ESSAY
The artificial curator
Artificial intelligence is an inescapable part of our lives now and is seeping into every corner of our culture. But does it have what it takes to run a museum?
Museums are institutions of cultural, historical and scientific significance, preserving the material and immaterial heritage of humanity. They play a critical role in education, research and the shaping of public memory. Traditionally, curators, who are trained experts in their fields, have been responsible for the selection, organization, interpretation and display of museum collections. However, with the advent of artificial intelligence (AI), a question arises: could an AI run a museum?
This inquiry involves not only the technical feasibility of AI curating but also the broader implications for cultural heritage, biases and the nature of curation itself.
AI and the technical feasibility of running a museum
AI has already demonstrated capabilities in various aspects of museum operations, from managing collections to engaging visitors. For instance, the Museum of Modern Art (MoMA) in New York has employed AI to analyze visitor behaviour and optimize exhibit placements. It has also deployed AI-driven chatbots and virtual assistants to help visitors get the most from their time in the museum.
This is clearly seen as a fruitful path for further development and led to IBM creating Watsonx specifically to provide virtual assistants in museums, enhancing the visitor experience by answering questions and providing information about exhibits.
Additionally, AI systems can efficiently manage databases of artefacts, cataloging and retrieving items with speed and precision that surpass human capabilities.
Moreover, AI can assist in identifying and categorizing artefacts by analyzing vast amounts of data, such as images, texts and historical records. Deep learning algorithms have been trained to recognize patterns and features in artworks, enabling them to attribute unidentified pieces to specific artists or periods. An example of this is the collaboration between Microsoft and the Rijksmuseum in Amsterdam, where AI was used to fill in missing parts of Rembrandt’s famous ‘Night Watch’ painting, based on the style and techniques of the artist.
While these examples illustrate that AI can handle many technical aspects of museum management, the question remains whether AI can truly replace human curators in the more subjective and interpretative aspects of curation, although it’s believed this has not stopped some institutions from trying.
AI curation and cultural biases
One of the most significant challenges in AI curation is the issue of cultural biases. Traditional museums have often been criticized for reflecting the prejudices of their time, such as Eurocentrism, colonialism and gender bias. The selection of what is displayed and how it is interpreted is influenced by the curator’s cultural background, education and the socio-political context.
AI models, however, are trained on existing data, which inevitably carries the biases of the sources. If an AI is trained on a dataset that is predominantly Western, it might perpetuate the same Eurocentric biases found in traditional museums. A study by Gebru et al (2018) highlights how biases in AI training data can lead to skewed outcomes, which in the context of a museum, could mean under-representation or misrepresentation of certain cultures and histories.
For example, if an AI is tasked with curating an exhibition on African art but is trained on a dataset that heavily emphasizes Western interpretations of African culture, it might fail to capture the nuances and diversity of African artistic traditions. This could result in an exhibit that reinforces stereotypes rather than challenging them.
In contrast, human curators, despite their own biases, have the ability to reflect critically on their work and make conscious efforts to address and rectify cultural prejudices. This reflexive practice is crucial in ensuring that museums become more inclusive and representative of diverse perspectives. While AI can be programmed to recognize and counteract certain biases, it is limited by the data it is trained on and lacks the deeper understanding required to navigate complex cultural issues.
Accuracy and understanding in AI curation
The ability of AI to identify, categorize, collate and label exhibits with the same accuracy and depth of understanding as a human expert is another area of concern. AI excels at pattern recognition and data analysis, but it often lacks the contextual knowledge and interpretative skills that human curators bring to the table. This can lead to obvious errors, although less-obvious mistakes are arguably a bigger concern.
For instance, when categorizing an artefact, a human curator considers not only the physical attributes of the object but also its historical context, cultural significance and potential connections to other items in the collection. AI, on the other hand, might rely heavily on superficial features, leading to misclassification or oversimplification. In a study conducted by Arora et al (2020),it was found that AI algorithms could identify artistic styles in paintings with high accuracy but struggled to understand the underlying cultural and historical significance.
Furthermore, AI might make mistakes that a human curator would likely avoid. For example, an AI could misinterpret an artefact from a minority culture by applying a dominant cultural framework, resulting in a misrepresentation of the artefact’s significance. Such errors could undermine the educational value of the exhibit and perpetuate cultural misunderstandings.
AI and breakthroughs in understanding
One of the most exciting potentials of AI in museum curation is its ability to make breakthroughs in understanding. AI’s capacity to process vast amounts of data and identify patterns that humans might overlook could lead to new insights into historical periods or cultures. For instance, AI could analyze a large corpus of ancient texts and artefacts, identifying previously unrecognized connections or trends that challenge existing models of history.
However, this potential is not without its limitations. AI is inherently conservative in that it tends to reinforce existing models unless explicitly programmed to explore alternative hypotheses. If an artefact does not conform to the established model, an AI might be more likely to dismiss it as an anomaly rather than considering it as evidence that could lead to a revision of the model.
This contrasts with the approach of a human curator, who might recognize the significance of an outlier and explore its implications for our understanding of history. Human curators are trained to question and revise existing narratives, a process that is essential for the advancement of knowledge. For example, the discovery of the Rosetta Stone led to a breakthrough in understanding ancient Egyptian writing, not because it fit neatly into the existing model, but because scholars recognized its potential to challenge and expand that model.
In this sense, AI may be more prone to maintaining the status quo rather than driving revolutionary changes in our understanding of history and culture. While AI can assist in identifying patterns and generating hypotheses, it ultimately requires human intervention to interpret the results and determine their significance.
The human qualities of curation
Curation is not just a technical task; it is an art that requires creativity, empathy and an understanding of human experience. A curator’s ability to craft a narrative, to evoke emotions and to engage with visitors on a deep, personal level is something that AI, at least in its current state, cannot replicate.
For example, an exhibition on the Holocaust curated by a human might be able to convey the emotional gravity of the subject matter through the careful selection of artefacts, personal stories and interpretative texts. The curator’s understanding of the human condition, the moral implications of the Holocaust and the need for sensitivity in presentation are qualities that AI lacks.
Moreover, curation involves notions such as ‘relevance’, ‘significance’ and ‘taste’, which are subjective and context-dependent. AI, which operates on logic and algorithms, might struggle to grasp these concepts fully. For instance, the decision to include a particular artefact in an exhibit might depend on its relevance to current social or political issues — a determination that requires an understanding of contemporary society that AI may not possess.
In addition, taste, which is informed by cultural, aesthetic and intellectual factors, is inherently subjective and constantly evolving. A human curator might choose to highlight a particular artist or movement because of its significance in the context of a particular cultural moment, something that an AI might not appreciate or recognize.
Deeper understanding
To sum up, while AI has the potential to enhance various aspects of museum operations, it is not yet capable of fully replacing human curators. AI can manage collections, assist in identifying and categorizing artefacts and even contribute to new insights, but it lacks the deeper understanding, creativity and empathy required for the more subjective aspects of curation.
Moreover, AI is susceptible to the biases present in its training data, which could perpetuate the cultural prejudices of traditional museums. While AI can be a powerful tool in the hands of human curators, curation ultimately demands human qualities that AI cannot provide — qualities such as critical thinking, emotional intelligence and a deep understanding of human culture and history.
As AI continues to evolve, it will undoubtedly play an increasingly important role in museums, but it is unlikely to replace the need for human curators. Instead, the future of museum curation will likely involve a collaboration between AI and human experts, combining the strengths of both to create more inclusive, accurate and engaging exhibitions. This partnership has the potential to push the boundaries of what museums can achieve, but it will require careful consideration of the limitations and ethical implications of AI in the curation process.
