Search papers, labs, and topics across Lattice.
This paper audits caste bias in LLM-mediated matchmaking using real-world matrimonial profiles across five LLM families (GPT, Gemini, Llama, Qwen, and BharatGPT). By varying caste identity (Brahmin, Kshatriya, Vaishya, Shudra, and Dalit) and income, the study evaluates how LLMs assess profiles along dimensions of social acceptance, marital stability, and cultural compatibility. The analysis reveals that LLMs consistently favor same-caste matches, with ratings up to 25% higher than inter-caste matches, reproducing traditional caste hierarchies.
LLMs used in matchmaking amplify existing caste hierarchies, rating same-caste matches significantly higher and perpetuating social biases in potentially harmful ways.
Social and personal decisions in relational domains such as matchmaking are deeply entwined with cultural norms and historical hierarchies, and can potentially be shaped by algorithmic and AI-mediated assessments of compatibility, acceptance, and stability. In South Asian contexts, caste remains a central aspect of marital decision-making, yet little is known about how contemporary large language models (LLMs) reproduce or disrupt caste-based stratification in such settings. In this work, we conduct a controlled audit of caste bias in LLM-mediated matchmaking evaluations using real-world matrimonial profiles. We vary caste identity across Brahmin, Kshatriya, Vaishya, Shudra, and Dalit, and income across five buckets, and evaluate five LLM families (GPT, Gemini, Llama, Qwen, and BharatGPT). Models are prompted to assess profiles along dimensions of social acceptance, marital stability, and cultural compatibility. Our analysis reveals consistent hierarchical patterns across models: same-caste matches are rated most favorably, with average ratings up to 25% higher (on a 10-point scale) than inter-caste matches, which are further ordered according to traditional caste hierarchy. These findings highlight how existing caste hierarchies are reproduced in LLM decision-making and underscore the need for culturally grounded evaluation and intervention strategies in AI systems deployed in socially sensitive domains, where such systems risk reinforcing historical forms of exclusion.