1

Say that I had upvotes and downvotes for a document. I'm trying to come up with a good "controversialness" measure from this. When both upvotes and downvotes are large, then this measure should be large.

This is my very naive attempt at doing this:

Agreement = Upvotes/(Upvotes + Downvotes)

Disagreement = Downvotes/(Downvotes + Upvotes)

Controversialness = ABS(Agreement - Disagreement)

Does anyone have an idea on how I could measure the controversialness of a post?

Community
  • 1
  • If you have this information as a time series, e.g., votes per week or month, then the variability and diversity of the sentiments over time could form a metric for "controversy." There are many metrics for diversity, as a google search will show you. One good one is the coefficient of variation defined as the std deviation of the sentiments divided by their mean. – user78229 Apr 04 '16 at 22:06
  • In addition to DJonhson's notes on variability, a good indicator of controversy is seeing a bimodal distribution, with peaks at "agreement" and "disagreement" having similar heights. – Alex R. Apr 04 '16 at 22:10
  • This link is to a discussion on linguistic "weirdness," a highly relevant topic area... http://languagelog.ldc.upenn.edu/nll/?p=5006 – user78229 Apr 05 '16 at 00:27
  • 1
    I'm missing something. If you had a billion upvotes and a billion downvotes, your Agreement and Disagreement would both be 1/2, and your controversialness would be zero. I'm pretty sure that's not what you intended? –  Apr 05 '16 at 02:31
  • Yeah - I couldn't think how to handle that. –  Jul 30 '16 at 16:18

0 Answers0