Let visitors choose how an artwork is explained. Give them control over vocabulary and depth, supported by audio and readable text.
On this pageOverview
Muse
Designing an AI experience that helps visitors understand art at their own pace.
Design decision
Keep the artwork and the visitor’s next question in context.
Muse uses conversational AI to meet visitors’ curiosity in the moment, grounded in the artwork they are looking at and the museum’s collection.
Problem
Why museum visitors need more than a fixed explanation
A visitor may enjoy an artwork and still leave unsure what it means or why it stayed with them. A label or recorded explanation gives them somewhere to start. The next question is more personal: a detail they noticed, a term they did not understand, or a connection they want to explore. That was the problem I focused on with Muse: helping visitors continue learning when the first explanation is not enough, and giving them a way to return to an artwork after the visit.
Research
I started by looking at how visitors make sense of art.
I visited museums in New York, Chicago, and Boston, observing how people interacted with artworks and interviewing visitors about their experiences. I wanted to understand how they explored a piece, what helped them understand it, and where they wanted more explanation.
Alongside these visits, I researched existing museum apps, compared their experiences, and reviewed App Store and Google feedback. Visitors wanted to ask questions, clarify meaning, and follow the details that caught their attention, with explanations that matched their interests and level of understanding.
Bloomberg Connects offered institution-authored audio, video, and text. Smartify combined artwork scanning with audio tours and personal collections. Review reports raised questions about navigation, explanation depth, and interruptions during use.
For Muse, the opportunity was to bring explanation controls, artwork-linked conversation, and optional personalization together. The interviews defined the needs; the comparison helped me focus on how the experience could support them.
Three goals that shaped Muse
Make room for the visitor’s next question. Keep the artwork in context while visitors ask questions, clarify meaning, and explore further.
Make personalization useful, even when setup is skipped. Offer guest access, optional interests, and saved collections for a later return.
Solution
Following one artwork through Muse
I organized the experience around the artwork and the visitor’s next question. Visitors can start as a guest, then discover a piece through scanning, browsing, or search.
Personalization needed more than one way to begin. Choosing interests gives Muse a starting point for relevance, but visitors can skip setup and still explore an artwork and choose how it is explained.
Scanning leads to artwork details, where visitors choose what to explore next. An unsuccessful identification has a retry action, so the journey has a way forward when recognition fails.
From artwork to understanding
The artwork stays attached when the visitor opens chat. Audio supports listening while looking at the piece; text supports reading, and chat gives visitors space to clarify meaning or follow a detail that interests them.
I made explanation depth a visible choice. Simple, Balanced, and Detailed let visitors choose the amount of context. Explanation detail and audio playback speed are separate, so choosing simpler language does not mean changing how quickly the audio plays.
Follow your curiosity
Saving gives visitors a way to return to an artwork or museum after the visit. Saved items also provide an input to personalization over time, alongside optional interests. What someone likes remains separate from how much explanation they need.
Design decisions
Design process
Mapping the journey from discovery to understanding: Scanning, browsing, and search lead to a detail page. From there, visitors can listen, read, ask a question, or save the artwork for later. Each choice keeps the same subject in view.
Interests and understanding are different inputs: Choosing a favorite artwork does not tell me how much someone already knows. I kept content preferences separate from explanation levels, with direct control over vocabulary and depth.
Building the app let me check the transitions: I designed and coded Muse, bringing opening chat, closing the keyboard, saving an artwork, and returning to it into the review alongside the individual screens.
Next steps
Testing and next steps
User testing is the next step. I want to examine whether visitors can choose a useful explanation level, continue with their own questions, and return to an artwork without losing their place.
The testing findings, revisions, and reflections will follow once that evaluation is complete.
Continue exploring
Pocket Patterns