It is basically hacking the algorithm. You seek to bias its output toward some other goals versus what the optimal math would give. One might argue that develop

It is basically hacking the algorithm. You seek to bias its output toward some other goals versus what the optimal math would give. One might argue that developing such algorithms for good intentions runs the risk of providing technologies for negative intents. But I think the research on the negative side is generally way ahead anyway, as security is so important.