So far the 6 sections follow roughly the data scientist activities: data, bias/modeling, evaluation, usage in production. However that is not that clear, as:

• section 2 and 3 cover together modeling, training, bias, evaluation, without a clear split
• section 4 covers responsibilities and reproducibility
• section 6 is transversal

We propose here to improve the sections scope to offer a clearer structure. The update sections could be:

• A first set of sections dedicated to the data scientists activities:
  • 1. Data preparation and protection
  • 2. Modelisation, training, evaluation
  • 3. Deployment, monitoring/MLOps
• A second set dedicated to organisation aspects, processes and externalities, more global to the org doing the assessment:
  • 4. Organisation, governance, responsibilities - internal to the org
  • 5. Organisation, governance, responsibilities - related to external parties
  • 6. Negative externalities

These new sections would include the following evaluation elements:

1. Data preparation and protection: 7 elements

• 1.1 and 1.2
• 1.5 and 1.6
• 2.1
• 3.1 to 3.3

2. Modelisation, training, evaluation: 8 elements

• 1.7 and 1.8
• 2.2 to 2.4
• 3.4 and 3.6

3. Deployment, monitoring/MLOps: 5 elements

• 3.5
• 4.3
• 5.1 to 5.4

4. Organisation, governance, responsibilities - internal to the org: 6 elements

• 1.3 and 1.4
• 3.7
• 4.1 and 4.2
• 4.4 (specific to internal parties)

5. Organisation, governance, responsibilities - related to external parties: 5 elements

• 1.9
• 4.4 (specific to external parties)
• 4.5 and 4.6
• 5.5

6. Negative externalities: 3 elements

• 6.1 to 6.3