How can the quality of AI be ensured?
If AI performs only specific subtasks within a system, or if a system is AI-based but contains conventional components (i.e., non-AI components), or if there are interfaces with conventional systems, the conventional parts of these systems—as well as the interfaces between AI and non-AI components—are tested using conventional methods. Existing testing procedures and quality metrics therefore remain applicable here as well, and a mix of different methods and approaches will become established. When it comes to AI, the testing process must also analyze the risks associated with its use. Testing is conducted to mitigate these risks or, at the very least, to make them transparent.
In this case, too, testing efforts should be focused where the risks are greatest (risk-based testing). These are largely the same risks as in any other system: What happens if the system fails or if the system fails to correctly recognize input information? What happens if the system provides incorrect data, makes incorrect decisions, or reacts incorrectly?
However, with AI, additional AI-specific risks arise. For example, the AI-based system may make incorrect decisions due to incorrect or insufficiently representative training data. Here, too, the corresponding risks must be analyzed and mitigated using appropriate quality metrics for AI.
Furthermore, due to the nature of AI-based systems, additional AI-specific risks may arise, such as a system that objectively meets all requirements but is not purchased or used because the users—or a certain segment of them—do not trust decisions made by artificial intelligence.
Regarding the question, “What risks can arise when using AI, and how can these be managed through appropriate quality assurance and testing procedures?” imbus can advise you and offer support in optimizing your development processes.
Ihr Ansprechpartner bei imbus
Mr. Tilo Linz