Towards Identifying Social Bias in Dialog Systems: Frame, Datasets, and Benchmarks

TL;DR

This paper introduces a new framework and dataset for detecting social bias in Chinese dialog systems, providing benchmarks to improve safety and reduce biases in conversational AI.

Contribution

It proposes the Dial-Bias Frame for analyzing social bias, creates the first annotated Chinese bias dialog dataset, and establishes benchmarks for bias detection at multiple levels.

Findings

The Dial-Bias Frame enables comprehensive bias analysis.

The CDail-Bias Dataset is the first annotated Chinese social bias dialog dataset.

Benchmarks show the importance of detailed analysis for bias detection.

Abstract

The research of open-domain dialog systems has been greatly prospered by neural models trained on large-scale corpora, however, such corpora often introduce various safety problems (e.g., offensive languages, biases, and toxic behaviors) that significantly hinder the deployment of dialog systems in practice. Among all these unsafe issues, addressing social bias is more complex as its negative impact on marginalized populations is usually expressed implicitly, thus requiring normative reasoning and rigorous analysis. In this paper, we focus our investigation on social bias detection of dialog safety problems. We first propose a novel Dial-Bias Frame for analyzing the social bias in conversations pragmatically, which considers more comprehensive bias-related analyses rather than simple dichotomy annotations. Based on the proposed framework, we further introduce CDail-Bias Dataset that, to our knowledge, is the first well-annotated Chinese social bias dialog dataset. In addition, we establish several dialog bias detection benchmarks at different label granularities and input types (utterance-level and context-level). We show that the proposed in-depth analyses together with these benchmarks in our Dial-Bias Frame are necessary and essential to bias detection tasks and can benefit building safe dialog systems in practice.