As AI rapidly approaches ubiquity, what role should it play in the heritage sector? How might AI affect the museum experience? What might it do for the museum experience? These are the questions behind CultureQuest: an interactive, AI-powered ‘quest experience’ for museums and galleries designed to unlock new ways for visitors to explore and engage with collections.
CultureQuest emerged from existing collaboration between two University of Bristol academics—Dr Richard Cole (Bristol Digital Game Lab co-Director, Lecturer in Digital Classics) and Dr Chris Bevan (Lecturer in Computer Science)—and gaming company Meaning Machine, known for using generative AI in their immersive ‘first-person talker’ games. The project’s aim was to bring this innovative technology into a heritage space, drawing on Richard’s research into interactive historical simulations and the expertise in immersive heritage experiences Richard and Chris gained through working together on the Virtual Reality Oracle. The team partnered with Bristol Museum and Art Gallery to build and test a quest prototype designed around its ancient Egypt gallery, with funding from Digital Catapult Creative Connect.
When Richard asked me to join the project as a research assistant, I knew I’d be mad to refuse. The quest experience, I learned, would make use of Meaning Machine’s Game Conscious™ technology to create non-playing characters (NPCs) that would be powered by large-language models (LLMs) fine-tuned on data provided by the museum, alongside further materials and resources approved by curatorial expert Amber Druce. By conversing with these NPCs and completing the tasks they set, the museum visitor is gently nudged around the gallery space and encouraged to engage with objects and ideas in an active, curious way. Our aim was to support—rather than replace—the gallery’s existing curatorial framework. Ethical responsibility, accessibility, inclusivity, sustainability, and scalability were at the heart of the project from the outset.
The plan—across a helter-skelter three-month sprint, April to June 2025—was to create a working prototype as proof-of-concept for the quest experience. In April, we defined workstreams, gathered knowledge about ancient Egypt and the museum’s Egyptian gallery, and drafted a paper prototype of the quest. Richard and I were surprised (and pretty excited) to find that we would be designing the structure of the quest ourselves, as ’quest architects’. With Meaning Machine’s support, we embarked on a complicated, frustrating, and fascinating process of mapping out characters, objects, and ideas we wanted visitors to encounter. Finding an order for those encounters that worked logistically and striking the right balance between NPCs, objects, and the gallery’s existing narrative wasn’t easy, but after four rounds of iteration we had our characters and our story.
Picture the scene: you wander into the hushed darkness of the Bristol Museum’s Egyptian gallery (or, in the presence of a school group, not so hushed…). Upon scanning a QR code you become Iy-en-Amen-nay-es-nebet-ta, an Egyptian woman from seventh century BCE Thebes whose coffin is displayed in the gallery. You are welcomed to the Egyptian afterlife (duat) by the god Osiris, who tells you that to reach the Field of Reeds—the realm of the righteous dead—you must first prepare both your body and your soul… So begins your quest, during which you meet a clutch of Egyptian deities—Osiris, Anubis, Maat, and Ra—rummage around for some containers for your internal organs, seek out instruments used to ward off evil forces, hunt down a mummified reptile, and—eventually—undergo the Weighing of the Heart test (nervously eyeing the crocodile-hippo monster, Ammit the Devourer, who lurks nearby). In the process, we hope, you become an enthusiastic amateur Egyptologist.
Our quest structure in place, May was devoted to designing and recruiting for a study aimed at helping us understand how members of the public experience the quest. Meanwhile, Meaning Machine began crafting the knowledge database behind the LLM and designing the WhatsApp-style interface through which visitors will interact with the NPCs (‘GodsApp’). This work carried through to June, during which Meaning Machine continued fine-tuning the parameters of the LLM while Richard and I worked on injecting some personality into our NPCs. We put together a mood board to guide Meaning Machine’s illustrator, studiously avoiding problematic internet tropes: ‘ripped Ra’, ‘zombified Osiris’, ‘sexy werewolf Anubis’… The final visuals, we think, nicely balance the original Egyptian iconography with the dynamism required for digital gaming.
We then began to imbue the NPCs with character quirks, to make each one unique and enjoyable to interact with. (Ra, for instance, has a secret phobia…but you’ll have to play the quest to find out what!) These personality details will be entered into Meaning Machine’s bespoke tooling, which sits on top of the LLM and includes Conversation Creator, an easy-to-use interface that enables even AI rookies like myself to influence how the NPCs come across.
We’ve learned a huge amount from this fast-paced project and came up against some knotty challenges. From the industry partner perspective, this included introducing Game Conscious™ NPCs into a physical space where they had to interact with real objects, people, and histories, handle sensitive topics appropriately (e.g. regarding an object’s colonial past), and nurture player curiosity while maintaining focus on the quest at hand. Balancing accuracy and characterisation—so NPCs are informative without sounding like academic databases—was also a tough nut to crack. From our perspective as researchers, stepping into new roles such as ‘quest architect’ presented a steep learning curve, and I for one have occasionally had to be reminded that ‘fun’ is an essential part of good game design. There were also the usual hurdles of interdisciplinary, cross-sector work to overcome: wrangling meetings, balancing competing priorities, keeping communication flowing. But, at the same time, working across organisations in this way offered all of us new perspectives and insights—I personally relished the opportunity to peek behind the scenes of commercial digital game design in action.
This initial phase of work culminated in a study on Monday 28 July, during which 40+ volunteers trialed the quest experience either solo or in small groups. We observed whether and how the quest influenced how participants navigated the gallery and engaged with its objects, and interviewed participants to discover if the quest helped them learn more and/or differently and whether—crucially—it was fun. Our aim was to answer three main questions:
- How do visitors respond to this quest experience?
- Does it enhance their engagement with the collection?
- Can conversing with these AI-powered NPCs help turn museum visitors into researchers?
We’ll publish the results of the study in due course—so watch this space!