Skip to main content



Triadic Closure in Movies

http://www.cs.rpi.edu/~magdon/courses/casp/projects/Jin.pdf

Triadic Closure is often studied in research regarding social networks and relationships between individuals, where growing networks trend towards forming triangles to settle psychological and social tensions. However, at the Rensselaer Polytechnic Institute’s Department of Science, this property is being studied in the scope of the Internet Movie Database (IMDB) and the results are remarkably similar.

By simulating actors as nodes, two of these nodes are connected when the two actors are seen in the same movie. New edges have a default weight of 1 and are incremented by 1 each time the pair of actors shares screen time in an additional movie. In order to determine the cutoff point between strong and weak edge considering weight, the researchers observed a decreasing trend between the number of relationships and number of actors in the IMDB that have that many relationships, concluding that the drop-off point after 2 movies is significant enough to set this cutoff at 2. The study on triadic closure operates under the assumption that all edges are positive and simply considers strength of the edge. When analyzing the network created by IMDB data as compared to a network with a degree of randomness, the ‘clustering coefficient’ was 0.67 versus 0.25, indicating a naturally higher occurrence of triangles and therefore a higher instance of the triadic closure property.

The qualitative forces behind this phenomenon are similar to those in friendship networks – actors with similar qualities end up in smaller ‘clusters’ where opportunities tend to be shared more quickly and easily. Additionally, in the entertainment industry, certain styles or demographics of actors often appear on screen together simply because movies often include social groups consisting of similar people.

Comments

Leave a Reply

Blogging Calendar

September 2017
M T W T F S S
« Aug   Oct »
 123
45678910
11121314151617
18192021222324
252627282930  

Archives