With the foundations of agent architecture, communication, and workflow orchestration in place, this chapter focuses on the reasoning and decision-making processes that drive agent behavior. We begin by examining how individual LLM agents can perform complex reasoning tasks. Subsequently, we will address how these individual capabilities can be extended to enable groups of agents to reason collectively and make coordinated decisions to solve shared problems.

This chapter will guide you through techniques for individual agent inferencing, including methods like ReAct and Chain-of-Thought. You will then learn about aggregating knowledge for collective reasoning and approaches to distributed problem resolution. We will also cover strategic agent interactions using concepts from game theory, introduce Multi-Agent Reinforcement Learning (MARL) for developing coordinated behaviors, and examine how agents can learn adaptively. The chapter concludes by considering agent mental state representations, such as the Belief-Desire-Intention (BDI) model.