One inherent problem for the Safety Team to solve is the constant tension between giving user reports the time and attention they deserve versus taking action to remove problems from the platform. The team obviously has limited resources to review reports. If they don't review and act on reports fast enough, the platform gets worse. But if they act too fast, they risk making mistakes. Either they miss problems, or they mistakenly take enforcement actions against users who did nothing wrong. So it's difficult to get this right without just spending more money to increase Safety Team resources, and I don't know how practical that is.

One option, if the user who is the subject of a report is active in some community, is to ask the moderators of that community (including the broadcaster) to give an opinion about the alleged policy violation. Moderators are normally in the business of enforcing the standards of their own communities, and so should be able to give useful opinions.

For example in my case, I was indefinitely suspended for having an unacceptable username (for a different account). Presumably what happened was that I spoke in somebody's chat, some other user pressed the "report user" button and complained about my username. At this point, the report would presumably be sent to the Safety Team, who would be expected to act on it.

As an aside: Some others on this forum have said that these reports are being handled by AI and that this doesn't work. I don't know if this is true or not, but it doesn't fundamentally change the analysis. The Safety Team must spend resources to evaluate reports in order to act on them in a timely manner. Whether their accuracy suffers because human reviewers have too many reports to review in a given time or because the AI isn't good enough doesn't really matter. The system as a whole is under pressure to review reports faster than it is able to do so accurately. Ideally, it would like some way to increase accuracy without just spending more resources, whether that means hiring more reviewers or paying for a more expensive AI or some combination of the two.

Suppose instead of sending the report directly to the safety team, it was first sent as an alert to the channel moderators. The alert would say something like "A user in your channel was reported for the safety violation UNACCEPTABLE USERNAME. Here are the details. Do you think this report is accurate? Do you want to ban this user from your channel?"

This could arguably put the Safety Team in a position where they are able to work slower. If there's an immediate issue, you prod the moderators to solve it immediately by banning the user. You also get the moderators to give an opinion, which makes the Safety Team's job easier since they don't have to do anything from scratch.

More importantly, this allows for a more fine-grained approach to safety. For example, suppose in one channel a user reports something (like an inappropriate username), and the moderators agree and ban the offender, but in another channel the same report is made but the moderators disagree and decline to ban the offender. That might mean that, for example, the users in the first channel skew younger and more immature, but the second channel has a more mature audience and thus there is no safety concern for them. In my opinion, this is a desirable outcome. If some conduct makes users in one community feel unsafe, but does not trouble users in another community, the ideal result is to ban the conduct from the one community but not the other. The issue need only be considered by the Safety Team if this approach is not working for whatever reason. This means that the Safety Team is able to move more slowly, and thus is able to be more accurate.

The Safety Team does not necessarily have to give any credit to the opinions of moderators. Just because a moderator responded that a report was not accurate or not an issue, the Safety Team could still ignore this and take enforcement action. But just asking for moderator opinions could make the job easier. And of course if it were discovered that 10 moderator agreements and 0 moderator disagreements in response to these requests for opinions accurately predicted that the Safety Team would choose to take enforcement action, or at least would agree in retrospect that this report was a high-priority-to-review-asap, that might lead to opportunities to optimize the whole report process.

Possibly, this sort of feature would not be appropriate for all kinds of user reports. But I think it could be good for some of them. And there's no reason such a system couldn't be rolled out incrementally for some kinds of reports but not others.