This thesis explores the use of deep reinforcement learning (DRL) to enhance dynamic option hedging by incorporating forward-looking market information, mitigating speculation, and optimizing portfolio rebalancing frequency. The first paper, Enhancing Deep Hedging of Options with Implied Volatility Surface Feedback Information, introduces a DRL-based hedging framework that leverages implied volatility surface data, improving hedging performance over traditional methods. The second paper, Is the Difference between Deep Hedging and Delta Hedging a Statistical Arbitrage?, examines whether deep hedging introduces speculative behavior in incomplete markets, demonstrating that proper risk measure selection prevents unwanted speculation. The third paper, Implied-Volatility-Surface-Informed Deep Hedging with Options, extends deep hedging by integrating implied volatility surface-informed decisions, no-trade regions, and multiple hedging instruments, improving cost efficiency and adaptability. This research contributes by defining frameworks that enhance existing techniques for managing risk in financial markets.