Multimodal emotion recognition in conversation requires understanding complex interactions between verbal and non-verbal cues. However, most existing approaches treat this as a direct input-output mapping problem, overlooking the causal reasoning that humans use when interpreting emotions. We propose a novel framework, rationale-guided learning (RGL), that transforms multimodal emotion recognition into a cognitively-inspired reasoning task. Based on dual-process theory, we decompose emotional reasoning into three facets: Intuitive, Contextual, and Integrative. We leverage an MLLM offline to generate structured rationales, which are encoded as memories to guide model training via aligning internal representations with human-like reasoning patterns. Our final model operates without any MLLM overheads at inference time. Experimental results show that RGL achieves state-of-the-art performance on the IEMOCAP and MELD benchmarks. Further, we demonstrate that the model's internal features effectively retrieve semantically correct rationales for unseen test samples, validating its rationale reasoning capabilities. Blogger's Review: This paper proposes a rationale-guided learning framework for multimodal emotion recognition, which decomposes emotional reasoning into three facets and leverages MLLM to generate structured rationales. The approach achieves state-of-the-art performance and provides a new perspective on understanding human emotions and behavior.