Machine learning-based risk analytics in combating the shadow economy: public indicators on Kaşif and the debate on procedural safeguards
Year: 2026
Type: article
Abstract: In efforts to combat the shadow economy and the use of fraudulent documentation, traditional auditing systems remain inadequate in the face of massive data volumes.This study aims to evaluate the implications of the machine learning-based KAŞİF and KURGAN systems, implemented by the Ministry of Treasury and Finance, for tax auditing.The study employs descriptive document analysis based on official statistics and bulletins.The compiled public indicators suggest that algorithmic governance narrows the time asymmetry between the commission and detection of the offence.The documents report that the system analyzes 4 million taxpayers per day; that, in illustrative cases, a fraudulent invoice network worth ₺246 million was detected; and that a transaction volume of ₺276 billion was reported as "risky" over the last three months.However, turning opaque autonomous risk scoring into administrative sanctions entails procedural risks in relation to the presumption of innocence, proportionality, and legal certainty.Consequently, for the sustainability of algorithmic auditing, the study recommends the adoption of Explainable Artificial Intelligence (XAI) standards and the integration of a human-in-the-loop model into administrative procedures, where final decisions are left to expert auditors. Citation
DOI: 10.56334/sei/9.6.21
Author: Salahattin Altundağ
Institution: Dicle University
Language: English
Subfield: Economics and Econometrics
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7168288917