AI evidence extraction

Beyond intention: Investigating the dynamics of AI tool usage and purchase behavior in scientific research

M. Çağrı Budak · Information Development · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

2/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1177/02666669261438768

Methodology & findings

Study design

Cross-sectional survey with partial least squares structural equation modeling (PLS-SEM) and Necessary Condition Analysis (NCA).

Sample

N = 557, 3 groups

Primary method

Partial Least Squares Structural Equation Modeling (PLS-SEM) using SmartPLS4 software; Necessary Condition Analysis (NCA) using CE-FDH ceiling technique (suitable for ordinal data); Bootstrapping procedure for assessing statistical significance (mentioned but not detailed); Permutation test with 10,000 permutations for NCA statistical significance; Harman's Single Factor test for Common Method Bias assessment; Factor loading analysis with threshold 0.7 (or 0.4-0.7 with AVE/CR adjustment); Cronbach's α and Composite Reliability (CR) for internal consistency reliability; Average Variance Extracted (AVE) for convergent validity (threshold >0.5); Heterotrait-Monotrait (HTMT) ratio for discriminant validity (threshold <0.9); Variance Inflation Factor (VIF) for collinearity assessment (threshold 0.3-5.0)

Main result

The study found that "performance expectancy, hedonic motivation, habit, and perceived ethics positively impact behavioral intention" to use AI tools. Additionally, "facilitating conditions and behavioral intention positively impact use behavior," and "price value and use behavior shape purchase behavior." A novel finding demonstrated that "personal innovativeness positively impacts behavioral intention indirectly through perceived ethics which reveals a novel pathway in AI tool adoption."

Reports effect sizes and confidence intervals.

Research paradigm

Positivist quantitative empiricism

Author conclusions

The authors conclude: "This study examines scholars' adoption of AI tools in scientific research by extending the UTAUT2 framework with perceived ethics and personal innovativeness, and by simultaneously examining behavioral intention, use behavior, and purchase behavior as target constructs." They further state: "These findings offer practical insights for institutions and policymakers, highlighting the importance of institutional support and ethical frameworks in promoting scholars' adoption and purchase of AI tools."

Risk of bias

Selection bias: Scholars selected only from Turkish higher education institutions with publicly available email addresses, limiting generalizability; Attrition: 28 participants excluded for failing attention-check question; 1 excluded for inconsistent response patterns; Cross-sectional design: Cannot establish causal relationships, only associations; Single-indicator measures: Perceived ethics, use behavior, and purchase behavior measured with single items, reducing reliability; Self-report bias: All data based on self-reported surveys; Geographic limitation: Study conducted only in Turkish higher education context; Selection bias: Scholars selected from Council of Higher Education Academic Search Database who had publicly available email addresses (excludes those without public email); Self-selection bias: Survey respondents self-selected into participation; Single-source data collection: All data collected via self-report survey instrument; Context-specific: Turkish higher education context may limit generalizability; Limited variance in ethics construct: Single-item measurement for perceived ethics (R²=0.182 suggests limited explained variance); Selection bias: Scholars were selected from publicly available email addresses in the Council of Higher Education Academic Search Database, which may exclude those without public listings.; Attention check: 28 of 586 initial responses (4.8%) failed the attention-check question, indicating potential careless responding.; Exclusion of one participant due to inconsistent response patterns.; Context-specificity: Study conducted in Turkish higher education context, limiting generalizability.; Single-indicator constructs: Perceived ethics, use behavior, and purchase behavior measured with single items, which may limit reliability.

Limitations

"This study directly investigated whether the use of AI tools is ethical
Findings from the literature and this study indicate that perceived ethics have a significant impact on the use of AI tools
However, evidence regarding the specific aspects of ethics remains context-dependent in the literature." Additionally, the authors note that "the explanatory power of this impact is quite limited (i.e., Adjusted R 2 = 0.182)" regarding personal innovativeness and ethics
The study also acknowledges that "Although the results provide an insight into academia as a whole, the behavior of academics in different fields and with different titles may vary."

Open questions raised

Future studies should examine different elements of ethics as a higher-order construct to reveal different aspects of ethics
Investigation of other factors influencing perceived ethics beyond personal innovativeness is needed, as the current explanatory power is limited (Adjusted R² = 0.182)
Multigroup analysis focusing on focal areas and academic titles, enhanced with NCA, would yield more detailed results
Future studies should examine the impact of various factors from different theories (e.g., Relative Advantage from DOI) or other UTAUT2 model factors on purchase behavior
Context-dependent variations across different academic fields and countries need examination
Limited understanding of specific ethical aspects: Authors note that "evidence regarding the specific aspects of ethics remains context-dependent in the literature" and recommend examining ethics as a higher-order construct

Data: Not mentioned as publicly availableCode: Not mentionedExtracted from: pdfAgreement 54%

Explore related topics

Productivity Effects Research Productivity Survey