Analysis of Our Own
Archive of Our Own (AO3) is a nonprofit, open-source fanfiction repository that has been in continuous operation for nearly 20 years.
In March 2021, the site administrators allowed their userbase to study the data they had collected over many years. Let’s see what we can find in this data!
First, we import our library and data, print column names and data types using the print (df1.dtypes) command to discover what types of data are accessible.
Column Names and Data Types:
creation date object language object restricted bool complete bool word_count float64 tags object
Frances Jones Bannerman, In the Conservatory, 1883
There is a common misconception that fanfiction is an inferior form of writing, and that both reading and writing it are associated with being immature, self-indulgent, cheap, and unimaginative because the field itself is derivative.
If this were true, it would be reasonable to assume that fanfiction writers are more likely to drop their ongoing projects, scattering them across different areas of fandom. Does our database support this idea?
Alas, most writers using AO3 do tend to finish their creations. In their defense, they had over 10 years!
Let’s see if we can see any spikes in the very first years of the site running?
It’s very hard for me to define the difference between 0,5% and slightly above 1% as significant.
Based on this stat alone, it’s fair to say that fanfiction writers seem to be quite determined and well organized.
Perhaps such personality traits can be defined by the length of their works? Let’s divide all the fiction by word count and calculate the percentage of the total number of works presented on the site.
Finally, I would like to see if the idea that fanfiction is self-indulgent and mostly being erotica for women is supported by the evidence.
However, the percentage I got from this analysis shows that only about 5% of the works were rated «R,» which includes not only eroticism but also graphic violence and excessive swearing, and the trend has been declining over the years, barely exceeding 1% in 2021.
Study, Frank Moss Bennett, 1874-1952
At the end of the day, judging purely from the data I’ve analyzed on ao3, writer trends don’t seem to vary that much over the years, and seem to be pretty set and stable.
For this project, I have used: AO3 Database for analysis, ChatGPT-4 to write and edit the code, DeepLWrite to proofread and edit the writing.
ipynb file: https://drive.google.com/file/d/1-AzyXPzX-73dSRwXwPmK9bFFNJdKWEG1/view?usp=sharing
