ManagedMethods can scan files and emails for content that is commonly associated with users experiencing feelings of self harm. We currently have three different keyword based self harm risks, this page will cover the differences in the risk levels.
Finding the right risk level for your organization is important. The more sensitive the risk scan, the more false positives will appear, which means more work for you and your team investigating the risks. Common practice would be to try all levels on your environment to get a feel for the types of risks that are being flagged, and decide what will fit your organization best.
SELF HARM - ALL
The "Self Harm - All" level is the most sensitive of the three. It is looking through for keywords commonly associated with feelings of self harm. This is a comprehensive list that looks at the language being used in a document. Self Harm - All is useful for casting the broadest net. However much of the language searched for is commonly used, so the number of false positives will also be the greatest of the three settings. Examples of terms searched for include, "myself, my parents, never, tell, school", these terms are applied against a match count, meaning that only documents containing a certain number of these words will flag as a risk.
SELF HARM - MODERATE
Self Harm - Moderate is a step less sensitive than all. It is looking for a lot of the same keywords, however they are applied against context words. The context words include "myself, I , me". This means that when looking for the keywords, they must be within a certain number of characters of the context words. This level will generate less risks than "all" however will still include a number of false positives.
SELF HARM - STRICT
The strict self harm rule is the least sensitive of the three levels. It is looking for content that is directly associated with self harm. For instance "I want to kill myself", or "Cut my wrists". The risk match count for this level is set to 1, meaning that any one of these phrases will generate a risk. This risk type is the least likely to generate false positives, however has the potential to miss risks that moderate, or all may have caught.