Governance of artificial intelligence applications in a business audit via a fusion fuzzy multiple rule-based decision-making model

Financial Innovation

Table 8 Gap ratio of AI in internal auditing by modified-VIKOR

Dimensions/criteria	Weights (global)	Weights (local)	Enterprise
Dimensions/criteria	Weights (global)	Weights (local)	\({\text{A}}_{1}\)	\({\text{A}}_{2}\)	\({\text{A}}_{3}\)	\({\text{A}}_{4}\)
AI application strategy (A)		0.287	0.282	0.303	0.361	0.319
AI competencies (\(a_{1}\))	0.075	0.261	0.357	0.329	0.371	0.314
AI outcomes and expected level (\(a_{2}\))	0.096	0.334	0.243	0.286	0.357	0.343
Ability of the AI provider (\(a_{3}\))	0.057	0.199	0.143	0.229	0.271	0.214
AI cognition of senior executives (\(a_{4}\))	0.059	0.206	0.386	0.371	0.443	0.386
AI governance (B)		0.308	0.362	0.299	0.352	0.379
Techniques of AI governance (\(b_{1}\))	0.082	0.266	0.343	0.257	0.357	0.400
AI activities and decisions (\(b_{2}\))	0.077	0.250	0.386	0.329	0.329	0.357
AI accountability and oversight (\(b_{3}\))	0.086	0.279	0.371	0.271	0.343	0.386
The necessary skills and expertise of AI responsibilities (\(b_{4}\))	0.063	0.205	0.343	0.357	0.386	0.371
Data infrastructure and data quality (C)		0.237	0.206	0.194	0.232	0.192
Data accessibility (\(c_{1}\))	0.065	0.274	0.214	0.243	0.229	0.229
Information privacy and security (\(c_{2}\))	0.079	0.333	0.171	0.129	0.257	0.186
Completeness, accuracy, and reliability of the data (\(c_{3}\))	0.093	0.392	0.229	0.214	0.214	0.171
Human factor (D)		0.169	0.318	0.304	0.326	0.336
AI test (\(d_{1}\))	0.057	0.337	0.257	0.271	0.271	0.243
Human error and biases (\(d_{2}\))	0.067	0.396	0.371	0.343	0.343	0.343
Black box (\(d_{3}\))	0.045	0.266	0.314	0.286	0.371	0.443
Total gap (\(S_{p}\))	–	–	0.295	0.276	0.322	0.311

For example enterprise \({\text{A}}_{1}\), AI application strategy (A): 0.282 = (0.357 × 0.261) + (0.243 × 0.334) + (0.143 × 0.199) + (0.386 × 0.206), and total gap ratio value: 0.295 = (0.282 × 0.287) + (0.362 × 0.308) + (0.206 × 0.237) + (0.318 × 0.169). The gap ratio \(r_{pj}\) is calculated by \([r_{pj} ]_{P \times n} = {{[\left( {\left| {f_{j}^{aspiration} - f_{pj} } \right|} \right)} \mathord{\left/ {\vphantom {{[\left( {\left| {f_{j}^{aspiration} - f_{pj} } \right|} \right)} {\left( {\left| {f_{j}^{aspiration} - f_{j}^{worst} } \right|} \right)}}} \right. \kern-\nulldelimiterspace} {\left( {\left| {f_{j}^{aspiration} - f_{j}^{worst} } \right|} \right)}}]_{P \times n}\) for enterprise (alternatives) p = 1,2,…,m and criteria j = 1,2,…,n