Contextual Acceptance of Artificial Intelligence in Commercial Policy-Making: An Integrated Model of Acceptance, Resistance, and Institutional Factors
Abstract
This study examines how senior decision-makers in Iranian commercial policy institutions evaluate and respond to artificial intelligence (AI), with particular attention to the interplay of adoption, resistance, and institutional considerations. It develops the notion of contextual acceptance to describe the importance of whether AI-supported decisions can be justified to relevant stakeholders. An exploratory sequential mixed-methods design was used. The qualitative phase involved 18 semi-structured interviews with experts in commercial policy, information technology, and academia. The authors report thematic saturation after the fourteenth interview and four additional interviews to consolidate the themes. In the quantitative phase, a 27-item five-point Likert questionnaire was completed by 248 senior managers and specialists selected through stratified random sampling. The reported analysis used partial least squares structural equation modeling (PLS-SEM) in SmartPLS 4. The qualitative analysis identified digital leadership, perceived usefulness, institutional trust, algorithmic opacity, organizational resistance, and perceived risk. In the structural model, digital leadership (β = 0.36, t = 6.3), perceived usefulness (β = 0.34, t = 5.9), and institutional trust (β = 0.28, t = 4.7) were positively associated with attitudes toward AI. Algorithmic opacity (β = −0.22, t = 3.8) and organizational resistance (β = −0.19, t = 3.4) were negatively associated with attitudes. The reported R² values were 0.64 for attitude and 0.69 for behavioral intention, with an overall GOF value of 0.61. All nine hypotheses were reported as supported, although the indirect-effect coefficient and test statistic for mediation were not provided. The findings are consistent with a context-sensitive account of AI acceptance in which perceived institutional accountability and the defensibility of algorithm-supported decisions complement assessments of technical usefulness. Potential implications include clearer accountability arrangements, explainable AI systems, and digital leadership development; causal and policy-effectiveness claims require further study.
Abstract
This study examines how senior decision-makers in Iranian commercial policy institutions evaluate and respond to artificial intelligence (AI), with particular attention to the interplay of adoption, resistance, and institutional considerations. It develops the notion of contextual acceptance to describe the importance of whether AI-supported decisions can be justified to relevant stakeholders. An exploratory sequential mixed-methods design was used. The qualitative phase involved 18 semi-structured interviews with experts in commercial policy, information technology, and academia. The authors report thematic saturation after the fourteenth interview and four additional interviews to consolidate the themes. In the quantitative phase, a 27-item five-point Likert questionnaire was completed by 248 senior managers and specialists selected through stratified random sampling. The reported analysis used partial least squares structural equation modeling (PLS-SEM) in SmartPLS 4. The qualitative analysis identified digital leadership, perceived usefulness, institutional trust, algorithmic opacity, organizational resistance, and perceived risk. In the structural model, digital leadership (β = 0.36, t = 6.3), perceived usefulness (β = 0.34, t = 5.9), and institutional trust (β = 0.28, t = 4.7) were positively associated with attitudes toward AI. Algorithmic opacity (β = −0.22, t = 3.8) and organizational resistance (β = −0.19, t = 3.4) were negatively associated with attitudes. The reported R² values were 0.64 for attitude and 0.69 for behavioral intention, with an overall GOF value of 0.61. All nine hypotheses were reported as supported, although the indirect-effect coefficient and test statistic for mediation were not provided. The findings are consistent with a context-sensitive account of AI acceptance in which perceived institutional accountability and the defensibility of algorithm-supported decisions complement assessments of technical usefulness. Potential implications include clearer accountability arrangements, explainable AI systems, and digital leadership development; causal and policy-effectiveness claims require further study.
Keywords: artificial intelligence; commercial policy-making; contextual acceptance; digital leadership; institutional trust; technological resistance
Introduction
Artificial intelligence is increasingly being considered for public-sector analysis and policy formulation. Differences between planned initiatives and operational deployment raise an organizational question: why might policy officials hesitate to rely on AI even when they recognize its potential usefulness? Accountability, regulatory clarity, and confidence in the decision process may be relevant alongside technical performance.
Established acceptance frameworks, including the Technology Acceptance Model (TAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT), emphasize perceptions of usefulness and other determinants of behavioral intention (Davis, 1989; Venkatesh et al., 2003, 2012). However, commercial policy officials operate in an institutional environment in which decisions may carry public consequences and require justification to senior authorities and other stakeholders (Janssen et al., 2020). Resistance and acceptance may therefore coexist rather than represent opposite ends of a single continuum.
This study asks why senior officials in Iranian commercial policy institutions may hesitate to adopt AI even when they regard it as beneficial. It aims (a) to explore perceived mechanisms of acceptance and resistance, (b) to assess relationships among the constructs identified, and (c) to discuss contextual acceptance as a concept linking technological evaluations with institutional defensibility. The mixed-methods design combines qualitative interviews with a survey-based structural model.
Theoretical Background
Technology acceptance and resistance
TAM proposes that perceived usefulness and perceived ease of use inform technology acceptance (Davis, 1989), while subsequent UTAUT formulations incorporate additional social and facilitating factors (Venkatesh et al., 2003, 2012). Resistance can also constitute an active, potentially rational response to organizational change, rather than merely an absence of favorable attitudes (Lapointe & Rivard, 2005; Kim & Kankanhalli, 2009). The lack of transparent explanations for algorithmic outputs is a related source of uncertainty (Burrell, 2016).
Institutional governance and digital leadership
The institutional dimension of public-sector AI involves formal rules, governance responsibilities, and perceived legitimacy. Research on digital transformation emphasizes changes in administrative processes and institutional structures (Mergel et al., 2019; Vial, 2019). The study also draws on the literature concerning algorithmic governance, human oversight, and the extent to which decision-makers can understand and defend automated recommendations (Alon-Barkat & Busuioc, 2023; Grimmelikhuijsen, 2023). Within this perspective, senior leadership may signal organizational permission and legitimacy for the responsible use of AI.
Prior research and the proposed contribution
Prior scholarship spans technology adoption, organizational resistance and risk, institutional governance, and digital leadership. Related Iranian research on AI-enabled recruitment scenarios (Rouhani et al., 2024) and organizational productivity (Ataei, 2024) offers contextual parallels, although neither examines AI acceptance in commercial policy-making directly. The proposed contribution is an integrated examination of favorable technological evaluations alongside perceived organizational and institutional barriers.
Research stream | Emphasis | Limitation identified in the source |
|---|---|---|
Technology adoption (Dwivedi et al., 2023) | Perceived value, trust, infrastructure | Resistance not modeled explicitly |
Algorithmic governance (Young et al., 2019) | Discretion and uncertainty | Institutional context requires further attention |
Public-sector AI (Wirtz et al., 2020) | Governance and public administration | Commercial policy context not examined |
Technology resistance (Kim & Kankanhalli, 2009) | Risk and status quo considerations | Joint acceptance–resistance model not central |
Digital government (Mergel et al., 2019) | Institutional transformation | Sector-specific decision processes differ |
Conceptual framework and hypotheses
The conceptual framework includes digital leadership, perceived usefulness, institutional trust, algorithmic opacity, organizational resistance, and perceived risk, with attitude toward AI as an intermediate construct and intention to use AI as the outcome. Structural guarantees are conceptually relevant but were not tested as a moderator. The nine reported hypotheses assess five predictors of attitude, two direct predictors of intention, the attitude–intention association, and mediation. A separate structural coefficient for perceived risk is not reported.
Methods
Design and qualitative phase
The study used an exploratory sequential mixed-methods design. Eighteen in-depth semi-structured interviews were reported with experts from the Ministry of Industry, Mine and Trade, the Islamic Republic of Iran Customs Administration, the Iran Chamber of Commerce, and universities. Purposive recruitment was used. The research team reported saturation at interview 14, followed by four additional interviews. The source describes thematic analysis alongside open, axial, and selective coding, with participant review, researcher triangulation, and peer checking as credibility procedures. The detailed audit trail and full interview protocol were not included in the supplied manuscript.
Survey and statistical analysis
The quantitative phase included 248 senior managers and specialists who completed a researcher-developed, 27-item questionnaire using five-point Likert responses. Stratified random sampling was reported. Data were analyzed using partial least squares structural equation modeling (PLS-SEM) in SmartPLS 4. The reported KMO statistic was 0.89, and Bartlett’s test was significant (p < .001). The first factor accounted for less than 35% of the variance in Harman’s single-factor test; this diagnostic alone cannot exclude common-method bias.
Hypotheses
H1. Digital leadership is positively associated with attitudes toward AI.
H2. Perceived usefulness is positively associated with attitudes toward AI.
H3. Institutional trust is positively associated with attitudes toward AI.
H4. Algorithmic opacity is negatively associated with attitudes toward AI.
H5. Organizational resistance is negatively associated with attitudes toward AI.
H6. Attitudes toward AI are positively associated with intention to use AI.
H7. Institutional trust is directly and positively associated with intention to use AI.
H8. Digital leadership is directly and positively associated with intention to use AI.
H9. Attitudes toward AI mediate relationships between the predictors and intention to use AI.
Results
Qualitative themes
The qualitative phase yielded more than 180 initial codes, which were synthesized into six principal constructs: digital leadership, perceived usefulness, institutional trust, algorithmic opacity, organizational resistance, and perceived risk. Four illustrative interview excerpts are shown in Table 2 in English translation, retaining the original participant identifiers.
Reported construct | English translation of quotation | Interpretive code |
|---|---|---|
Algorithmic opacity | “I do not know on what basis this algorithm made this recommendation. It is not defensible at all.” (P7, ministerial adviser) | Difficulty defending a decision |
Institutional trust | “If we had a transparent legal framework specifying who is ultimately responsible, we would feel more reassured.” (P5, IT manager) | Absence of a legal framework |
Digital leadership | “If senior management does not support it, nobody will take the risk of using these systems.” (P3, director general) | Hierarchical support |
Organizational resistance | “I prefer to rely on my 20 years of experience. At least if I make a mistake, I am personally accountable.” (P11, deputy) | Personal accountability risk |
Measurement model
The manuscript reports Cronbach’s alpha values above 0.80, composite reliability values above 0.85, average variance extracted (AVE) values above 0.50, heterotrait–monotrait ratios (HTMT) below 0.85, and variance inflation factors (VIF) below 3.3. The maximum HTMT was reported as 0.74 for attitude and intention. Table 3 preserves the individual values exactly as provided.
Construct | Cronbach’s α | CR | AVE |
|---|---|---|---|
Digital leadership (DL) | 0.89 | 0.92 | 0.72 |
Perceived usefulness (PU) | 0.87 | 0.91 | 0.70 |
Institutional trust (TR) | 0.85 | 0.89 | 0.66 |
Algorithmic opacity (OP) | 0.83 | 0.88 | 0.62 |
Organizational resistance (RE) | 0.82 | 0.87 | 0.60 |
Attitude (AT) | 0.84 | 0.89 | 0.69 |
Behavioral intention (BI) | 0.86 | 0.90 | 0.71 |
Hypothesis testing
Eight structural paths have reported coefficients and t statistics (Table 4). Mediation (H9) is identified as supported in the study results, but no indirect-effect coefficient, confidence interval, or corresponding test statistic is reported; its statistical support cannot therefore be independently evaluated.
Hypothesis / path | β | t | Source decision |
|---|---|---|---|
H1: Digital leadership → Attitude | 0.36 | 6.3 | Supported |
H2: Perceived usefulness → Attitude | 0.34 | 5.9 | Supported |
H3: Institutional trust → Attitude | 0.28 | 4.7 | Supported |
H4: Algorithmic opacity → Attitude | −0.22 | 3.8 | Supported |
H5: Organizational resistance → Attitude | −0.19 | 3.4 | Supported |
H6: Attitude → Intention | 0.47 | 7.5 | Supported |
H7: Institutional trust → Intention | 0.23 | 4.2 | Supported |
H8: Digital leadership → Intention | 0.29 | 4.9 | Supported |
H9: Attitude as mediator | Not reported | Not reported | Reported supported; test unavailable |
Structural-model indicators
Indicator | Reported value | Threshold in source |
|---|---|---|
R² (attitude) | 0.64 | > 0.26 |
R² (behavioral intention) | 0.69 | > 0.26 |
Q² (attitude) | 0.41 | > 0.35 |
Q² (behavioral intention) | 0.45 | > 0.35 |
GOF | 0.61 | > 0.36 |
The model accounted for reported variance in attitude and behavioral intention (Table 5). Comparisons of model performance with other studies require directly comparable samples and analytical procedures. The global GOF statistic is retained as reported but should not be considered a substitute for contemporary PLS-SEM diagnostics.
Discussion
Interpretation of the reported associations
Attitude had the largest reported coefficient predicting intention to use AI (β = 0.47; t = 7.5), consistent with the importance of cognitive evaluation in technology acceptance frameworks (Davis, 1989; Venkatesh et al., 2003). However, attitudes were described as arising from several perceived organizational and technological conditions rather than from usefulness alone.
Digital leadership was positively associated with attitude (β = 0.36) and had a reported direct association with intention (β = 0.29). The qualitative observations suggest that support from senior leadership can be interpreted as organizational permission to use AI. Institutional trust was also associated with attitude (β = 0.28) and directly with intention (β = 0.23), suggesting that formal assurance and perceptions of accountability may matter alongside individual evaluations of technology.
Algorithmic opacity (β = −0.22) and organizational resistance (β = −0.19) were negatively associated with attitudes, according to the reported model. Participants’ statements highlighted uncertainty about the basis of algorithmic recommendations and the personal consequences of endorsing them. This evidence supports treating explainability and institutional accountability as potentially relevant topics for adoption research; it does not establish that officials are unconcerned about technical errors or that accountability is always more important than accuracy.
Theoretical contribution and practical implications
The study uses contextual acceptance to describe the joint consideration of technical value and institutional defensibility. This framing is compatible with bounded rationality and the organizational context of technology adoption (Simon, 1997). The study additionally brings acceptance and resistance into the same proposed account, with attention to how legitimacy, accountability, and leadership may shape managers’ attitudes and intentions. These contributions remain conceptual and context-specific until further replication and direct measurement of the proposed institutional mechanisms.
Potential policy implications include articulating formal responsibilities for AI-assisted decisions, specifying conditions for human review, and establishing organizational procedures for explaining and auditing algorithmic recommendations. For managers, structured digital leadership development may be relevant. For system designers, making recommendations interpretable to accountable decision-makers is a sensible design objective. The cross-sectional associations in this study do not establish that any of these interventions would have a particular causal effect.
Limitations and future directions
This study used cross-sectional survey data, which do not establish causal direction or changes in attitudes over time. The sample was restricted to Iranian commercial policy institutions. Indicator loadings, detailed recruitment information, the qualitative audit trail, and the bootstrap statistics were not reported in the manuscript. The structural moderating role of guarantees was not tested, mediation lacked indirect-effect statistics, and perceived risk did not have a separately reported structural path.
Future research could assess the model longitudinally, test structural guarantees and other moderators, provide bootstrap confidence intervals for all indirect effects, and examine the separate roles of human oversight and explainability. Comparisons with additional policy sectors and jurisdictions would also help assess transferability.
Conclusion
The qualitative and survey results characterize AI acceptance in Iranian commercial policy-making as an organizational decision shaped by technology appraisals and institutional context. Perceived usefulness, digital leadership, and institutional trust were positively associated with attitudes; algorithmic opacity and organizational resistance were negatively associated. Institutional trust was also positively associated with behavioral intention. Contextual acceptance is proposed as a framework for considering how technological judgments intersect with institutional accountability. Further research is required to evaluate the proposed explanatory mechanisms and indirect effects.
Declarations
Ethical considerations. This study involved interviews with professional experts and a questionnaire administered to managers and specialists. Interview quotations are identified by participant codes rather than names.
Competing interests. No competing interests are disclosed in the materials accompanying this article.
Funding. None
Data availability. The interview transcripts and individual-level survey responses are not publicly included with this article. Sharing potentially identifiable material is subject to appropriate confidentiality safeguards and participant consent.
Author contributions. Hamidreza Goldouz: Conceptualization, Investigation, Data curation, Writing – original draft. Ezatollah Abbasian: Supervision, Methodology, Project administration, Writing – review and editing. Nastaran Hajiheydari: Methodology, Validation, Writing – review and editing.
References
Alon-Barkat, S., & Busuioc, M. (2023). Human–AI interactions in public sector decision making: "Automation bias" and "selective adherence" to algorithmic advice. Journal of Public Administration Research and Theory, 33(1), 153–169.
Rouhani, A., Amin-Kazemi, B., & Arasteh, M. (2024). Developing scenarios for the use of artificial intelligence in human resource recruitment in Iran. Transactions on Data Analysis in Social Science, 6(3), 185–197. https://doi.org/10.47176/TDASS.2024.185
Ataei, E. (2024). Identifying the Factors Affecting Human Resource Productivity. Transactions on Data Analysis in Social Science, 6(2), 127-139. https://doi.org/10.47176/TDASS.2024.127
Bian, X., Wang, B., & Yang, A. (2025). The trust trifecta: How transparency, ethics, and benefits shape public confidence in government AI. Government Information Quarterly, 42(4), 102083.
Burrell, J. (2016). How the machine 'thinks': Understanding opacity in machine learning algorithms. Big Data & Society, 3(1), 1–12.
Criado, J. I., Sandoval-Almazán, R., & Gil-Garcia, J. R. (2024). Artificial intelligence and public administration: Understanding actors, governance, and policy from micro, meso, and macro perspectives. Public Policy and Administration, 40(2), 173–184.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340.
Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., & Kar, A. K. (2023). Opinion paper: So what if ChatGPT wrote it? International Journal of Information Management, 71, 102642.
Grimmelikhuijsen, S. (2023). Explaining why the computer says no: Algorithmic transparency affects the perceived trustworthiness of automated decision-making. Public Administration Review, 83(2), 241–262.
Janssen, M., Hartog, M., Matheus, R., Yi Ding, A., & Kuk, G. (2020). Will algorithmic governance cause the downfall of democracy? Government Information Quarterly, 37(3), 101504.
Kim, H. W., & Kankanhalli, A. (2009). Investigating user resistance to information systems implementation: A status quo bias perspective. MIS Quarterly, 33(3), 567–582.
Lapointe, L., & Rivard, S. (2005). A multilevel model of resistance to information technology implementation. MIS Quarterly, 29(3), 461–491.
Mergel, I., Edelmann, N., & Haug, N. (2019). Defining digital transformation: Results from expert interviews. Government Information Quarterly, 36(4), 101385.
Mikalef, P., Conboy, K., & Krogstie, J. (2021). Artificial intelligence as an enabler of B2B marketing: A dynamic capabilities micro-foundations approach. Industrial Marketing Management, 101, 80–92.
OECD. (2024). OECD AI Policy Observatory: Government AI Readiness Index 2024. OECD Publishing.
Simon, H. A. (1997). Administrative Behavior (4th ed.). Free Press.
Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27(3), 425–478.
Venkatesh, V., Thong, J. Y. L., & Xu, X. (2012). Consumer acceptance and use of information technology: Extending the unified theory of acceptance and use of technology. MIS Quarterly, 36(1), 157–178.
Vial, G. (2019). Understanding digital transformation: A review and a research agenda. Journal of Strategic Information Systems, 28(2), 118–144.
Wirtz, B. W., Weyerer, J. C., & Sturm, B. J. (2020). The dark side of artificial intelligence: An integrated AI governance framework for public administration. International Journal of Public Administration, 43(9), 818–829.
Young, M., Bullock, J., & Lecy, J. (2019). Artificial discretion as a tool of governance: A framework for understanding the impact of artificial intelligence on public administration. Perspectives on Public Management and Governance, 2(4), 301–313.
