In January 2025, Global Dialogues conducted its second conversation involving 1,120 participants. In partnership with Mistral AI, we asked people how they wanted to personally interact with AI, and how they wanted AI assistants and agents to interact with them. This survey round examined the global perspectives on these AI interaction preferences and cultural expectations, based on the collected survey data from participants across Africa, Asia, Europe, North America, and South America.
Clear preferences emerged regarding how AI should interact. Beyond the general inclination towards formal interactions (the preference strength for which varies regionally, as noted), participants also favored detailed explanations over brief answers.
There was a strong preference for analytical AI capabilities over creative ones. Many expressed a desire for AI to ask clarifying questions to better understand their needs, though some also valued AI's ability to appropriately challenge ideas.
Adaptability to the task context was highly valued across cultures. A significant number of participants also expressed a desire for AI to remain impartial and avoid adopting specific human-like personas, highlighting concerns about potential bias.
The strong preference for AI tailored to local contexts (favored by 76% of participants, who raised concerns about potential misinterpretations by Western-centric systems) encompasses appropriate language use, cultural understanding, and access to locally relevant information. This extends to how AI handles cultural missteps: 86% of participants emphasized that AI must recognize cultural errors, apologize respectfully, explain the mistake, and demonstrate a commitment to learning.
While 84% of participants agreed that AI should operate on scientific principles, they also believed that traditional and cultural knowledge should inform its cultural sensitivity and relevance.
Participants reported primarily using AI for three purposes: productivity (enhancing work efficiency), creativity (generating content like emails), and learning (academic support and skill development).
While generally perceived as effective, concerns regarding the reliability, accuracy, and environmental impact of AI were noted across all surveyed regions.
Regarding information delivery, participants preferred direct answers for simple queries but valued discussion for complex or open-ended topics, indicating a need for AI to offer flexible interaction based on the nature of the question.
The qualities central to trustworthy AI—identified by respondents as honesty, reliability, experience, empathy, clear communication, and respect for privacy—underpin the broader ethical expectations for these systems.
This emphasis on trustworthy qualities aligns with the strong consensus (over 74%) that AI systems should defer to human wisdom or community leaders when dealing with complex situations. These include ethical dilemmas, matters of cultural sensitivity, legal issues, mental health concerns, and topics related to religion, politics, and personal relationships.
Explore what the world is really saying about AI. Access our complete dataset to conduct your own analysis and contribute to the growing body of research on global AI perspectives.
