Fidelity spans visual, content, and interaction dimensions

What it is

Fidelity is not a single axis from sketch to ship. One artifact can look release-ready while carrying placeholder copy; another can be a wireframe wired to live records; a third can look crude yet implement a full state machine. Later work calls this mixed-fidelity: how it looks, what it contains, and how it behaves are independently tunable. Collapsing the three into a single “high/low fidelity” label mixes three kinds of evidence into one verdict.

Why it happens

The dimensions constrain different perception and behavior. Visual fidelity supplies finish and brand cues, so attention and aesthetic judgment follow pixels. Content fidelity decides whether people can read, search, and verify with real vocabulary; dummy strings evaporate those actions. Interaction fidelity decides whether timing, error recovery, and state understanding can be observed, because those depend on transitions rather than stills. Evaluators rarely split comments by dimension on their own; whichever dimension is raised attracts the talk. The engineering implication is to make real only the dimension the current question needs, and keep the rest disposable.

Studying it

Comparative studies treat fidelity as separable factors, not a total score. A typical design holds task and script constant, varies visual rendering, data realism, or clickable depth, and records success, time, problem counts, and comment themes. Walker, Takayama, and Landay’s paper-versus-computer comparison is useful less for “which is better” than for whether error types remain comparable. McCurdy and colleagues showed that one prototype can keep navigation highly interactive while visuals stay low-fidelity. Coding feedback into visual, content, and interaction bins is what reveals where the evidence actually landed.

Where it stops holding

The three-way split comes from screen-based evaluation. Material and scale in physical products, or rhythm and turn-taking in speech interfaces, do not map onto visual–content–interaction. In games, visualization, or brand-critical surfaces, the look is part of the interaction problem; suppressing visual fidelity changes the task. Looking finished in the lab also is not engineering completion: motion, performance, and accessibility are often still fake.

Applying it

• Label the current prototype on visual, content, and interaction separately; do not write only “medium fidelity.”
• Raise only the dimension that must yield evidence; mark the rest as undecided with wireframes, placeholders, or stills.
• Log review comments in three columns so visual remarks do not count as flow findings.
• After the session, if most comments landed on a dimension you were not asking about, retune the prototype and retest rather than treating the comments as requirements.

Related

• Same group: Q5.01.2 Match fidelity to the question currently being asked · Q5.01.3 High fidelity suppresses participant criticism
• Adjacent: Q5.07 How prototypes mislead · Q5.10 Prototype misleadingness and over-polishing · Q5.11 Realism of prototype data
• Search terms: visual-content-interaction fidelity · mixed-fidelity · prototype fidelity