Blog: What knowledge and tools should the Supervisory Board have in order to be able to assess the risks of applying artificial intelligence?

The growth (in importance) of AI

The increasing use of algorithms in various business sectors has led to the need for effective supervision and regulation to mitigate potential risks. A Supervisory Board has the task of supervising the strategic decision-making of management and ensuring compliance with laws and regulations. The increased usage of (generative) Artificial intelligence has recently complicated these tasks further. This new technology can provide value in several ways: problem solving, efficiency, decision making, optimization, and scalability. Yet, the Supervisory Board may face challenges in exercising effective supervision of algorithms. Challenges might include: understanding complex algorithms, identifying potential bias and discrimination, ensuring transparency, and safeguarding privacy protection. In addition, the Supervisory Board must provide the right expertise and skills to understand and assess the technical aspects of algorithms.

It is important to first demystify and grasp the concept of AI. Essentially, it is powerful software that predicts and is completely dependent on the patterns that can be found in a large mountain of data. Moreover, the technology is human motivated. However, due to the large scale at which algorithms work and hence the possible negative consequences, organizations are faced with ethical dilemmas.
Addressing these ethical dilemmas requires a combination of technical, legal and social measures. It is important to build accountability, transparency and ethical considerations into the design, implementation and use of algorithms. In this blog I focus on the administrative measures, namely the supervisory role of the Supervisory Board. As a result, are you as part of a (supervisory) board looking for more directions and control over AI? Do take into account the below aspects.

• With regard to privacy and data protection, the primary supervisory authority lies with the Data Protection Officer (DPO) and the Privacy Officer (PO). The relationship between the DPO, PO and the Supervisory Board may vary depending on the specific organization. In general, however, we can set the following preconditions for both functions: 1) Independence, the DPO and the PO must be independent in the performance of their tasks and responsibilities. 2) The Supervisory Board can and should make use of the advisory expertise of both officers in making decisions regarding data protection and privacy policy. 3) Reporting obligation, the DPO and the PO should  be obliged to report regularly to the Supervisory Board on compliance with data protection laws and regulations within the organization. 4) Cooperation: Although the DPO and the PO must be independent, cooperation with the Supervisory Board and other stakeholders is of great importance. the DPO and the PO can work together with the Supervisory Board to promote the implementation and compliance with data protection measures. They should strive for a shared vision and understanding of data protection and privacy as a strategic and ethical issue within the organization.

• To prevent discrimination and bias, algorithms must be tested in advance and continuously monitored on those characteristics. For the Supervisory Board, it is a must to check what measures the company has taken to prevent bias and discrimination.  

• Given the large scale on which the output of an algorithm can have an effect, it is essential for a Supervisory Board to have a good understanding of the liability and responsibility for the use of algorithms in the company. The company must have insight into the responsibility and liability. As with bias and discrimination, it is good if a company has an ethical assessment framework that shows the determination of responsibility and liability. In addition, it is good for a company to set up an algorithm register.        

• An organization must always strive for transparency and comprehensibility in how an algorithm arrives at an output. It is therefore important to determine in which context an algorithm is used. Using an AIto assess people's performance is probably ethically irresponsible, but an AI that chooses between machine A or B is justifiable. This can also come to the fore when using a well-functioning ethical assessment framework.      

• For an organization, it is almost daily work to determine the consequences for autonomy and human involvement in the event of process changes. The usual business procedures such as a request for advice for the works council are suitable for testing this in the case of the use of algorithms. It is necessary that the works council is informed at the moment that human involvement is going to be affected. In addition, it is important that the works council members have sufficient knowledge of the algorithms that will be implemented.

• It is often difficult to determine whether the implementation of an algorithm has a social impact, especially the social impact outside the direct context of the company is difficult to determine. Here too, an ethical assessment framework can work well to determine whether the impact is acceptable or desirable.

• A Supervisory Board also has the responsibility to test whether a company responds well enough to new developments and innovation. This is certainly the case in the field of Artificial Intelligence because developments there are going fast. That is why it is wise to devote a meeting of the Supervisory Board to innovation at least once a year, in which new developments are discussed and what the company intends to do or not do with them.

• Due to context changes such as market conditions and policy adjustments, algorithms can evolve. As a result, it is necessary that the algorithms are periodically monitored and registered (tested and possibly retrained). Hence, it is necessary that there is a specialist department that deals with this. For a Supervisory Board, it is good to regularly ask the question what this department has adapted to these algorithms, given certain context changes. For every company it is also advisable to have a register in which Algorithms used in the company are registered.

Conclusion

Due to the large increase in the use of Artificial Intelligence and algorithms in particular, the importance of good supervision of this is a must. Especially when we look at the risk of harm to a company when things go wrong on the ethical aspects of Artificial Intelligence. In addition, there is also an increasing need for good supervision from legislation, such as the AI act that is in the making from the EU. Given the high expectations in the field of Artificial Intelligence, it is important for the Supervisory Board to be able to serve as a sparring partner to promote innovation in the field of Artificial Intelligence in a responsible manner within the company.     

Frank Ferro
Director Insights
PostNL

This blog is a summary of a longer paper, for the longer version or additional question and remarks feel free to contact Frank.