Mr. Ronan Costello:

That is no problem. I can jump in there. I will address content moderation in general given that it was the first issue the Deputy raised. I included an annexe in my opening statement that addresses some of this and goes to the heart of one of the things that Twitter wants to emphasise, in that we have been much more proactive in the past four or five years in reducing the burden on users to feel that all content moderation is initiated by them reporting content to us and that that is always the starting point of content moderation. That is not the case at all any more and hopefully into the future that burden on users will reduce further. At the moment, 68% of the content that we remove for violating Twitter's rules on abuse is proactively surfaced by machine learning. This goes to a key point, which is that the most effective and scaleable way to do content moderation now and into the future – although user reports are always helpful from a contextual point of view – is to deploy machine learning in such a way that it gets better and better at identifying abusive behaviours. Abusive behaviours are common to all accounts, they are spamming a particular person or they are spamming a hashtag to try to get a topic trending or something like that. The better we can be at deploying machine learning and looking at abusive behaviours, the less the burden there will be on a user to report things to us and hopefully then the better their experience will be on the platform.