I think it's related to the exploration exploitation tradeoff in reinforcement learning. Creativity is related to exploration and randomness (and a high learning rate) while exploitation to using past knowledge to extract the maximum benefit. The brain switches off the creative/learning part in order to maximise the latter. In reality we need a mix of both, exploration is necessary for improvement and exploiting past knowledge is necessary for achieving high rewards. In neural nets, a too high or too low learning rate would not lead you to the optimum, you need to find the exact sweet spot. Anxiety would turn the balance too much into exploitation mode when we need to be creative.