New faith-based university coalition find systematic religious bias in AI models (www.deseret.com)

A new coalition, the Consortium for Evaluating Faith and Ethics in AI (CEFE-AI), composed of researchers from four prominent faith-based universities, has revealed significant religious bias in AI models. This announcement was made at the Athens Summit on Faith and Artificial Intelligence, one day after Pope Leo XIV emphasized the need for AI to uphold human dignity and faith. The coalition introduced the AllFaith Benchmark, a set of tests designed to assess how well AI engages with various religious perspectives. Findings showed that AI models overwhelmingly neglect religious viewpoints when responding to existential questions, with only 3% of responses incorporating any religious context, despite over half of surveyed individuals valuing religious input in such discussions. The implications of this research are profound for the AI/ML community, highlighting a systematic pattern of exclusion that undermines the representation of diverse faiths in consequential conversations. Additionally, the studies identified biases in guiding users about religious conversion, favoring certain faiths over others, such as a negative bias against Jehovah’s Witnesses and a positive bias toward Catholicism. CEFE-AI aims to partner with AI companies to redesign language models in a way that respects and reflects the ethical and religious wisdom of billions worldwide. By fostering collaboration between faith leaders and AI developers, the consortium seeks to ensure that AI technologies accurately represent diverse belief systems and contribute positively to public discourse.