Data-Driven Decoding of Russell's Circumplex Model of Affect

2026-06-15

AI Summary

This study addresses the long-standing challenge of verifying whether deep learning models in affective computing implicitly encode the psychologically grounded Russell circumplex model of emotion within their latent representations. By analyzing embeddings from RoBERTa, wav2vec 2.0, and multimodal Transformers on both the MSP-Podcast dataset and LLM-generated data, the work provides the first data-driven evidence that the circular topology of the Russell model naturally emerges in multimodal deep representations—without reliance on manual annotations. Experimental results demonstrate a high degree of topological alignment between multimodal fused embeddings and the canonical ordering of emotions in the Russell circumplex. Furthermore, generic text embeddings, even in a zero-shot setting, accurately map fine-grained emotion terms near human-annotated coordinates, thereby establishing a meaningful bridge between psychological theory and representation learning.

Abstract

Affective computing increasingly relies on deep learning to represent emotions, yet latent spaces often remain opaque, high-dimensional black boxes. This paper investigates whether Transformers' embeddings recover the geometric regularities of Russell's circumplex model. We unify two complementary experiments testing the hypothesis that, after training models on text and speech, their resulting latent spaces encode a topology consistent with valence-arousal and reproduce human-like neighborhood relations. Specifically, we evaluate deep representations extracted from Transformer-based text (RoBERTa) and speech (wav2vec 2.0) encoders, along with a multimodal Transformer fusion architecture, across naturalistic datasets like MSP-Podcast and controlled LLM-generated stimuli. Our analysis reveals that multimodal fusion of text and audio yields perfect topological alignment with Russell's primary emotion ordering. Furthermore, in a zero-shot setting using generic text embeddings, projected fine-grained emotion terms fall close to their established human-mapped coordinates. Our contribution is a novel, data-driven framework for validating emotion models, demonstrating that Russell's circumplex structure is intrinsically encoded in the embeddings of these modalities rather than being solely an artifact of human labeling, thereby bridging the gap between psychological theory and representation learning.

Problem

• Russell's circumplex model
• affective computing
• latent space
• emotion representation
• multimodal fusion

Innovation

• Russell's circumplex model
• multimodal fusion
• Transformer embeddings
• affective computing
• topological alignment