By: Vivian Ma

What is Accessibility within a Museum Context?

Although accessibility has been a progressively emphasized part of museum practices for the past six decades,[1] what exactly does accessibility mean in the context of a museum? What does it encompass? And what are the guiding principles for assessing museum accessibility?

In general, accessibility means providing equitable access to everyone, regardless of their abilities and experiences.[2] It further refers to how organizations create space to accommodate the unique characteristics each person brings.[3] Since most museums are mission-driven institutions that serve the public, accessibility becomes a key factor in evaluating whether they have successfully achieved their core mission.

Various kinds of disabilities can hinder an individual’s ability to access museum spaces freely. In addition to physical, developmental, and cognitive disabilities, invisible challenges, such as emotional, social, or educational barriers, also fall into this list.[4] While each museum may adopt different strategies to facilitate effective communications and design accessible exhibitions and programs, common practices often include building infrastructures to eliminate physical barriers, providing specialized assistance, and offering audio-guiding services.

To evaluate the effectiveness of museum accessibility practices, Dr. Sheryl Burgstahler defined a set of criteria called Universal Design (UD).[5] Out of the seven principles to assess whether a product or environment can be used by everyone, some important qualities include:

1.      Equitable and flexible use that accommodates people with varying abilities and preferences

2.      Simple and intuitive use, with perceptible information accessible to all sensory abilities

3.      Effective size and space that cater to users of different sizes and mobility levels

With the UD criteria in mind, how well are museums improving accessibility? Although it is very hard to reach a generalizable conclusion across all institutions around the world, recent studies offer insights into this question by presenting the perspectives of the disability community and scholars. A July 2021 survey of visually impaired individuals in France revealed that over 80% would visit museums more frequently if the spaces and content were more accessible. They also identified some key barriers, with the top three being: a lack of adapted content, inaccessible museum layouts, and insufficiently adapted guided tours.[6] In recent research conducted by Silvia Pireddu of the University of Turin, Italy, the availability of audio guides showed uneven quality of translations that need improvement in clarity and engagement.[7]

If some common accessibility practices in museums require improvement, what should be their next step? With the inevitable rise of artificial intelligence (AI), some museums have already begun to implement AI solutions to enhance their accessibility efforts in order to better serve a broader audience.

What is AI and How is It Related to Museum Visits?

According to McKinsey & Company, AI is defined to be a machine’s capability to perform cognitive functions often associated with the human mind, such as perceiving, reasoning, learning, interacting, problem-solving, and even creativity.[8] It is generally classified into three interconnected categories:

1.      Machine learning, which is a form of AI that takes a wide range of inputs and eventually makes predictions based on patterns detected from data processing.

2.      Deep learning, which is an advanced version of machine learning that requires less human intervention but can produce more accurate results on a wider range of data.

3.      Generative AI, which is an AI model that automatically produces content in response to a prompt.[9]

AI’s ability can be summarized into two key functions, also referred to as “AI capabilities” by Dr. Mathilde Pavis: one is generating content in the form of texts, images, or sound; the other is analyzing large and complex bodies of information.[10]

Currently, museums generally gain access to this powerful technology in two ways. The first approach is to develop a bespoke AI system, either by collaborating with a third-party supplier or utilizing their own technology development team. The second approach involves employing off-the-shelf tools created by private sector businesses.[11]

With AI’s ability to automatically produce content and process complicated datasets with high accuracy, museums worldwide are exploring ways in which this innovative technology can help them better achieve their missions. To advance accessibility efforts, museums are actively leveraging AI to enhance language accessibility for foreign visitors, create tactile experiences for visually impaired visitors by integrating AI with 3D-printing systems, and generate visual descriptions that can be displayed across various devices.[12] Many institutions also utilize AI’s data-processing capabilities to accelerate content digitalization, removing mobility barriers for the public.

While there are numerous examples of museums employing AI to improve accessibility for different purposes, the next section of this discussion will focus on three areas of its applications — content digitalization, language accessibility, and visual description — by introducing specific case studies for each.

Case Study 1: AI-Empowered Content Digitalization to Account for Mobility Disability

Every museum possesses a professional archive dedicated to systematically collecting, organizing, preserving, and providing access to items with long-term or permanent research value.[13] Beyond documenting the museum’s activities and tracking the conditions of its collection, most museums also manage special collections, which may include any related materials like books, academic journals, and art monographs that align with their overarching mission.[14] The extensive scope of museum archives suggests the complexity of managing them, not to mention the challenge of selecting and making some resources available to the public.

Despite many large museums having long maintained a digital presence, it wasn’t until the COVID-19 pandemic that more institutions recognized the urgent need to fully leverage their online platforms.[15] This realization led to increased efforts in digitalizing exhibitions and making part of their archives accessible online, which allows them to transcend the limitations of their physical spaces.[16] And this is where Codemantra, a digitalization initiative offering an AI-powered document processing platform, steps in to support institutions make their resources truly accessible.[17]

In partnership with The Met, Codemantra helped to digitize over 700 art monographs into high-quality, ADA-compliant e-books.[18] The team also collaborated with Yale University Art Gallery to expedite the conversion of more than 150 rare publications into ADA-compliant formats.[19] By integrating AI technology, Codemantra can quickly extract information and properties from documents while ensuring a high level of accuracy and quality, which would otherwise take days or even months for an archival team to complete.[20]

Working with AI-driven companies like Codemantra, museums have digitalized content in line with the UD principles, enabling broader and equitable access to cultural knowledge for those unable to visit in person during and after the pandemic. Converting resources into ADA-compliant formats also ensures inclusivity for individuals with various disabilities. Further, AI’s efficiency in processing large volumes of information has expanded digitalization to a broader scale, aligning with UD’s emphasis on effective reach.

Case Study 2: Natural Language Processing (NLP) and AI QR Codes for Multilingual Support

Museums aim to convey information about aspects of human history through multiple modes of communication, like wall labels, pamphlets, and audio guides. While most institutions provide translations and multilingual audio options, it is simply impossible to cover every language for global audiences. For example, the Tokyo National Museum thoughtfully provides wall labels in Japanese, English, Korean, and Mandarin Chinese. Yet, these four languages may still be insufficient to meet the needs of visitors who don’t understand them. Moreover, multilingual support should extend beyond basic translation to accurately and comprehensively convey meanings to ensure equitable access to information for both local and foreign visitors. As Manfredi stated, “A museum translator should not only be a language expert, not a general translation expert, but rather an expert of meaning-making production of exhibition texts at all linguistic levels.”[21]

It is challenging to enhance language accessibility in these regards, but AI technology, particularly in Natural Language Processing (NLP), seems to offer a promising solution. As a subfield of AI, NLP enables computers to understand human languages and perform translations without human intervention.[22] Although it is not yet perfect at the current stage and may sometimes produce inaccuracies, the technology is expected to become more refined in the future, making it feasible to freely translate museum materials into more languages.

On top of this, AI and augmented reality (AR) have been integrated into an advanced version of QR codes that allow visitors to instantly access exhibit information on their mobile devices through simple scanning.[23] These AI-powered QR codes offer multilingual support with native pronunciation and cultural nuances, which will greatly improve the accuracy and quality of translations.[24] In other words, the translations generated by NLP can be stored within these AI QR codes, letting museums keep their original wall labels while offering equitable information to a global audience. The use of QR codes is also intuitive, which meets the UD criteria, as users only need to scan the codes to access the content.

Case Study 3: AI-Driven Audio Assistant for Visually Impaired People

The July 2021 study on visually impaired individuals in France found that 97% of respondents used some kinds of auditory assistance during their museum visits.[25] Among these participants, the most popular mediation tool was the traditional audio guides.[26] However, these audio guides are not specifically designed for visually impaired people, meaning that the scripts often contain less detailed descriptions of the exhibits since they assume the user can already see the item clearly.

As early as 2019, way before ChatGPT had introduced AI into the public consciousness, Smithsonian Museums had already partnered with Aira to implement an accessibility technology that combines artificial intelligence and augmented reality.[27] By downloading the Aira app to smartphones for free, this service connects visitors with trained sighted live agents who can remotely see what is in front of or near them.[28] Essentially, it offers on-demand verbal description and agent assistance in navigating the museum. While this is certainly a step forward in providing an audio service specifically for visually impaired people, relying on real-person agents can still introduce a level of inefficiency due to potential waiting times for connections. Yet, the rapid evolution of AI over the past five years has successfully eliminated the human element from this kind of audio service and enabled fully automated audio descriptions to better serve the visually impaired community, as demonstrated by a recent collaboration between Microsoft and the Rijksmuseum.

The Rijksmuseum has a massive collection of over a million artworks available to the public online; however, those who are blind or have low vision face limited access to this collection.[29] To address this, the museum partnered with Microsoft to create detailed audio descriptions for each artwork, capturing both the scene and the mood in a vivid manner.[30] Such a vast undertaking would be nearly impossible to manually complete, but with Microsoft’s Azure AI Computer Vision and Azure OpenAI, this task becomes feasible and can be accomplished in just a few hours.[31] The audio descriptions powered by generative AI have successfully allowed visually impaired individuals to equally explore the rich collection of the Rijksmuseum.

Conclusion: What is the Future of Museum Accessibility?

In just five years between the Smithsonian’s collaboration with Aira and the Rijksmuseum’s project with Microsoft, we can already see the rapid pace at which AI technology has revolutionized museum accessibility.

While many accessibility practices in these case studies could be performed without AI, its real advantage lies in delivering accuracy and efficiency. With vast collections to manage, museums can leverage AI to share information on a larger scale and in a more customized way.

Indeed, the applications of AI in museums are appealing, but can they really replace existing practices and become widespread across institutions globally?

The museums discussed in these three case studies are all large and internationally renowned institutions that benefit from high visitor numbers. This steady influx of visitors can somewhat ensure that the new implementation of the AI approach is a cost-effective deal. However, is it necessary for smaller local museums to follow in their footsteps? This decision requires careful consideration by museum board members and directors after a thorough investigation and discussion.

Additionally, there are always trade-offs associated with AI convenience:

1.      The potential loss of job opportunities replaced by AI can raise ethical concerns.

2.      AI systems are not free from flaws; false or biased outcomes may harm a museum’s reputation.

Nonetheless, artificial intelligence has certainly opened up a world of possibilities to make museum experiences more accessible to all — let’s see what the future holds.