Conducted in the Sukolilo and Gubeng Districts of Surabaya, the study examines how digital systems are changing the way tax authorities assess taxpayers’ financial activities and compliance risks. The findings are relevant to Indonesia’s digital tax administration reforms because the success of technological transformation should not be measured solely by a system’s ability to detect irregularities. Tax audit systems must also ensure that decisions can be explained, reviewed, and corrected fairly.
AI Is Changing How Tax Authorities Assess Taxpayers
Digital transformation has shifted tax administration from processes that primarily rely on manual document examination toward systems that use integrated data, predictive analytics, and algorithms. These technologies can match transaction records, identify unusual patterns, develop risk profiles, and determine which taxpayers should receive further examination.
According to the Organisation for Economic Co-operation and Development (OECD) in its Tax Administration 2024 report, 80% of tax administrations across 52 jurisdictions covered in the report had adopted big data. Meanwhile, 52.9% used AI for risk assessment, and 49% applied it to detect tax evasion and fraud.
These capabilities offer governments opportunities to improve oversight and allocate audit resources more efficiently. However, algorithmic systems also introduce challenges. Incomplete data, classification errors, or biased historical patterns may influence how taxpayers are assessed and categorized.
In this context, fiscal reality is more than a collection of figures representing income, transactions, and tax liabilities. It is also an institutional representation shaped by data, regulatory frameworks, auditors’ professional judgment, and technological recommendations about whether a taxpayer is considered compliant or high-risk.
Wibowo examines this issue by connecting AI use in tax audits with procedural fairness, trust in tax authorities, and ethical tax compliance consciousness. This approach places the relationship between people and technology at the center of responsible tax administration.
Study Combines Interviews and a Survey
Wibowo employed a mixed-methods approach combining qualitative and quantitative research. The study began with interviews to explore the experiences and perspectives of individuals involved in digital tax administration. The qualitative findings were then used to develop indicators and questionnaire items for statistical testing through a survey.
The qualitative phase involved 15 informants: three tax officials or auditors, four tax consultants, four corporate tax or accounting practitioners, and four individual or business taxpayers.
The quantitative phase included 200 respondents from Sukolilo and Gubeng Districts in Surabaya. Survey data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM), a statistical technique used to examine relationships among multiple factors within a research model.
The study measured seven constructs: AI dominance, fiscal reality reconstruction, algorithmic transparency, human oversight, perceived procedural fairness, trust in tax authorities, and ethical tax compliance consciousness.
The findings should be interpreted within the study’s scope. The research examined respondents’ perceptions rather than directly comparing official tax audit records or experimentally measuring changes in actual taxpayer behavior. Its results therefore primarily explain how AI governance relates to perceptions of fairness, institutional trust, and ethical compliance.
Human Oversight Plays a Stronger Role in Perceived Fairness
One of the study’s main findings is that algorithmic transparency and human oversight are positively associated with perceived procedural fairness. However, human oversight demonstrated a stronger relationship with fairness than transparency alone.
The statistical model reported a path coefficient of 0.38 for human oversight and procedural fairness. Algorithmic transparency had a coefficient of 0.31, while fiscal reality reconstruction had a coefficient of 0.21. All three relationships were statistically significant.
These figures represent relationship coefficients within the research model, not percentage increases in fairness. They suggest that meaningful human involvement and understandable algorithmic decisions are both relevant to how taxpayers perceive the audit process.
The interviews reinforced this conclusion. Fourteen of the 15 informants, or 93.3%, highlighted the importance of algorithmic explanations in supporting procedural legitimacy. Thirteen informants, or 86.7%, emphasized the continuing need for human review.
The findings indicate that AI-generated recommendations should not automatically determine the outcome of a tax audit. Auditors need to understand why a system flags a transaction as risky, verify the accuracy of the underlying information, and have the authority to correct recommendations that do not reflect the facts.
Transparency must also go beyond simply informing taxpayers that AI has been used. Meaningful explanations should help affected parties understand the basis of an assessment, identify possible errors, and obtain opportunities to provide clarification.
Fairness and Trust Support Ethical Tax Compliance
Wibowo found that perceived procedural fairness was strongly associated with trust in tax authorities. The relationship had a path coefficient of 0.55 and was statistically significant. Trust in tax authorities, in turn, had a coefficient of 0.47 in its relationship with ethical tax compliance consciousness.
Procedural fairness also had a direct relationship with ethical compliance consciousness, with a coefficient of 0.24.
These results suggest that taxpayers consider more than the government’s ability to detect violations. The way an audit is conducted, the opportunity to explain transactions, and confidence in fair decision-making are also associated with taxpayers’ willingness to fulfill their obligations.
An important finding emerged when the researchers examined the direct relationship between AI dominance and ethical compliance consciousness. This relationship was not statistically significant, with a coefficient of 0.08 and a p-value of 0.190.
In other words, greater reliance on AI does not automatically translate into a stronger moral commitment to paying taxes. Within the study’s model, technology’s influence was more evident through its relationships with fiscal reality reconstruction, procedural fairness, and institutional trust.
The mediation analysis supported this pattern. Human oversight had an indirect effect coefficient of 0.098 on ethical compliance consciousness through procedural fairness and trust. Algorithmic transparency had an indirect effect coefficient of 0.080. Both relationships were statistically significant.
These findings distinguish compliance motivated primarily by fear of audits or penalties from compliance grounded in a belief that the tax system operates fairly and deserves public trust.
Implications for Indonesia’s Digital Tax Administration
The findings offer several practical directions for the governance of AI-supported tax audits.
First, tax authorities should provide meaningful explanations of why particular transactions or tax reports are classified as risky. These explanations should help taxpayers understand the basis of an examination while maintaining appropriate protections for confidential tax information.
Second, human oversight must be substantive rather than symbolic. Auditors need sufficient information, relevant expertise, and the authority to review and override inaccurate system recommendations. Merely approving an algorithmic recommendation without examining it does not provide effective accountability.
Third, taxpayers should have accessible procedures for correcting inaccurate data, submitting explanations, and requesting reviews of decisions they consider incorrect. Such mechanisms can help reduce the risk that data errors or algorithmic misclassifications lead to unfair examinations.
Fourth, the success of AI should not be evaluated exclusively through audit speed, the number of detected risks, or tax revenue. Authorities should also consider the quality of explanations, consistency of procedures, opportunities to correct mistakes, data protection, and public trust.
These recommendations are particularly relevant as governments expand digital public services. Technology can strengthen administrative capacity, but its long-term legitimacy depends on how fairly and responsibly it is implemented.
Nevertheless, the findings have limitations. The study involved 15 informants and 200 respondents from two districts in Surabaya, using non-probability sampling. The results may not represent all Indonesian taxpayers. Future research involving broader geographic coverage, longitudinal observations, and actual tax compliance records could help establish whether similar relationships hold in practice.
Author Profile
Danny Wibowo is affiliated with Sekolah Tinggi Ilmu Ekonomi Indonesia Surabaya (STIESIA), Surabaya, Indonesia. His article examines issues related to taxation, digital tax administration, artificial intelligence governance, procedural fairness, and ethical tax compliance. His academic degree and professional title were not specified in the article information provided.
Research Source
Article title: Fiscal Reality Reconstruction in Artificial Intelligence Dominated Tax Audit Ecosystems and Its Consequences for Human Ethical Compliance Consciousness
Author: Danny Wibowo
Affiliation: Sekolah Tinggi Ilmu Ekonomi Indonesia Surabaya (STIESIA), Indonesia
Journal: International Journal of Management and Business Intelligence (IJMBI)
Publication year: 2026
DOI: https://doi.org/10.59890/ijmbi.v4i5.43
Journal website: https://journalijmbi.my.id/index.php/ijmbi
0 Komentar