The only proxy we have for measuring self-awareness is "responds in a way that maps to human behaviour in X scenario"—e.g. things like the mirror test—which, while it's the best we've got, isn't very compelling given the diversity we observe even in primates.

Without any convincing evidence in either direction, I find it perplexing that people seem to default to asumming the negative (i.e. the more convenient assumption)