Warning: This website has been archived and some features are not available anymore.

Reply Creation Ratio

The reply creation ration consists of two underlying metrics: the (1) first reply created ratio and the (2) last reply created ratio. They consider whether a user has made the either the first or the last reply in a comment-chain. These metrics indicate the tendency of the users to participate in discussions and are presented in one section. Hacker et al. (2015) and Viol et al. (2016) propose the metrics, which can be calculated on an ego-centric scope.

The calculation schema for the first reply ratio $fr$ and the last reply ratio $lr$ are similar. The mean is used instead of the median, as most users have a value of zero or one and no outliers are expected.

(1): 1. select all posts from a user
     2. foreach post:
          2a. retrieve the thread
          2b. check if it is the first reply
     3. divide the count of first replies by count of all replies:
        fr := first replies / all replies
(2): 1. select all posts from a user
     2. foreach post:
          2a. retrieve the thread
          2b. check if it is the last reply
     3. divide the count of last replies by count of all replies:
        lr := last replies / all replies

According to Hacker et al. (2015) a high ratio of first or last replies indicates a person, who is not engaging in discussions for an extended period of time. Instead a reply in the middle of the thread indicates an active discussion.

As a niche expert can answer a question with only one post, this metric characterises a niche expert as stated by Viol et al. (2016). Therefore low value for the first reply and last reply ratios indicate engaging users and thus is a measure of Bonding Social Capital for the individual.

Bonding Social Capital establishes a common ground for the userbase. The experts can freely share their knowledge and expertise with other users. Shared knowledge and regular communication are the basis for effective collaboration in an organisation (Riemer et al., 2015).