Co-creating responsible AI for public safety: A mixed-method approach toward understanding stakeholder perceptions

Responsible deployment of AI-assisted video surveillance (A 2 VS) for public safety requires input from community members and frontline users like security personnel, law enforcement, and small business owners. This study explores their perceptions of A 2 VS through three in-person focus groups (n = 21) held in June 2025 and a public survey (n = 410) conducted in Charlotte, NC from August to September 2023. Findings show broad support: over 63% of focus group participants expressed positive views, and 85% of survey respondents saw the technology as at least somewhat beneficial. Key concerns included real-time anomaly detection, scalability, and system learning. Random forest analysis of the survey showed education and ethnicity as the top predictors, with white, more educated individuals showing greater support. To synthesize insights, we introduced the Alignment–Tension–Gap (ATG) quadrant map by quantifying cross-group attention and within-group agreement. Alignment topics like notification features were discussed in 100% of groups with 100% agreement. Tension topics like real-time detection appeared in 55% of groups but had only 10% agreement. Emerging gaps like business insights were raised in 25% of groups and showed 70% agreement, pointing to opportunities for innovation.