Risk management
Overview
The risk management suite is a set of features tailored to game community operations. Its main goal is to help community operators quickly locate content that poses risks and send alerts about it.
Risk management trends
When you open the Risk Management module, you see the overall trend of recent community risk management and the latest flagged content, which gives you an overview of the risk situation. Click View All under the latest flagged content section to open the risk management details page and view the specific content. You can also view and configure risk management alerts in the alert area at the top.
Risk management details
Click View All to open the risk management details page. In this list, you can see all abnormal content that triggered risk detection within the selected period, including the risk type it triggered, its risk level, and the reason it was identified as violating content.
Click View Details in the actions column to view the details of the selected abnormal content. If you find that a piece of content is a false positive, click Remove in the actions column; the content is then no longer treated as abnormal. In later versions, the system will use the content you remove manually to improve the accuracy of risk detection. Use Bulk Remove on the right of the filter bar to remove multiple items at once: select the items, then click Remove Selected (N).
In the upper-right corner of the list, there are also a refresh button and a Run Manually button: click Run Manually and select a date to run risk detection again on historical content. Finally, click Rule Details in the upper-right corner of the page to open the risk management rules page.
Risk management rules
On the details list page, click Rule Details in the upper-right corner to open the risk management rules page. This page shows the risk detection rules currently in effect, that is, the LLM prompts used by the risk detection task. The upper-right corner of the page has two buttons, Run Manually and Edit Rules. To edit the risk management rules, click Edit Rules to enter edit mode.
On the rule editing page, you can edit the rules for the five major risk types. Rules are divided into detection rules and exception rules: detection rules describe which content should be identified as violating content of that risk type, while exception rules describe which content should not be identified as violating. Adjust them according to your game and community guidelines.
If you have adjusted a rule but want to undo the changes, click Revert to reset it to its state before editing; after you make changes, a Revert button also appears in the header of the editing page, which reverts the rules of the whole page at once. Detection rules are required, while exception rules are optional. After you finish modifying all the rules, click Save to make the new risk management rules take effect. New data the system receives afterward is evaluated against the new rules.
-
Custom risk management rules are supported
- If a preset risk management rule name does not meet your needs, edit the name to change it.
- Add custom rules as needed. You can have up to 10 risk management rules.
Risk management alerts
On the community risk management landing page, you can view and enable risk management alerts. The entry for configuring risk management alerts is shown in the following figure:
In Alert Settings, you can configure the Monitoring Frequency, Alert Schedule, Notification Methods, and more.
- Monitoring Frequency
Hourly and daily monitoring are currently supported. At the specified time (for example, 20:46 every day in the figure below), the system counts the new flagged content from the previous period. If there is new flagged content in the previous period, it tries to send an alert.
- Alert Schedule
The time range during which alert notifications can be sent. Alert tasks within the time range send notifications immediately when needed. Alert tasks outside the time range still run but do not send messages; the alert is sent only when the next alert task within the time range is triggered, and all previously unsent alert data is combined into a single alert message.
For example, for an alert monitoring task that runs at 20:46 every day with the allowed period set to Monday through Friday, alerts from Monday to Friday are sent immediately after they are triggered, while alerts on Saturday and Sunday are deferred and sent together at Monday's alert task time. The data sent is the total new flagged content from last Saturday to Monday.
- Notification Methods
The target that messages are sent to. Alerts can currently be sent through DingTalk Chatbot, WeCom Chatbot, and Feishu Bot. You can configure bots for multiple platforms, but only one bot per platform. To learn how to configure each platform's bot, click Setup Instructions at the end of the row to view that platform's bot guide. After you finish configuring the bot on the platform, enter the bot's webhook URL under the notification method to complete the notification configuration.
- Disable alerts
After you enable alerts, you can disable them in the following ways to stop receiving alert notifications (the risk detection task still runs and is not affected). First, once alerts are enabled, hover over the alert menu entry and click the close button next to it to disable alerts.
Alternatively, go to the alert settings page and click Disable Alerts at the bottom to disable risk management alerts.
After you disable alerts, the previous configuration is kept, so you can enable alerts again at any time without reconfiguring them.