Ensuring Ethical and Responsible AI Use in Museums
When considering how to ensure ethical and responsible AI use in their museums, leadership is navigating a crucial question: How can our institution benefit from artificial intelligence without causing harm?
With careful planning, it may be easier to balance the two than you think.
During “Navigating the Transformative Impact of Artificial Intelligence on Museums,” a session at the AAM’s Future of Museums Summit, 60% of poll respondents cited ethical concerns as their biggest concern about implementing AI at their museums.
As cultural stewards, it’s clear that museums recognize the necessity and importance of ethical and responsible AI use. However, many leaders and stakeholders are left asking how to build an AI use strategy for their institution that fosters fairness, transparency, and cultural integrity.
If you’re wondering the same thing, you’re not alone. Though using AI ethically and responsibly in your museum may seem like a Herculean task, it is manageable with careful planning, special considerations, and a solid strategy.
Want to dive deeper?
Watch the recordings for our webinar series, Navigating AI for Museums: A Purposeful Approach to Innovation.
The second webinar, “AI in Museums: Upholding Ethics, Cultural Sensitivity, and Data Privacy,” explores ethical and responsible AI use in more detail.
Ethical vs. responsible AI use in museums
At an operational level, Microsoft defines artificial intelligence as “the capability of a computer system to mimic human-like cognitive functions such as learning and problem-solving.”
Though ethical and responsible AI use in museums are closely related concepts, there are distinctions between the two that museums must keep in mind.
Ethical AI is broadly associated with morality (or the “right vs. wrong” of AI), while responsible AI relates to the operational and tactical side of AI, or how it is applied (though ethics also play a part).
In the simplest terms, ethical AI asks us, “What should we do?”, and responsible AI asks, “How will we do it right?” For museums:
Ethical AI: Involves deploying AI applications that uphold fairness, transparency, and respect for cultural heritage and visitor privacy and align with museums’ educational and social missions.
Responsible AI: Means using AI to enhance visitor experiences and operational efficiency while mitigating risks, preserving the integrity of collections, and ensuring inclusive access to cultural knowledge.
Let’s say your museum uses an AI solution to generate artwork descriptions.
Ethical AI would ensure that the solution is trained on diverse, unbiased datasets to generate descriptions that won’t reinforce colonialist or other problematic narratives. Responsible AI would involve having a human curator review and validate its outputs to safeguard against bias and inaccuracies.
Ethical AI frameworks for museums
Not sure where to start when creating a practical ethical AI framework for your museum? Depending on where you are in your journey, the following resources may help guide you:
UNESCO’s Recommendation on the Ethics of Artificial Intelligence outlines ethical AI practices and actionable policy advice for implementation. Though not specific to the GLAM sector, it is a helpful starting point.
The Smithsonian Institution’s AI values statement and the process its team used to develop it are explained in “Developing responsible AI practices at the Smithsonian Institution,” a journal article in Research Ideas and Outcomes.
Funded by the Arts and Humanities Research Council (UK), The Museums + Artificial Intelligence Network’s “ethically robust framework” supports museums in responsibly engaging with AI and offers a free planning toolkit.
In the book, AI in Museums: Reflections, Perspectives and Applications, Sonja Thiel’s essay, “Managing AI: Developing Strategic and Ethical Guidelines for Museums,” presents guiding principles and explores questions related to implementing ethical AI solutions in museums—even recommending potential KPIs.
While these resources can provide useful recommendations and guidelines, your museum’s mission, vision, and values should form the foundation for your unique ethical AI framework.
Considerations before using artificial intelligence
When deciding what ethical and responsible AI use looks like at your specific museum, several important considerations must be made.
As with integrating any new technology, a balanced approach is best, and thorough preparation is key before you decide which solutions to implement.
Before using AI in your museum, consider the following:
Data privacy and usage: Ensure that any museum visitor data you collect for personalization is ethically used and safeguarded in line with current data protection guidelines.
Transparency: Be open with museum visitors about how your institution uses AI. Before collecting and using visitors’ data, you must ask for consent and share how you will protect any personal data that is collected.
Job Impacts: Consider how AI adoption may affect jobs at your museum. Be mindful that your staff may be concerned about their job security if you implement AI tools. Though these tools are becoming very advanced, human review is still an essential part of using AI ethically and responsibly in museums. Highlighting this can help to provide reassurance.
Bias: Just like humans, AI tools can be biased. Prevent inherent biases from creeping into your machine learning systems by carefully deciding which information to include and having a human reviewer check the outputs.
Decolonization: The outputs from AI tools are only as good as the data they are trained on. Consider whether the data you are inputting into your tool may have a colonialism bias, and diligently review the tool’s outputs to avoid perpetuating colonial mindsets.
While these are a helpful starting point, you may also need to keep other considerations in mind, depending on your museum’s unique objectives and challenges.
Responsible AI implementation strategies
After considering what ethical and responsible AI use in your museum looks like in practice, you are ready to begin laying the groundwork for your AI implementation.
Below, you’ll find practical approaches that can support you on your journey.
Develop an AI ethics statement
Having a public-facing ethics statement that clearly outlines your museum’s approach to AI serves two important purposes:
It acts as a guide as you decide which solutions to implement, when, and how
It showcases your commitment to transparency and builds trust with your stakeholders, including donors, your staff, visitors, and the public
Your ethics statement doesn’t need to be unnecessarily complex.
At a minimum, it should include a clear definition of how your museum uses AI, the principles guiding your AI application(s), and the safeguards you use to reduce potential risks, such as bias or privacy concerns.
Address AI data biases
Though AI is not a sentient being, it can develop biases just like humans can. Spotting biases at the data level is crucial for preventing social inequalities from being reflected in AI outputs and ensuring responsible AI use in museums.
To combat AI-related biases, it’s essential to examine your existing datasets for potential biases or gaps. Some questions you may consider are:
Given the content of this dataset, and the historical period in which the artifacts were situated, where might I expect to find biases?
Was the language used to catalogue the artifacts in this dataset biased?
What data might be missing from this dataset (non-official records, community histories, alternative interpretations, etc.)?
Is it possible that the way these artifacts were catalogued overemphasized certain aspects and underemphasized others?
Do many of the works contained in this dataset contain missing values, and if so, how does this impact our predictions and conclusions?
Your museum’s current processes for bias detection can also serve as a helpful guide in preventing AI biases.
Train museum staff on AI
Your staff members are one of your greatest assets. To ensure adequate support, provide thorough, ongoing training on your museum’s AI tools to ensure staff understand their functionality and applications.
Training is also an excellent opportunity to champion what responsible AI use in museums looks like—in general and at your specific museum. This training should include information on how to spot potential biases and adhere to ethical guidelines.
You may also want to consider creating internal user guides to distribute after training. These can help ensure that expectations are clear, all staff are on the same page, and new team members and volunteers can be trained quickly.
As mentioned, museum staff may also be concerned about how AI will impact their jobs. Training is an excellent place to provide reassurance on job security and explain how AI tools can help them in their day-to-day lives without threatening their crucial contributions.
Implement human review
Ethical and responsible AI use in museums is not possible without human review. To prevent biases and mitigate potential harm, implement a verification system in which museum staff review any AI outputs before using them.
To illustrate what this might look like, let’s consider the example of an AI-backed metadata tagging tool. This solution can help museum staff save valuable time and streamline image cataloguing by using AI to identify keywords for objects, people, text, and colours. Without review, the AI tool may misidentify important components or perpetuate biases.
Terentia’s AI-powered metadata tagging tool facilitates human review by assigning confidence scores to each keyword generated. These confidence scores—which essentially measure the likelihood that the tool has identified the item correctly—can be reviewed by museum staff.
Before finalizing the outputs, these staff members can make any edits necessary to ensure accuracy and prevent biases from becoming problematic.
Integrate AI into existing policies
After deciding how to ensure ethical and responsible AI use in your museum, be sure to update your existing policies.
In addition to crafting an ethics statement, update your data governance, privacy, and security, and DEIA (diversity, equity, inclusion, and accessibility) policies to include AI considerations.
With data breaches becoming more prevalent, it’s essential to be clear on how you are protecting data privacy and ensuring that your AI tools have robust protections in place.
Relatedly, prioritize solution vendors with clear AI use policies—like Terentia provides on our website—that offer transparency into how your data is stored, used, and protected.
Be transparent about AI use
Transparency is another important cornerstone of responsible AI use in museums. Broadly speaking, it means clearly communicating how the AI tools you use were built, the data you use to train them, and how they make decisions.
Being transparent about how your museum uses AI helps build stakeholder trust and ensures that your visitors can benefit from all AI offers without worrying about their privacy.
From the beginning, be sure to adopt AI tools and systems with strict data protection policies, and consider:
Labelling AI-generated content to ensure it isn’t confused with human-generated content
Documenting how AI content was produced so it’s clear which elements of objects or metadata are AI-generated
Explaining how humans are involved in reviewing AI outputs to ensure accuracy, prevent biases, and mitigate harm
Sharing how and when AI is being used to collect data from visitors, and how this data is being protected
When you aren’t transparent about how you’re using AI, your stakeholders may come to their own conclusions. It’s best to be proactive and share this information to avoid potential misunderstandings.
Conduct AI ethics audits
Setting up a formal monitoring process is important to ensure your AI tools don’t accidentally cause unintended consequences.
Ethics audits formally monitor your AI solutions and ensure their outputs still align with your institutional values. Choose a schedule for your audits that makes sense for your museum, and during the audits, review your AI systems’ outputs for possible biases or ethical concerns.
If you notice a lack of alignment, it’s time to prepare an action plan.
Say your yearly audit identifies that one of your AI tools is perpetuating negative stereotypes about a particular group. This could be because it’s picking up biases in your collections data, in which the historical portrayal of certain groups or events may be prejudiced.
Your action plan could include reviewing the collections data to identify what is causing the negative stereotypes to be perpetuated and taking steps to mitigate the biases. For example, perhaps the object descriptions in your collections data could be revised to include more inclusive language.
During these ethics audits, let your museum’s values be your compass and have a formal process to document how you investigated and resolved any concerns raised.
Collaborate and share best practices
Lean on one of the museum community’s greatest strengths—its collaborative culture—as another resource in your AI toolkit.
By participating in AI-related conversations and events with other museums, you can stay updated on the latest developments in AI for museums, share your experiences, and learn from others.
Partnering with AI specialists experienced in working with museums and other GLAM institutions can further help bridge knowledge gaps during implementation to help you ensure you are using AI tools ethically and responsibly.
As time passes and AI tools evolve, the museum community’s ability to come together to discuss best practices surrounding AI use in the sector will be a valuable asset.
Benefits of using AI ethically and responsibly
When implemented responsibly, AI can offer significant transformative benefits to museums.
The foundation for getting these benefits is ensuring your institution has robust, transparent ethical and responsible AI use guidelines and a privacy policy that ensures any data collected is kept safe and private.
Specific advantages of using ethical and responsible AI in your museum could include:
Enhanced visitor experiences through personalization: With AI, providing a personalized experience to each museum visitor is becoming a reality. For example, by using an AI chatbot, you can empower a group of friends viewing the same artifact to ask the chatbot the questions they are most interested in, sparking further discussion and helping them connect more deeply with your collections.
Better accessibility for diverse audiences: Carefully implemented AI has the power to transform museum accessibility and deliver richer experiences to diverse groups. For example, AI tools can create audio descriptions for Blind or visually impaired visitors, help visitors with mobility impairments navigate your museum via an app, or provide real-time captions that make spoken content more digestible for those who are Deaf, hard of hearing, neurodivergent, or non-native speakers.
Automated DAM and collections management tasks: AI-driven digital asset management and collections management solutions, such as those offered through Terentia’s platform, can help your museum staff automate processes and streamline workflows to save valuable time. These tools, powered by natural language processing (NLP) and machine learning, can assist with metadata tagging, cataloguing, organizing collections, and more.
Deeper insights into audience behaviour and preferences: AI analytics can empower your museum to understand better how visitors behave, what they are looking for, and how you can enhance their experiences. These outputs help you inform programming, marketing, staffing, and exhibit design decisions based on audience needs. AI can offer valuable, actionable insights—from generating heat maps of visitor movement to segmenting visitors based on their preferences, so you can communicate with them more effectively and build engagement.
By championing ethical and responsible AI use, museums can leverage AI in exciting ways while upholding their responsibility for “social good” within communities.
Interested in using AI at your museum?
Terentia’s AI-powered platform can help you enhance asset discoverability, gain fresh insights, and save valuable time. Streamline your image cataloging with automated metadata tagging, transform videos and images into discoverable data, and create smart links across your collections data.
Microsoft Azure is the powerhouse behind our solutions, so you can be confident that your data will be secure.
Explore the potential of AI for museums
Building a solid institutional AI foundation focused on ethical and responsible use will go a long way toward ensuring your successful implementation.
With this foundation in place and stringent data protection and privacy policies, you’re ready to move toward using AI to tackle your cataloguing backlog, better meet your visitors needs, and create a more inclusive, accessible institution.
A responsible AI solutions provider, experienced in working with cultural institutions, can help guide your implementation and help you cover all your bases. At Terentia, we’re proud to champion ethical AI use through our comprehensive AI use policy, which ensures our client data remains private, protected, and under your control.
Ready to take the next step in your AI journey? Reach out to Terentia to receive a personalized demo of our AI-powered features.