Bridging the Knowledge Gap: Understanding User Expectations for Trustworthy LLM Standards

How do users define “trustworthy” when it comes to large language models? We’re excited to share that Michaela Benk, newly minted PhD, from Mobiliar Lab for Analytics at ETH Zurich presented our latest research at AAAI 2025!

Our study, “Bridging the Knowledge Gap: Understanding User Expectations for Trustworthy LLM Standards”, explores how users interpret and prioritize the values behind AI standards. Through in-depth interviews and a large-scale survey, we identified 68 user-driven criteria for assessing trust in LLMs.

Findings show:
– Transparent governance matters
– Technical solutions need non-technical complements
– Strategic communication can close the user-expert divide

Proud to contribute to the AAAI 2025 Special Track on AI Alignment, and to help move the conversation forward on aligning AI with societal values!

Read the full paper here:
https://lnkd.in/eR-wQERQ

Learn more about the Lab: https://lnkd.in/eXmfE78U