Keywords: AI, public opinion, deliberation, South Africa, Nigeria, autonomy, personalization, user education, emotional cues, cultural differences

This paper examines the findings of a national deliberation on artificial intelligence (AI) in South Africa and Nigeria. The deliberation gathered a diverse sample of respondents in both countries and explored their views on the use of AI agents for various tasks, as well as their preferences on AI autonomy, personalization, and education.

Participants in both South Africa and Nigeria were asked about AI’s potential for completing tasks autonomously and providing personalized experiences. Nigerian participants expressed strong support for AI’s ability to save time, with 92% agreeing that “having AI agents complete tasks autonomously will save users’ time”. South African participants showed less enthusiasm for AI autonomy and personalization, particularly regarding the proposal that AI agents should remember past behaviors or preferences by default. This proposal saw the highest post-deliberation approval in Nigeria (82.8%), while South Africa exhibited the least support (58.1%). In contrast, proposals that allowed users to select and update their settings to specify which past behaviors and preferences AI agents should remember were met with high support across both countries, especially in Nigeria (90.9%).

When it came to user education, participants strongly favored proposals to inform users about AI agents' personalized responses and data usage. Both proposals received over 80% support post-deliberation, with Nigerian participants showing particularly high approval (92.9%) for real-time education on AI data usage. Additionally, when presented with proposals about AI’s human-like qualities, Nigerian participants demonstrated the most support for AI agents responding to emotional cues, especially when the user is informed (88.5%), while South African participants expressed the least support for this idea.

The findings suggest that Nigerians were overall more supportive of AI autonomy and user control over AI behaviors, while South Africans were more cautious and less supportive of certain AI features, especially related to data retention and emotional responses. The paper highlights important cultural differences in AI perceptions across these two countries, underscoring the significance of user education and transparency when it comes to AI agents’ behaviors and data usage. The results emphasize the importance of including diverse perspectives in shaping AI policies that are inclusive and tailored to regional attitudes.