Average Replies per Thread

The metric average replies per thread is an indicator of an active discussion. It is proposed by Angeletou et al. (2011), Hacker et al. (2015) and Viol et al. (2016). It is of global scope as it is calculated for the entire network. The average replies per thread $at$ metric is calculated as an average for the network. The median is used instead of the mean, as some threads can have an unusual high count of responses.
1. select all posts grouped by threads
2. at := calculate median over count of posts

A high number of replies per thread characterises an engaging network with knowledge and ideas being discussed (Viol et al., 2016). Such a network thrives on active users, who initiate conversations and attract answers, leading to long discussion threads.

Hacker et al. (2015) argue that a high average value of replies per thread is an indicator for a lively discussion. Therefore in a network with a high average replies per thread, discussions take place. It indicates an engaging base of users, who interact and exchange ideas with each other.

Angeletou et al. (2011) describe a healthy network, in which there is high participation in topic discussions, that generate lots of replies. Discussions that are not driven forward, tend to die down, resulting in a large portion of unanswered threads. This can lead to a downwards spiral, in which the activity in the community decreases continuously.

A lively exchange of ideas and knowledge is a typical attribute of the closure theory. It is grounded in the creation of strong ties between the discussion partners facilitating collaboration and group work. Therefore, a high average replies per thread indicates Bonding Social Capital. A low average replies per thread indicates the lack of engaging discussions and community interactions and thus suggests a low level of Bonding Social Capital.