Mock-up for Moderation Queue of Reported Chat
- Status: Pending
- Preis: $500
- Einträge eingereicht: 13
- Gewinner: H8888
Kurzbeschreibung des Wettbewerbs
GOAL: Create a layout that allows a paid human moderator to very quickly at a glance get all the key information they need to make a decision if a user is breaking the community guidelines by being abuse and should be kicked out of the game for 24 hours. The information is supported by AI hints that highlight what is important.
BACKGROUND: Everyday millions of users play the games we help moderate. Sadly sometimes users abuse other users, who in turn file an abuse report. Hundreds of humans are then paid to look at these reports and to try to determine if it violates the rules and if the offending user should be kicked out of the game for 24 hours. Unfortunately about 80% of reports are useless. People just press the button to see what happens or they may be over-reacting. Other times the report is severe including abuse, physical threats, cyber-bullying, grooming. To resolve that we use AI to highlight what is important such as we suggest what the report is about, what action they should likely take, what is the ongoing behaviour of the actors and what did they say that is really bad in context of everything they said in that period.
The task is to create a highly optimized layout for our paid staff to go through reports quickly and accurately. The way the system works is a game user will file a report (not in scope), the report will be added to a database of waiting items, a moderator will click on a link to get the next most important report (not in scope) then the moderator will see the report (in scope) and choose the appropriate action (in scope), then the next report will show up.
Key Page Elements
- Action Buttons (Ignore, Ban)
- What language it is in
- What topic is predicted
- The reason the user reported the other user
- The actors (the person who was reported, the offended users, other chatters in the room)
- The chat logs
- What language was detected (inform the AI if it got it wrong)
- Most likely topics it is about (ex. sextortion)
- What action other moderators typically took in this case (recommend Ignore by 92%)
- Each actors personality type (ex. troll)
- Highlights the most important group of words
Preference for design
- The most important elements should be above the scroll. Scrolling makes the task really slow. Rarely used details like the full chat log could require scrolling.
- The information need to stay in the same location. Moderators spend 40 hours a week in the same layout so can use their subconcious to notice small changes when they are always in the same spot (for example, user1 is always the reported user, the label or icon "troll" is always the same, if it becomes "narcissist" then they will see that subconciously since the space changed.
- 1 Page Mockup
- Detailed Requirements attached.