Comprehensibility is an evaluation metric that acts as a constraint to ensure automatically generated metaphors remain meaningful. This metric is particularly crucial for metaphors compared to similes, as understanding metaphors often requires specific background knowledge or contextual information.