Available AI-dependence and attachment instruments follow three non-interchangeable measurement logics and do not converge on one stable construct¶

Claim · All claims
Evidence · 1 study · review r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Measurement instruments for AI dependence, attachment, and reliance follow three broad logics (addiction-component, attachment, reliance) with different item content; internal consistency within a scale does not show they measure one latent construct. → Yan D 2026

Evidence¶

Yan D 2026¶

Yan D (2026). A critical narrative synthesis of psychological correlates, measurement, and reported findings on conversational AI engagement and dependence-related constructs. Front. Psychol. 17:1827795. https://doi.org/10.3389/fpsyg.2026.1827795

q2 · i? · review · r2

The review's psychometric synthesis (Section 3.3, Table 3) of instruments developed 2024-2026, finding that "Addiction-component instruments emphasize impaired control and harm" while attachment and reliance instruments target different content. Evidence for invariance, temporal stability, and discriminant validity remains uneven.

"The measures follow three broad logics. Addiction-component instruments emphasize impaired control and harm. Attachment instruments emphasize emotional closeness, anxiety, avoidance, or social substitution. Reliance instruments emphasize cognitive outsourcing and task delegation."

Discussion¶

Related Claims¶

• The survey instrument's nine constructs show convergent and discriminant validity in this sample — related
• Reliance, trust, and dependency are distinct constructs requiring separate measurement from AI-literacy totals — related
• Strategic reliance is positively associated with AI literacy, and the reliance types differentiate students across writing process and outcome variables — related
• The post-study survey instrument showed high internal consistency (alpha = 0.90) — related
• Questionnaire constructs showed acceptable to strong internal consistency (α = 0.740 to 0.945) — related
• Exploratory factor analysis of survey items yields a unidimensional 9-item AI pedagogical orientation scale with strong reliability — related
• Critical Evaluation and Dependent reliance emerge as empirically distinct constructs, with near-zero correlation between the two facets — a narrower finding that bears on this claim
• Within-method convergence (RSAT coding vs human coding; G vs d') indicates stability, not construct validity — a narrower finding that bears on this claim
• Literature and expert opinion converge on a limited number of key noncognitive attributes that matter for student success, with multiple available instruments — reports the opposite
• The post-interaction reflection instrument showed satisfactory internal consistency (Cronbach's α = .863) — related
• TAILS dimensions show high internal consistency, with Cronbach's alpha values above 0.90 — related
• A low alpha does not necessarily indicate high content validity, because reliability and validity can both be high — related
• Cronbach's alpha does not generally indicate the reliability of knowledge tests because knowledge is a heterogeneous construct — related