Empirical Evidence of the Limits of Automatic Assessment of Fictional Ideation