Name in original language
Panoramic au service des Assises de la protection sociale des Français de l'étranger
Initiative overview
Objectives:
- To democratise access to information on a complex subject (international social protection) by offering all citizens a simple, clear and reliable tool for understanding the issues at stake;
- to promote informed citizen participation by allowing everyone to ask questions freely and obtain contextualised answers from public and verified sources;
- and to strengthen the inclusiveness of public debate by involving French citizens living outside France—who are often distant from participatory processes—in an innovative, transparent and accessible approach.
Results, outcomes and impacts
Panoramic has logged 963 total queries: 560 open-ended questions (58.2%). Some users conducted extended exchanges (more than 20 questions), indicating deeper exploration needs. The platform supported targeted, mostly on-topic information seeking and provided source-grounded answers based exclusively on the CNDP-validated corpus. Reported impacts include access to reliable information for hundreds of citizens, an innovative way to equip large-scale public debates, and making the topic more accessible, widening participation and giving a voice to people who often feel excluded. Measurement relies on platform analytics (volumes, open versus click-based actions, session depth) and answer-versus-no-answer categorisation by theme.
Challenges and lessons learned
Key risks and challenges included (1) handling questions outside the validated scope; and (2) addressing legitimate but sensitive or interpretive questions (feelings of unfairness, complex personal situations) that a strictly document-based assistant cannot always resolve. The response was to answer only from the CNDP-validated corpus and to refuse to answer when content fell outside it, avoiding approximations and reinforcing trust. Lessons learned: users often arrive with concrete personal questions and may engage in long conversational exploration, so the tool must support both quick Q&A and deep dives; credibility depends as much on knowing when not to answer as on answering well; and “critical” prompts need dedicated handling distinct from information requests. Conditions for success: a high-quality knowledge base, clear scope boundaries, and continuous monitoring of question types to keep answers clear, contextualised and verifiable.
