Affinity rises with human likeness until near-likeness triggers revulsion, then recovers only at full likeness.
Mori's 1970 observation. The dip is real and its cause is disputed, with category ambiguity the leading explanation — the figure is hard to classify as person or object, and the difficulty registers as unease.
In practice
- A cartoon face with two dots for eyes is likeable; the same character rendered with real skin and slightly wrong eyes is not.
- A near-human interface is harder to trust than an obviously mechanical one, which is why voice assistants that sound almost human are found unsettling.
- Possible explanations
- The leading explanation is category ambiguity: a figure that is hard to classify as person or machine is costly to process, and the difficulty registers as unease. Others are pathogen avoidance, since near-human faces can resemble illness, and mismatch, where a realistic face moves wrongly.
- In its favour
- Designers report the dip consistently, and several experiments find reduced liking at high but imperfect human likeness.
- Against it
- Measured curves vary with the stimuli and the rating scale, and a number of studies find no dip at all, so the shape may depend on which faces are chosen.
- Standing
- The dip is real but its cause is disputed; category ambiguity is the leading explanation.
- Further reading
- Wikipedia
- Subject
- Design & Systems
- Essay
- The limits every design runs into