This work presents a comprehension-based evaluation approach for computationally constructed stories containing failed actions, focusing on how readers understand plot lines generated by our HeadSpace system. We conducted empirical studies to measure reader comprehension of narratives where characters’ actions fail due to mistaken beliefs, moving beyond purely analytical evaluations of story generation systems. This was a key part of my PhD dissertation research, providing crucial validation that our sophisticated narrative planning algorithms actually produce stories that human readers can understand and engage with meaningfully.

Click here to access the paper.