The AI would answer questions, but we would have to check whether the answers are correct.

It is important to design AI systems in such a way that errors or unintended behavior can be detected in the first place. Then it must also be ensured that a human can intervene competently.

No question

AI is not always right. Real human discourse will become more important compared to an artificial flood of information.

There are numerous examples of applications in occupational safety, for risk identification, to simplify documentation, for data analysis of sensor data, etc.

No question

No question

If necessary, standard-compliant and certified AI procedures for specific functions could be considered “legally compliant” insofar as they meet the requirements of the legal provisions.

We work on open questions regarding the design of AI and assistance systems to the best of our ability in research projects, and we also implement our findings as a standardization player and in advisory processes for further regulation.

These are conceivable applications; in addition to optical sensors, other sensors are also conceivable. This results in large and complex data volumes that may be particularly suitable for analysis with AI. In addition to other requirements, the reliability of the analysis results must be guaranteed. This is also a topic of our research, for example.

The work of a SIFA can certainly be supported by AI applications. However, the advisory process is always operationally and personally individual and influencing factors, situations and conditions must be taken into account for which an AI is not prepared. I don’t see this being replaceable in the foreseeable future.

Theoretically at least, but for us it is still a research topic as to how the reliability of such systems can be assessed and guaranteed.

There are two levels here:

Health and safety professionals who use AI tools wisely and safely in their work can become more productive and achieve even better prevention. However, it is necessary to know the possibilities and limitations of AI applications very well in order to be able to correctly assess the significance of an AI recommendation. It would be important to convey a basic understanding of AI functions and exemplary applications during training. However, in order to use a specific tool in the company, it is also necessary to know its exact characteristics and possible applications, which requires application-related training.

At the second level, occupational health and safety experts should also be able to recognize AI-related risks – as well as opportunities for good design solutions with AI – in companies. The risks can be of a technical nature (e.g. malfunctions), but can also relate to psycho-social aspects (e.g. stress when dealing with the systems) and should then also be able to be dealt with using risk assessments. Training should also prepare students as well as possible for this challenge of practical occupational safety.