Familiar faces rendered strange: Why inconsistent realism drives characters into the uncanny valley

Area of Science:

• Cognitive Psychology
• Computer Graphics
• Human-Computer Interaction

Background:

• The uncanny valley describes the eerie feeling evoked by highly realistic, yet not perfectly human, computer-generated characters.
• Previous explanations focused on uncertainty about a character's humanness, but this fails to explain the specific eeriness.
• A new perspective is needed to understand the psychological mechanisms behind the uncanny valley phenomenon.

Purpose of the Study:

• To investigate the role of realism inconsistency in the uncanny valley effect.
• To propose and test the hypothesis that perceptual narrowing due to realism inconsistency causes eeriness.
• To differentiate the uncanny valley from general unfamiliarity responses.

Main Methods:

• Conducted an experiment with 365 participants rating objects, animals, and humans with manipulated realism consistency.
• Quantified categorical perception using a Bayesian model.
• Employed structural equation modeling to analyze relationships between realism inconsistency, familiarity, eeriness, and warmth.

Main Results:

• Reducing realism consistency significantly decreased familiarity and increased eeriness for animals and humans, but not for objects.
• Objective measurement of realism inconsistency in the visual cortex-inspired wavelet domain was established.
• Eerie feelings were elicited specifically by anthropomorphic characters when realism inconsistency reduced familiarity.

Conclusions:

• Realism inconsistency, not uncertainty, is the primary driver of the uncanny valley effect.
• Perceptual narrowing, triggered by inconsistent realism in anthropomorphic characters, leads to feelings of unfamiliarity and eeriness.
• Findings suggest that the uncanny valley is a specific response to perceived fakeness in human-like entities, prompting threat avoidance.