Bias-Adjusted LLM Agents for Human-Like Decision-Making via Behavioral Economics