Towards Trustworthy and Aligned Machine Learning: A Data-centric Survey with Causality Perspectives

TL;DR

This survey reviews recent advances in trustworthy machine learning from a data-centric perspective, unifying various methods through causal frameworks and exploring their application to large pretrained models.

Contribution

It introduces a unified language based on Pearl's causal hierarchy to connect diverse trustworthy ML methods and extends this understanding to large pretrained models.

Findings

Causal hierarchy provides a unifying framework for trustworthy ML methods.

Connections between ERM and techniques like fine-tuning and prompting are established.

A cohesive understanding of robustness, interpretability, and fairness methods is developed.

Abstract

The trustworthiness of machine learning has emerged as a critical topic in the field, encompassing various applications and research areas such as robustness, security, interpretability, and fairness. The last decade saw the development of numerous methods addressing these challenges. In this survey, we systematically review these advancements from a data-centric perspective, highlighting the shortcomings of traditional empirical risk minimization (ERM) training in handling challenges posed by the data. Interestingly, we observe a convergence of these methods, despite being developed independently across trustworthy machine learning subfields. Pearl's hierarchy of causality offers a unifying framework for these techniques. Accordingly, this survey presents the background of trustworthy machine learning development using a unified set of concepts, connects this language to Pearl's causal hierarchy, and finally discusses methods explicitly inspired by causality literature. We provide a unified language with mathematical vocabulary to link these methods across robustness, adversarial robustness, interpretability, and fairness, fostering a more cohesive understanding of the field. Further, we explore the trustworthiness of large pretrained models. After summarizing dominant techniques like fine-tuning, parameter-efficient fine-tuning, prompting, and reinforcement learning with human feedback, we draw connections between them and the standard ERM. This connection allows us to build upon the principled understanding of trustworthy methods, extending it to these new techniques in large pretrained models, paving the way for future methods. Existing methods under this perspective are also reviewed. Lastly, we offer a brief summary of the applications of these methods and discuss potential future aspects related to our survey. For more information, please visit http://trustai.one.