Interference: visual steering biases one of the choices
Aliases: visual interference · misdirection · salience manipulation
What it is
Interference steers the user's choice toward the operator's preferred option using visual and attentional manipulation: contrast, position, motion, and preset focus push the compliant option to the front of perception and push the refusal option out of the attentional range. Among the dark-pattern classes it is the one that leans hardest on perceptual mechanics — it deletes no option and changes no option's existence, only the option's salience.
Why it happens
Visual salience decides attention allocation, and attention allocation decides the real comparison set: users compare not "the options on the page" but "the options pushed in front of them." The interference toolbox is a list of perceptual levers — a contrast gap that makes one button the sole focal point, motion that captures exogenous attention, preset focus that lands the Enter key on "accept," scroll position that puts the crucial term below the fold. The boundary against legitimate visual design runs along direction: design always steers attention, but legitimate steering serves the user's task (the primary action is obvious), while interference serves a choice against the user's interest (the option that hurts the user is made inconspicuous). Judgement is therefore an empirical question — measure the depth-of-processing gap between the mutually exclusive options, not argue about visual taste.
Studying it
The taxonomic evidence comes from large-scale content analysis: systematic crawls and manual coding of tens of thousands of shopping sites established this class framework and the frequency distribution of its tactics. Mechanism evidence comes from eye-tracking studies of attentional shifts under interfering layouts and controlled experiments comparing symmetric versus interfering designs on selection rates. Common dependent variables: the selection-rate gap, decision time, and post-hoc recall (can users remember what they declined?). Methodological caution: the lab can establish the direction of the bias, but "how much bias counts as manipulation" has no natural threshold — classify by the severity of the decision's consequences, not by p-values alone.
Where it stops holding
Interference judgement depends on task context: making the primary button bright is ordinary design; making the decline button the same colour as the background is interference — between the two ends lies a continuum, single elements cannot settle it, and only the relative treatment of the option pair counts. Accessibility contrast standards point the same direction as anti-interference but from different motives: the standard governs "can be seen," anti-interference governs "the degree of visibility is not manipulated." Layout conventions (a dialog's close button top-right) are themselves neutral; judgements must exclude convention effects or they will punish ordinary design.
Applying it
- Option-symmetry review: score every mutually exclusive pair (accept/decline, subscribe/skip) side by side on three dimensions — visual weight, path length, copy clarity; significant asymmetry is an interference suspect.
- Set anti-interference ceilings: write numeric limits on contrast, size, and path-length gaps between same-level exclusive options into the design system.
- Verification: A/B a symmetric version against the live version and compare refusal rates — a redesign after which refusals drop significantly is the evidence of interference; let the data testify rather than the intent.