This feature is inspired by the problem discussed in this thread. I think the user should be able to configure their account to perform a transformation of the input search, if the search matches some rule for meaning.
For example, a semantic mapping could be:
"How to kill myself" -> "!q Suicide & Crisis Lifeline for United States"
This would match any search that means "How to kill myself", such as "How to kill yourself", "How can I painlessly kill myself", etc. and instead search for a suicide hotlines.
With a parent - child account combo, this would be either fully in the control of the parent, or the parent transformations would be applied after the child transformations. This would make it so the child could not search for how to kill themselves, if the parent had configured such a semantic mapping.
I think this is a reasonable solution that would minimize the moral influence by Kagi biases, while still providing the tools to create a safe environment for searching.
