Selection Algorithm

The selection algorithm is an algorithm that was created to optimally select the single most impactful tweet that from a list of tweets about a specific debunked claim for the DebunkBot to respond to.

Run Intervals

We currently set the run intervals to 2 hours, so that we won’t bump into other operations running and we can easily fly under our api rate limit.

Metrics:

• created_at
• quote_count
• retweet_count
• reply_count
• verified [user]
• followers_count [user]

Steps:

• Assign weight to each of the metric we’ll be using in the first phase depending on how we think each contributes to spread. (Reason for adding a weight to each tweet is because, somethings contribute more to spread than the others so, even if the count on them is not as much as others, their effect might be more. For instance, retweeting a tweet spreads that tweet faster than responding to it. The weight basically means we’re taking into account what each metric contributes to the tweet going viral)
  • With this, we use the following weight, depending on how much factor each metric adds.
    • retweet_count: 4
    • quote_count: 3
    • replies_count: 3
    • favorite_count: 3
    • verified: 1
    • followers_count: 1
• Calculate a score for each tweet based on the weight for each metric available
  • This will basically be multiplying each count by its weight then adding those all up together.
• Get the tweet with the most score and return as our pick
• If there is a tie, with the highest score
• Check which tweet was created earlier of the tie and return that as our winner because there’s a possibility it’s been seen more than the newer one.