Designing and Building a Platform for Close Looking

First published on the Barnes Foundation's Medium, May 2026.

How a pedagogical problem became a teaching platform for art history and appreciation, and what we learned by listening to our students.

Some readers will already know the Barnes Visual Experience Platform (or VXP), some may have already taken a class on it! The short version of our “why”: online education platforms were not built for visual material. Zoom can advance a slide deck, or share a painting on a screen. It cannot let a student linger over brushstrokes at their own pace while an instructor steers the group’s attention across a composition.

The Barnes Method depends on close looking, dialogic inquiry, and sustained engagement with works of art. When we moved our adult education programs online in 2020 in response to the pandemic, no off-the-shelf platform could support that pedagogy without collapsing it into a slideshow. We moved in-gallery classes to Microsoft Teams, and while it kept classes running, it did not preserve the Barnes way of teaching.

The Barnes Method is an art-centered approach developed by Albert C. Barnes and Violette de Mazia, drawing on Dr. Barnes’s long collaboration with philosopher John Dewey. In Art as Experience (1934), a book dedicated to Barnes, Dewey argued that fine art gets sequestered on a pedestal, disconnected from daily life, and he distinguished between recognition, which is shallow and immediate, and perception, which takes time and work. To perceive a work of art is to study it, to take in its parts and reconstruct the whole through one’s own observation. Barnes built his collection and his teaching practice around that distinction, arranging works in dense, deliberate juxtapositions and teaching directly in the galleries so ordinary visitors could learn to see. Barnes pedagogy drives the shift from recognition to perception. Instructors guide students through structured close looking, sustained over time, with intentional movement between overview and detail, whole and part. The work of art is not the subject of a lecture. In the Deweyan tradition, it is the site of an experience, and the experience is what produces the learning. A well designed virtual environment removes barriers to participation (health-related, geographic, psychosocial) and dissolves the separation between art and daily life that Dewey worked to undo.

Put simply, screenshare cannot host that experience. A painting on a shared slide is static, presented at a downgraded resolution and a fixed crop, with the student’s field of vision tied to the instructor’s frame and cursor. When detail images are supplied, they come sequentially, one slide after another or shared after the fact, fragmenting a work that should be encountered as a whole and considered with the group. There is nowhere to linger, no way to follow your own attention and explore a work at your own pace. The image becomes a reference to the real object rather than an object you can examine. For disciplines where visual material is incidental, this is acceptable, but for art history and appreciation, and for any field rooted in close visual analysis, it collapses the practice completely.

I’m excited to share the research and design process that produced the first version of the VXP, designed and built in-house in 2022, what shipped in 2023, and how the asynchronous Replay feature emerged from structured user research with our students. A subsequent post will cover what happened when another institution began teaching on the platform, and what we are building next.

Research and Design Foundations

We started with a constraint set. The constraints were pedagogical first, but they had immediate technical consequences:

• Shared attention. Participants need to look at the same thing, at the same time, with minimal tool friction.
• Instructor-led pacing. The instructor must be able to guide focus and move the group through a sequence.
• Close looking. Images must support deep zoom without degrading quality or responsiveness.
• Individual agency, bounded. Students should be able to explore details when invited, without dissolving the shared frame.
• Accessibility. Keyboard navigation, captions, and predictable interaction patterns built in, not added later.
• Reliability under live conditions. The classroom cannot fail like a prototype can.

These constraints came from Barnes pedagogy as practiced by our adult education faculty, not from any external research framework. We have since come to recognize that they also align closely with empirical findings from Penn’s Center for Neuroaesthetics, and in particular with Anjan Chatterjee’s “aesthetic triad” model of how aesthetic experience is structured. What Barnes and Dewey arrived at through teaching practice, neuroaesthetics has more recently demonstrated experimentally: sustained, structured close looking, with intentional shifts between sensory, emotional, and interpretive modes, produces deeper aesthetic impact than passive viewing. The list above describes what it takes to support that pedagogy in a digital environment.

Two partnerships helped refine the thinking. LaPlaca Cohen ran an audience and strategy research project, delivering a Strategic Brief in September of 2022 that validated the three audience pillars we had been working from, K-12, Adult, and Institutional, and confirmed market signal for the institutional licensing model that had been part of our strategy from the start. Think Company took early hand drawn wireframes we put together and refined them through its UX practice, contributing ideas for technical solutions and design possibilities for the infinite-canvas interaction model that became central to the build. We brought our vision to the table and worked with both partners to refine them into Figma designs and a marketable minimum viable product.

In-House Design, Development and Testing

From there we steered the work internally. The platform was designed, built, tested, refined, and shipped by our three-person Web Development team, in close collaboration with our Adult Education and AV teams, our instructors, and our students. Design lived in Figma and engineering ran on two-week Agile sprints, with planning, design, development, testing, deployment, and review folded into a single cycle.

User testing happened alongside the build, not as a discrete research phase but as a continuous loop: prototypes tested with instructors teaching across contexts; AV staff supporting live delivery; iteration based on what surfaced in real classes. Bi-weekly sprint demos with stakeholders and release notes to users were published continuously, keeping a running channel for updates and known issues, and making feedback from instructors, staff, and students an expected input on the backlog rather than an occasional check-in.

The recurring themes from internal testing were consistent: reduce cognitive load during teaching; make roles explicit, instructor and student and support; prefer predictable controls over flexible ones; keep the visual field primary and everything else secondary.

An early and major structural choice was to move away from a slide metaphor and toward a shared canvas, one that could hold multiple works, preserve spatial relationships, and support fluid transitions between overview and detail. The interaction model that became Focus and Explore modes fell out of that decision: a shared canvas requiring no screenshare, enabled instructor-controlled focus and release to explore independently, and making sure that student exploration was intentional and pedagogically grounded.

The interface was derived from the temporal flow of a Barnes class rather than from conventional information architecture. The system’s primary job was to protect the instructor’s teaching rhythm. Discoverability was constrained in favor of focus, navigation was subordinated to sequencing, accessibility was treated as architectural rather than additive (keyboard navigation, predictable focus states, captions, and legible tool affordances shaped the core interaction patterns rather than getting bolted on at the end). By the time the platform was ready to ship, it met a practical and load-bearing definition of “minimum viable product”.

Initial Launch and Early Success

The first version of the VXP launched in 2023 for live, fully online classes. Its core interaction model, Focus and Explore, solved the fundamental problem: when the instructor focuses the group, everyone sees the same view of a work of art; when the instructor releases control, students can explore the image independently at full IIIF deep-zoom resolution. The transition between these states is legible and predictable for both roles.

Beneath that surface, the VXP is a custom web platform built on AWS. We use standards-based deep-zoom imaging (IIIF) for high-resolution close looking, integrated audio and video for live teaching, and a recording pipeline of our own design, which powers Replay. The platform has supported four board types since launch: high-resolution flat images, panoramas for 360-degree material, 3D models, and video.

Three pilot classes ran in February and March of 2023: Zooming in on Van Gogh (eighty-seven students), Color in Painting (forty-five students), and Impasto: Gesture and Texture in Modern Art (twenty-one students). By April, every Barnes online class was running on the platform. By the end of the year, thirty-two course topics, 344 hours of live content, and 1,292 students had moved through the VXP, taught by thirty instructors. Against a baseline of platform ratings between 3.9 and 4.7 out of five, by the fall of 2023, eighty-three per cent of students reported that their VXP class experience had been highly or extremely impactful on the way they thought critically about art, seven percentage points above non-VXP students.

Replay: A Research-Driven Build of an Asynchronous Mode

The live experience was a success, but at the time, a recording of a live class was a flat screen capture, the chat window and the instructor’s video baked into the frame, posted to YouTube. Students who could not attend a live session lost the deep-zoom canvas, the Focus and Explore interaction, the ability to navigate by chapter. The platform that made live close looking work was reduced, after the fact, to a static video.

Our research from 2022 had already named this as the next use case. Course evaluation questions added in November and December of 2023 confirmed it. Forty-two per cent of students were watching the legacy recordings anyway (n=84). Of those, forty-five per cent watched one, forty per cent watched two, and thirteen per cent watched four or more. The asynchronous use case was already real. The platform had not yet been built for it.

Replay was the answer, and what we built is not a screen recording. Replay preserves the interactive canvas itself. Students can pan and zoom the same deep-zoom images the instructor used live, enter Explore mode at any moment, and navigate by chapter or progress bar. The recording stays synced to the instructor’s view by default. The only thing absent is live interaction with instructors and fellow students. In certain ways, this makes Replay more flexible than the live class: a student can pause to look longer, jump back to a specific moment, or wander into the details at their own pace.

Replay was scoped as a research-led project. The research question, as it was posed at the time, ran like this:

How might we design an intuitive and user-friendly asynchronous interactive VXP experience that meets user expectations of the live in-class VXP experience, offers additional opportunities for independent exploration, and ensures accessibility and engagement for users of all ages, experience levels, and abilities?

The study ran in two phases between December 2023 and early January 2024.

Phase 1. Eleven thirty-minute sessions with Barnes staff. Participants spanned varying degrees of VXP familiarity: instructors, AV staff, members of the web team, evaluation staff, and colleagues who had never used the platform. The phase incorporated A/B testing of two patterns for the Focus and Explore indicator.

Phase 2. Eight thirty-minute sessions with members and students. Our colleagues on the Membership team identified longtime students, newer students, and non-students to invite, in order to capture a range of platform familiarity. We offered a Barnes online shop gift card in exchange for their participation.

The A/B test isolated the central design question: how does a student tell, at a glance, whether they are following the recorded instructor, or have unfollowed in order to explore independently?

• Path A: a Focus/Explore indicator nested in the existing toggle
• Path B: a switch button placed in the instructor video component

The research surfaced five recurring user types:

• The Multitasker. Plays the recording in the car or while making dinner. Mobile-first, sit-back posture. Wants a seamless, unobtrusive visual experience with maximum screen real estate and no distractions.
• The Note-Taker. An attentive learner. Uses a split-screen view, plays and pauses to take notes, navigates back to specific moments. Values simplicity.
• The Live Class Wannabe. Watches recordings as if they were live. Needs obvious “class mode” indication and interactive, immersive content.
• The Needs a Helping Hand. Low technical confidence. Hesitant to click “for fear of messing something up.” Needs clear and empowering prompts.
• The Explorer. A novelty user. Explores and unfollows class unprompted. Wants maximum screen real estate, minimal tools, and engaging learning beyond the traditional lecture format.

The personas drove the recommendation backlog:

1. Maximize screen real estate
2. Streamline recording navigation
3. Refine the Focus/Explore toggle design
4. Provide user-centered tutorials and instructional material

The research also settled the naming. Users tested “Replay” against the alternatives and consistently preferred it, especially when seen in context with the priming materials.

Replay launched on March 1, 2024.

The Replay development process is the clearest example of how the VXP product moves. Course evaluation data flagged the use case, structured user research, with internal and member participants, defined the design problem. A/B testing settled an interaction pattern, and the findings produced an explicit ticket backlog tied to four named commitments. An instructional plan ran in parallel with engineering toward a single launch date.

What’s Next

A subsequent post will cover what happened when another institution began teaching on the VXP, the partnership model we built around it, and the next-generation Admin experience we are developing, moving the platform from staff-supported to self-serve so more institutions can run it on their own.

Experience the VXP in this preview lesson, or take a Barnes online class as a student. For institutional inquiries about the platform and partnerships, contact us or visit our VXP overview page. Follow the Barnes Foundation on Medium for the rest of the series. This is the second post in a series on digital work at the Barnes. If you missed the first, start here.

Designing and Building a Platform for Close Looking was originally published in Barnes Foundation on Medium, where people are continuing the conversation by highlighting and responding to this story.