As the application of artificial intelligence expands to support more and more critical decisions about the economy and society, the more concerns will grow as to whether the predictions of such systems are valid, impartial, and socially acceptable.
The basis for the validity, reliability and acceptance of decisions based on artificial intelligence systems
by Thanos S. Chonthrogiannis
©The law of intellectual property is prohibited in any way unlawful use/appropriation of this article, with heavy civil and criminal penalties for the infringer.
For an artificial intelligence system to be valid and reliable for its predictions it must be characterized by the following parameters:
1. Be legal so that it always complies with current laws and regulations.
2. Be ethical to conform to ethical values.
3. Be reliable since technical errors or safety gaps can lead to unintended consequences and independent intentions.
An artificial intelligence system looks for correlations in the data it analyzes, which it then uses to come up with the requested prediction. It then adjusts itself according to whether the prediction was in practice correct or incorrect.
For this reason, and to predict ‘failures’ in the predictions of artificial intelligence systems, the following should be applied:
1. Reliable data on system training to draw objective and non-discriminatory conclusions from the system.
A prerequisite for this is the application of good data management practices so that the undertaking or organization can maintain good quality data while complying with the relevant regulatory framework.
2. Adequate security mechanisms shall be in place, in the case of unreliable data or the proposal/provision of unlawful or socially unacceptable decisions of the system, the mechanisms shall ensure the legitimate and ethical conduct of the system.
3. Adequate human control & supervision including continuous monitoring and evaluation of the system with a view to early detection of deviations from socially acceptable behaviour.
Through the implementation of the above policies, businesses and organisations can build strong and timeless relationships of trust with customers, partners, supervisors and other stakeholders, while at the same time they can gain more confidence and understanding by making the best use of artificial intelligence technology, rapidly promoting their digital transformation by achieving long-term sustainable development.
