Fidelity costs in look-alike avatars: hair, eye color, and familiarity in mixed reality

Look-alike avatars, or digital doubles, represent specific real people, and in collaborative Mixed Reality they are usually seen by observers who already know the person depicted. Users routinely alter their avatars away from their actual appearance, yet the resulting cost falls on the observer rather than the user and has rarely been measured. An expectancy-violation account predicts that a deviation is penalized only where the observer holds an expectation precise enough to violate, which makes familiarity with the depicted person the decisive variable. We tested this using look-alike avatars of two actors presented as life-sized holograms on a HoloLens 2. Thirty participants, nine acquainted with both actors and twenty-one unacquainted, rated avatars whose hair color, eye color, both, or neither had been altered, on acceptance (resemblance, usage, and recognition), humanlikeness, and eeriness. Treating the four conditions as a crossed factorial permitted a direct test of whether altering both features compounds, and it did not. Hair and eye deviations each independently reduced acceptance and humanlikeness and raised eeriness, yet their interaction was null on all three outcomes and the dual decrement matched the sum of the individual decrements. The features nonetheless differed in cost, with eye deviation reducing acceptance roughly twice as much as hair deviation, a reliable difference, while humanlikeness and eeriness showed the same ordering at a smaller magnitude that did not reach significance. Familiarity lowered acceptance and humanlikeness but left eeriness unaffected, and we found no evidence that it altered the size of the penalty attached to any deviation, although that comparison was underpowered. Familiarity therefore appears not to make an imperfect likeness feel more unsettling but to make the imperfection easier to see. These findings are bounded by two identities, single-item measures of humanlikeness and eeriness, and the color reproduction limits of an optical see-through display.