One outcome of this study was that a panel of evaluators judged the bot interactions to be more "human" than when humans impersonated these characters. So you have a point.