The concept of probably approximately correct (PAC) learning has been foundational in computational learning theory.
Psychology must be a science of observation, not speculation.
The most rewarding part of science is seeing your work help others.