Type: article
Abstract: The use of Generative Artificial Intelligence (GenAI) for financial advice has become increasingly prevalent. While GenAI can provide accessible guidance for high-stakes decisions, including salary negotiations, such advice may also reproduce unequal recommendations through seemingly irrelevant prompt details, such as names, phrasing, or community references, that implicitly signal cultural orientation. Using a large-scale factorial audit, we generate 14,400 responses via OpenAI API calls to controlled salary-negotiation prompts that vary cultural orientation, cultural cue type, location anchoring, model version, and negotiation stakes. We find that collectivistic cues lead to lower recommended salary opening offers than individualistic cues. This disparity is strongest for name-based cues, attenuated but not eliminated by location anchoring, and persistent in newer GenAI models. A low-stakes bicycle-purchase replication shows weaker effects. Our findings reveal a responsibility and contestability gap: equivalent users can receive unequal economic advice through subtle cues that remain difficult to observe or challenge.
Source: AIS Electronic Library (AISeL)
Institution: Erasmus University Rotterdam
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
SDG: Reduced inequalities
Open Access status: green
OpenAlex ID: W7222907699
