I've recently been exploring Reinforcement Learning (RL) methods in my work and the exploration-exploitation dilemma is always mentioned. It almost feels like the exploration-exploitation dilemma stemmed from RL research. However, when we think about Simulated Annealing (SA) and Genetic Algorithms (GA), isn't there also the notion of exploration and exploitation? Specifically, in SA, you take a random step (explore) with some probability and an improving step otherwise (exploit). In GA, you mutate (explore) with some probability and select good parents for crossover (exploit).
My question: which paper first mentioned this? It feels as though https://pubsonline.informs.org/doi/abs/10.1287/orsc.2.1.71 is when it was first mentioned but I'd like expert opinions on this.