Reinforcement Learning vs Retraining

I have created a complex ML model using supervised learning. For the sake of discussion, let's say my model identifies dogs and a human labels the output as correct or not correct

Now, I want to improve my model. I would like to understand when to use Reinforcement Learning and when to use Retraining.

Approach 1 (Reinforcement Learning) - My current understanding is that in reinforcement learning, we create an agent with the goal of maximizing reward. In my example, the goal could be maximize the number of correct predictions of dogs and use the reinforecement learning to improve the model. This is one way to improve the prediction quality of the model.

Approach 2 (Retraining) - Another way is that I collect all the instances of correct and not correct labeled by the human and add them to my training data to retrain the model using the larger training dataset. Assuming the labels are correct, this approach would also improve the prediction quality of model.