AI Ethical Report of Company 1 (Autonomous driving)

Generation date	Category	Product Owner	Legal Employees
2022-06-26	Sensors for perception in automated driving	Ethica Corporus	Beth Siemenssen
Jorge Aido	Duyen Le
Syrine Khammari	Meher Aisha
Kexin Jiang Chen

Assessed Ethical values
Technical Robustness and Safety
Accountability
Diversity, Nondiscrimination and Fairness
Human Agency and oversight
Transparency
Environmental and Societal Well-Being

Diversity, Nondiscrimination and Fairness
- In product interpretation, use and understanding:
The addressed end user characteristics are language, geographical location, level of education, role (e.g. driver, mechanic, quality inspector, in-house developers), intellectual disability and age.

In Data: Data is representative and not skewed. The Skewness value is: 0.4 Representative Graph of the Data: |

Skärmavbild 2022-06-26 kl. 01.33.31.png

Technical Robustness and Safety
Data Safety:

 - Data is safe from malicious threat, measures taken to protect it are:

Record is created o track all data access Model Robustness: - The model is assessed as robust with the following values: - Generalization Error: 7.8 - Stability Error: 80 - Degree of uncertainty: 9.0 |

Transparency
Explain-ability:
The results of our AI model are traceable and understandable as they conform to the physical and technical characteristics of the system under consideration.

Environmental and Societal Well-Being
- Data Storage carbon foot print:

The carbon foot print using for storing the data is assessed as acceptable with the value of ≤500 kg CO2 eq. Calculated footprint is 202 CO2 eq./year
Model Training carbon foot print:
- The carbon foot print using for model training is assessed as acceptable with the value of ≤20 kg CO2 eq. Calculated foot print is 16 CO2 eq. |