|
The Transaction Norm Drift Model in Algorithmic Bank Transaction Monitoring and Its Implications for Auditing in Iran
|
Mohammad Habibpour fatideh *    |
|
|
|
Abstract: (18 Views) |
The use of machine-learning algorithms to monitor bank transactions rests on the assumption that a relatively stable distinction can be drawn between normal and suspicious transaction behavior. This assumption is difficult to sustain in an economy where price levels, nominal transaction values, transaction rules, payment channels, and customer behavior continually change. This article examines how changes in transaction norms affect the reliability and auditability of bank monitoring models. Drawing on an integrative and critical review of the literature, documentary analysis of reports issued by central banks and regulatory authorities, and secondary analysis of data from Iran’s payment system, the article formulates the concept of transaction norm drift. The proposed model distinguishes among three forms of change: behavioral-distributional drift in transaction characteristics, relational-semantic drift in the association between transaction patterns and risk, and institutional-normative drift in the formal and operational definition of suspicious behavior. The model also incorporates the feedback loop among algorithmic decisions, bank interventions, and adaptation by customers and offenders. Within this framework, the reliability of algorithmic monitoring depends on both the model’s technical structure and the distance between the norm learned during the training period and the prevailing transaction norm. In Iran, inflation, rapid changes in the nominal value of transactions, regulatory changes, and shifts among payment channels may cause fixed thresholds and outdated models to produce alert inflation or fail to detect emerging patterns. The scope of auditing should therefore encompass the reference period of the data, peer-group formation, adjustment of nominal variables, threshold setting, drift monitoring, model versioning, and human intervention. The transaction norm drift model provides a framework for evaluating banks’ algorithmic monitoring systems and designing auditing arrangements suited to Iran’s economic and institutional conditions.
|
|
| Keywords: Algorithmic monitoring, artificial intelligence, concept drift, Iranian banking system, model auditing, transaction norm drift |
|
|
Full-Text [PDF 1033 kb]
(7 Downloads)
|
Type of Study: Research |
Subject:
Auditing
Received: 2026/09/1 | Accepted: 2026/10/12 | Published: 2026/09/1
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
