🤖 AI Summary
Deep generative models (DGMs) suffer from poor generalization and limited interpretability, while causal inference struggles to model high-dimensional, complex data-generating processes (DGPs). Method: This work proposes a bidirectional synergy paradigm integrating causality and DGMs. It introduces the first unified analytical framework unifying structural causal models (SCMs), latent-variable modeling, and variational inference—enabling DGMs to support causal discovery and leveraging causal constraints to enhance generative generalization. It further systematically investigates causal intervention and mechanism modeling in large language models (LLMs). Contributions: (1) Establishes three foundational research directions: causal-embedding generative models, generative causal identification, and LLM causality; (2) clarifies theoretical boundaries and distills ten open challenges; and (3) provides a systematic roadmap toward next-generation generative AI that is interpretable, generalizable, and amenable to causal intervention.
📝 Abstract
In the field of artificial intelligence (AI), the quest to understand and model data-generating processes (DGPs) is of paramount importance. Deep generative models (DGMs) have proven adept in capturing complex data distributions but often fall short in generalization and interpretability. On the other hand, causality offers a structured lens to comprehend the mechanisms driving data generation and highlights the causal-effect dynamics inherent in these processes. While causality excels in interpretability and the ability to extrapolate, it grapples with intricacies of high-dimensional spaces. Recognizing the synergistic potential, we delve into the confluence of causality and DGMs. We elucidate the integration of causal principles within DGMs, investigate causal identification using DGMs, and navigate an emerging research frontier of causality in large-scale generative models, particularly generative large language models (LLMs). We offer insights into methodologies, highlight open challenges, and suggest future directions, positioning our comprehensive review as an essential guide in this swiftly emerging and evolving area.
