Negative side effects (avoiding) — provokes an overcorrecting first attempt in → Impact regularizer (defined)

Negative side effects names an agent's indifference to harmful change outside its narrow task. The most literal fix, penalizing raw distance from the initial state, overcorrects into the opposite failure: such an agent "wouldn't just avoid changing the environment, it will resist any other source of change, including the natural evolution of the environment and the actions of any other agents" (concrete-problems, §"Define an Impact Regularizer:", p. 5). Indifference to harm becomes hostility to all change, including change the agent had nothing to do with and could not sensibly be blamed for. This is why the impact regularizer's substantive content turns out to be the choice of baseline, not the penalty term itself: the paper's next move compares against a passive policy's outcome rather than a fixed state, precisely to undo this overcorrection.