AI Models Show Gender-Based Risk Behavior Variations
- Research from Allameh Tabataba’i University shows AI models alter risk behavior based on gender prompts.
- DeepSeek and Google’s Gemini became more risk-averse when acting as women, reflecting real-world trends.
- OpenAI’s GPT models maintained a neutral stance, unaffected by gender identity prompts.
- Meta’s Llama and xAI’s Grok showed inconsistent responses to gender-based prompts.
The study highlights the influence of gender stereotypes on AI decision-making, with some models adapting their risk profiles based on assumed gender roles. This could lead to unintended biases in AI applications affecting high-stakes decisions like finance and healthcare.
Researchers emphasize the need for “bio-centric measures” to ensure AI systems reflect human diversity without reinforcing stereotypes, as demonstrated by DeepSeek and Gemini’s increased caution when prompted as female. Source
