Antisocial Behaviour in Online Discussion Communities
User-generated
content is critical to the success of any online platforms. Sites like Facebook,
Stackoverflow engage their users by allowing them to contribute and discuss
content, strengthening their sense of ownership and loyalty. While most users
tend to be civil, others may engage in antisocial behaviour, negatively
affecting other users and harming the community. Many platforms implement
mechanisms designed to discourage antisocial behaviour. These include community
moderation, up- and down-voting and the ability to report posts, mute
functionality, and more drastically, completely blocking users’ ability to
post.
Antisocial
behaviour is a significant problem that can result in offline harassment and
threats of violence. This motivates us to address several issues related to antisocial
behaviour:
- The instant of inception of this behaviour with respect to
the lifetime of the user on online discussion forums.
- The role of community on the behaviour gradient of
antisocial users.
- Feasibility of an effective prediction mechanism to identify
such users early on.
In the rest of this
text we refer to Future-Banned Users as FBUs and the Never-Banned Users as NBUs.
Data was collected from three online discussion forums viz. Brietbart, IGN, CNN
and the presented observations are with respect to this data. In a broad way it
is observed that FBUs tend to write less similarly to other users, and their
posts are harder to understand. They use less positive words and use more
profanity as seen in Figure1 (a), (b) and (c). They receive more replies than
average users, suggesting that they might be successful in luring others into
fruitless, time-consuming discussions. The behaviour of FBUs worsen over
their active tenure in a community. Communities may play a part in incubating
antisocial behaviour. On the other hand, while communities appear initially
forgiving, they become less tolerant of such users the longer they remain in a
community. This results in an increased rate at which their posts are deleted,
even after controlling for post quality.
A user’s posting
behaviour can be used to make predictions about who will be banned in the
future. With the help of features that capture various aspects of antisocial
behaviour: post content, user activity, community response, and the actions of
community moderators. We find that we can predict with over 80% AUC (Area under
ROC curve) whether a user will be subsequently banned. The features indicative
of antisocial behaviour that we discover are not community-specific.
On average, the
deletion rate of an FBU’s posts tends to increase over their life in a
community. In contrast, the deletion rate for NBUs remains relatively constant
as shown in Figure 2. The increase in the post deletion rate could have two
causes: (H1) a decrease in posting quality— that FBUs tend to write worse later
in their life; or, (H2) an increase in community bias — that the community
starts to recognize these users over time and becomes less tolerant of their
behaviour, thus penalizing them more heavily. Further it is observed that while
both FBUs and NBUs write worse over time, this change in quality is larger for
FBUs.
The goal to be
achieved here is building tools that could allow for automatic, early
identification of users who are likely to be banned in the future. The developed
methodology can accurately differentiate FBUs from NBUs with only a user’s first
ten posts. By finding these users more quickly, community moderators may be able
to more effectively police these communities. State of the art
mechanisms result in misidentifying one of five users as antisocial. Whereas
trading off overall performance for higher precision and have a human moderator
approve any bans is one way to avoid incorrectly blocking innocent users, a
better response may instead involve giving antisocial users a chance to redeem
themselves.
Further Reading:
https://cs.stanford.edu/people/jure/pubs/trolls-icwsm15.pdf
No comments:
Post a Comment