Designing Interactive Science Visualizations for Exhibitions

An interactive science visualization in a museum or gallery has a different job from an analytical dashboard on a researcher’s desktop. Visitors may spend seconds rather than hours with it. They arrive with different levels of subject knowledge, may interact in groups and often encounter the display without instructions.

The Max Planck Research Networks project addressed this challenge with a multitouch installation that combined a collaboration network, maps and direct selection of institutes. Its design illustrates a broader principle: public science interfaces must support immediate orientation and deeper exploration at the same time.

Design for three levels of engagement

If the display only works after a tutorial, most casual visitors will never reach the data.

Make interaction visible

Touch interfaces remove the physical cues of a mouse and keyboard. Interactive elements therefore need visual affordances. A selected node should clearly change state; connected partners should highlight; maps should update in a way that makes the relationship between views obvious.

• Use selection states that remain visible after the finger moves away.
• Avoid small targets that require precision.
• Let users recover easily from zooming or panning too far.
• Keep gestures conventional where possible.
• Ensure that animation explains a transition rather than merely decorating it.

Preserve the data model

Public-facing visualizations are often simplified, but simplification should not distort the meaning of visual variables. If node size encodes publication count and line width encodes joint publications, those mappings should remain consistent through every interaction state.

A legend can be brief, but it must answer the core decoding questions: what is a node, what is a link, and what do size and thickness mean?

Use linked views instead of overloading one view

Scientific collaboration has both structural and geographic dimensions. Trying to place every variable in one network view quickly becomes unreadable. Linked panels allow each representation to do one job well.

Selection should propagate across views so the visitor experiences them as one system.

Design for group use

A large multitouch table may be used by several people simultaneously. That changes layout priorities. Labels should not assume one viewing orientation, and interactions should avoid unnecessarily resetting another visitor’s context.

True multi-user interaction is technically demanding, but even single-focus systems can support social exploration by making state changes legible to people standing around the display.

Performance is part of communication

Dense networks can contain thousands of objects. Slow response breaks the connection between action and result. An exhibition system should therefore be designed around the hardware budget and expected dataset size.

• Precompute expensive network metrics.
• Use level-of-detail strategies for labels and edges.
• Animate only transitions that help users understand change.
• Test with the final display resolution and touch hardware.
• Plan for unattended recovery after software or input errors.

Explain uncertainty without overwhelming visitors

A collaboration map is built from a defined dataset and methodology. A public interface does not need to show the full data-processing pipeline on the main screen, but it should provide a route to methodology information: time window, source database, meaning of a link and major limitations.

Measure success by questions, not screen time

The strongest exhibition visualization gives visitors a question they can answer through interaction: Which institutes collaborate? Where are their partners? What happens when I select this node? Which relationships are strongest?

When the interface makes those questions discoverable, interaction becomes part of scientific explanation rather than an attraction added on top of the data.