Can you trust an Artificial Intelligence (AI)? How is trust established?
This is a very relevant question when important decisions are based on AI predictions or recommendations. The European Commission provided a self-assessment and nailed down 7 requirements which make up a Trustworthy Artificial Intelligence System. The full document you can find below.
We have summarised the headlines for you:
1.) Human agency and oversight
Translation: How are humans involved in the decision making process?
2.) Technical robustness and safety
Translation: What is the Plan B if something goes wrong?
3.) Privacy and data governance
Translation: What happens with the sensitive data?
4.) Transparency
Translation: What is the plan, so that humans understand the basis for the prediction/recommendation?
5.) Diversity, non-discrimination and data governance
Translation: How can I be sure that the recommendations are not biased and unfair?
6.) Societal and environmental well-being
Translation: How to avoid any negative impacts on society, environment or democracy?
7.) Accountability
Translation: How can I audit an Artificial Intelligence system?
Author: DI. Mag. Markus Loinig
Email: markus@senzoro.com