To support appropriate user reliance and responsible opinion formation, emerging mandates like the European AI Act enforce strict AI disclosure requirements. However, static labels and hidden watermarks are proving insufficient. Poorly designed disclosures can induce label fatigue or trigger effects that legitimize unlabeled misinformation. Aligning with HCOMP 2026’s core focus on Human-AI Complementarity and Alignment and the theme of “Connections,” this workshop unites researchers across disciplinary boundaries. Together, we aim to chart a course for transforming AI disclosures from legal checkboxes into interactive socio-technical tools that can empower user metacognition.

• Provenance, Watermarking, and Transparency: Evaluating the efficacy of cryptographic watermarking, metadata frameworks, and source-tracking tools for synthetic media.

• Countering Misinformation and Informing Opinion Formation: Assessing how AI disclosures alter user susceptibility to deepfakes, automated propaganda, and algorithmic echo chambers.

• Human Computation for Disclosure Design: Leveraging crowd annotations, preference learning, and human-in-the-loop workflows to evaluate the clarity and neutrality of disclosure language and disclosure designs.

• Algorithmic Literacy and Inclusivity: Tailoring disclosure architectures for diverse demographics, accounting for variations in technical literacy, cultural contexts, and age groups.

• Policy and Standardized Frameworks: Translating high-level legal mandates (e.g., the EU AI Act, the Digital Services Act) into standardized, open-source user interfaces and measurable compliance metrics to bridge the regulation-practice gap.

• Longitudinal and Ecological Validity: Examining how reliance behaviors and trust calibration evolve over sustained AI use in real-world settings, moving beyond single-session laboratory studies to capture habituation effects, disclosure fatigue, and adaptive

user strategies.