Is This Film Review Algorithm Better than Ebert?

A new study stakes a film’s significance on how many times it’s cited in the work of other directors.

A new study out of Northwestern University suggests that an algorithm is as good or even better at identifying quality films than a movie critic. Using the film's "significance" as the key denominator, the group of scientists led by Professor Luís Amaral studied how many times a movie was referenced in another movie. Based on these results, the group was able to pick out the films with the most cultural impact, essentially placing the vote into the hands of fellow movie directors rather than critics.

Published in this week's issue of Proceedings of the National Academy of Sciences (PNAS), the study explains how the team of scientists broke down 15,425 US films by various metrics: critical reviews, awards, public opinion, citations and box office sales before finding citations to be the best signifier of quality. By comparing the results of each approach to the particular movie's inclusion in or exclusion from the National Film Registry of the U.S. Library of Congress, the team was able to narrow down its approach.

Keep Reading