sure that Al technologies remain relevant and beneficial to the communities they serve. Last, but not least, is the aspect of promotion and campaigns, which related to two aspects: new services and raising awareness. Libraries need to actively promote the new Al-based services they offer to users, explaining the benefits and how to use them. This may involve marketing campaigns, demonstration sessions, or promotional materials such as brochures and videos. Raise awareness not only about the new services, but also about the potential and impact of Al in general in a library context. These campaigns can be targeted at various audiences, including regular library users, the academic community, or the general public. The goal of these promotions and campaigns is to ensure that the implemented Al technologies are utilized to their full potential and that users understand the added value that these technologies offer in their library experience. Fortunately, today there are also online outlets, such as the library’ official website and social media, that can help spread the word.

Libraries can improve accessibility for users with disabilities through the implementation of Al technologies such as speech recognition and screen readers among the many ways they can be used. While there are challenges in implementation, libraries have an obligation to address these by providing appropriate infrastructure, training staff and users, and collaborating with other stakeholders. Libraries also play an important role in promoting inclusion through technological advancements by providing special programs and tailored support for users with disabilities. In addition, there are further opportunities for research into the effectiveness and efficiency of using Al in libraries, as well as the integration of inclusive library services with AI. With the rapid adoption of Al technologies, libraries have a broader responsibility in Al-related applications and education, creating an environment conducive to continuous research by various stakeholders such as scholars and researchers; library professionals; as well as observers from both educational and library-related fields.