Reward Model Perspectives: Whose Opinions Do Reward Models Reward?
This work addresses the propagation of unwanted social biases in language technologies, which is a critical issue for AI alignment and fairness, though it is incremental in its formalization and analysis of existing methods.
The paper tackled the problem of reward models (RMs) in language model alignment by investigating their sociodemographic biases and misalignment with various demographic groups, showing that RMs can systematically reward harmful stereotypes and that prompting alone is insufficient to correct these issues.
Reward models (RMs) are central to the alignment of language models (LMs). An RM often serves as a proxy for human preferences to guide downstream LM behavior. However, our understanding of RM behavior is limited. Our work (i) formalizes a framework for measuring the alignment of opinions captured by RMs, (ii) investigates the extent to which RMs demonstrate sociodemographic biases, and (iii) explores the effects of prompting to steer rewards towards the preferences of a target group. We study the subjective and diverse perspectives on controversial topics, which allows us to quantify RM perspectives in terms of their opinions, attitudes, and values. We show that RMs are poorly aligned with several demographic groups and can systematically reward harmful stereotypes, and steering alone is not enough to overcome these limitations. Our findings underscore the need for more careful consideration of RM behavior in model alignment during preference learning to prevent the propagation of unwanted social biases in the language technologies that we use.