Map the Complexities of Human Values for AI
The Global AI Values Challenge is an invitation-based, research-oriented competition initiated by Microsoft Research Asia (MSRA).
It invites participants from around the world to help tackle one of the most fundamental and unresolved questions in artificial intelligence today:
Can AI systems reason, judge, and act in ways aligning with human values with real-world complexity？
Join Us 
The Global AI Values Challenge is not about quick answers or easy benchmarks.
It is about raising the questions that matter most and helping AI learn to respond with wisdom, responsibility, and respect for human values.
We invite you to contribute, challenge, and help define the future of value-aligned AI.
What Is This Challenge About?
Large language models (LLMs) have made remarkable progress in language understanding and mathematical reasoning. Yet when confronted with questions involving human values, moral trade-offs, and ethical judgment, which are essential and common in humans' everyday life, current AI systems might struggle, or fail entirely, to produce answers that are value-aware, meaningful and helpful to the users asking them.
This Challenge asks participants to contribute the hardest, most complex value/morality-related questions and answers imaginable that can expose the limitations of today’s advanced reasoning based AI systems.
Can you think of a real question where current AI cannot provide a reasonable answer that aligns with the universal values and morality?
Together, we aim to build the most challenging and comprehensive global dataset of value-oriented questions and answers ever assembled, grounded in real-world dilemmas rather than abstract thought experiments.
Why Does This Matter?
By collecting and curating these difficult cases, the Challenge seeks to:
Advance academic research on value reasoning, moral judgment, and ethical alignment in AI
Promote transparent, reproducible standards for value-related AI evaluation across academia and industry
Provide clearer moral and value-oriented guidance for the responsible development of future AI systems
Our long-term vision is to support the creation of AI agents that can offer value-based decision support in everyday contexts, to help users solve dilemmas and improve well-being, while operating in a safe, reliable, and responsible manner for the benefit of society.
Who Should Participate?
The Challenge is open to participants worldwide. We warmly welcome graduate and doctoral students, postdoctoral researchers, faculty members and academic researchers
from the humanities and social sciences, including (but not limited to):
psychology, sociology, social science, philosophy, ethics, law, literature, journalism, and communication;
working at the
intersection of AI and social sciences
We especially encourage participation from individuals with diverse cultural, disciplinary, and social perspectives, as understanding human values requires more than a single worldview.
What Do Participants Do?
Participants are invited to submit:
Challenging, real-world questions involving human values and ethics
Accompanied by carefully reasoned answers and justifications that reflect value-aware judgment
Submissions will undergo a review process to ensure quality, originality, and relevance to the Challenge's goals.
What Can You Gain by Participating?
If you agree to the challenge consent and your submission passes review, you will receive:
Monetary rewards of RMB 200 or RMB 500 will be granted for each accepted Q&A pair, depending on the quality and impact of the contribution.
An official Certificate of Participation in the MSRA Global AI Values Open Research Project
Public recognition as a contributor to this global values dataset
The priority to receive information about MSRA Societal AI internships
Beyond these tangible benefits, participants become part of a global research effort shaping how future AI systems understand and respect human values.
Process
The Challenge is conducted fully online through this website. Participants are invited to contribute value-related questions that current AI systems cannot answer reliably, focusing on human values, ethics, and moral judgment in real-life contexts.
