Conceptual Framework: AI Wellbeing, Privacy, Trust, and Disclosure

Quantitative specification of the conceptual framework

The model is specified as a theory-driven latent-variable framework integrating Privacy Calculus Theory and Trust Theory. Willingness to Disclose is the primary endogenous outcome variable. Privacy Concern, Trust in AI, and Perceived Benefit are exogenous predictors. Privacy Calculus functions as a mediating construct that transmits part of the effect of predictors to disclosure.

Construct roles and directional hypotheses

• H1: Privacy Concern -> Privacy Calculus (negative structural path).
• H2: Trust in AI -> Privacy Calculus (positive structural path).
• H3: Perceived Benefit -> Privacy Calculus (positive structural path).
• H4: Privacy Calculus -> Willingness to Disclose (positive structural path).
• H5: Privacy Concern -> Willingness to Disclose (negative direct effect).
• H6: Trust in AI -> Willingness to Disclose (positive direct effect).

Operationalization and measurement model

• Each construct is operationalized using multi-item Likert indicators and aggregated into composite scores.
• Internal consistency reliability is assessed via Cronbach's alpha for each construct.
• Bivariate dependence among composites is examined using Pearson correlation coefficients with two-tailed p-values and Holm adjustment.
• The criterion equation can be estimated using OLS regression (Disclosure ~ Privacy + Trust + Calculus), with Perceived Benefit represented in the calculus mechanism.

Suggested academic wording (ready to cite in methodology)

"This study adopts a theoretically specified conceptual framework in which willingness to disclose personal wellbeing concerns to AI is modeled as the principal endogenous variable. Privacy concern, trust in AI, and perceived benefit are specified as exogenous antecedents, while privacy calculus is modeled as a mediating mechanism linking antecedents to disclosure intention. The model includes both mediated and direct effects to evaluate partial mediation. Constructs are operationalized through multi-item Likert measures, reliability is evaluated using Cronbach's alpha, and inter-construct associations are tested using Pearson correlations with multiple-comparison adjustment. Criterion relationships are estimated via regression."