The ability to identify similarities between narratives has been argued to be central in human interactions. Previous work that sought to formalize this task has hypothesized that narrative similarity can be equated to the existence of a common summary between the narratives involved. We offer tangible psychological evidence in support of this hypothesis. Human participants in our empirical study were presented with triples of stories, and were asked to rate: (i) the degree of similarity between story A and story B; (ii) the appropriateness of story C as a summary of story A; (iii) the appropriateness of story C as a summary of story B. The story triples were selected systematically to span the space of their possible interrelations. Empirical evidence gathered from this study overwhelmingly supports the position that the higher the latter two ratings are, the higher the first rating also is. Thus, while this work does not purport to formally define either of the two tasks involved, it does argue that one can be meaningfully reduced to the other.
Identifying similarities among stories is a central part of the process of making sense of stories, and building machines for the latter task will presumably require some solution to the former. In this work we have provided overwhelming psychological evidence that the more appropriate a given story is as a common summary of two other stories, the more similar the latter two stories are to each other. The validity of this hypothesis offers a sufficient condition to test for similarity, or more precisely, offers a way to lower bound the degree of similarity. The condition is not, however, necessary, since the failure of a candidate summary to be an appropriate common summary of two stories does not indicate lack of similarity between the two stories, since some other candidate summary could exist that would be appropriate. Devising a method to produce candidate summaries that would be the most specific common summaries of two stories would offer the missing link to establish the necessity of the condition as well. The role of expectations in stories [18, 19] would seem to be important to that end.
The present study was a first step towards the confirmation of our hypothesis for a certain sample of the possible types of stories. Further research could examine the applicability of this hypothesis to other genres of stories. In a different direction, we could analyze stories extracted automatically from online sources, in order to avoid manually selecting specific types of stories, and any bias this choice may bring to the empirical study.
It would be interesting to generalize our hypothesis beyond stories, and to examine whether similarity between two concepts is effectively equivalent to saying that the two concepts share a common abstraction which is appropriate for both of them. Such concepts could be short videos, simple images, or sound clips. We believe that the empirical methodology developed herein, and the type of analysis performed, could be applied equally well to such more general settings.