Abstract

Large language models (LLMs) have significantly advanced the field of automated code
generation. However, a notable research gap exists in evaluating social biases that may be
present in the code produced by LLMs. To solve this issue, we propose a novel fairness
framework, i.e., Solar, to assess and mitigate the social biases of LLM-generated code.
Specifically, Solar can automatically generate test cases for quantitatively uncovering
social biases of the auto-generated code by LLMs. To quantify the severity of social biases
in generated code, we develop a dataset that covers a diverse set of social problems. We
applied Solar and the crafted dataset to four state-of-the-art LLMs for code generation.
Our evaluation reveals severe bias in the LLM-generated code from all the subject LLMs.
Furthermore, we explore several prompting strategies for mitigating bias, including Chainof-
Thought (CoT) prompting, combining positive role-playing with CoT prompting, and
dialogue with Solar. Our experiments show that dialogue with Solar can effectively reduce
social bias in LLM-generated code by up to 90%.
Beyond single prompts, we studied social bias in multi-agent LLM workflows using
FlowGen, where agents act as requirement engineers, architects, developers, and testers.
The results show that the design of the workflow, the fairness-aware role instructions, and
the composition of the roles affect the fairness of the code.
Our findings demonstrate that social bias is a systemic issue in LLM-based code generation.
Solar offers a practical tool for assessing bias risks, tracing their origins, and applying
targeted mitigation strategies, including collaborative workflows.