MiALA 2026 Annual Conference

This poster reports findings from a small-scale study comparing Google Lens' assisted and human-only descriptive metadata creation for murals in Detroit. AI-supported workflows aided visual identification and reduced initial descriptive effort but introduced epistemic friction, including misidentifications and non-authorized terminology, requiring verification. Results position Google Lens as workflow support rather than replacement, underscoring the continued necessity of human involvement in authority control, contextual interpretation, and culturally grounded description in AI-integrated environments.