Many CIOs wanting to ensure their AI systems comply with regulatory frameworks and laws will bring in their legal or risk professionals to lead a team in evaluating a system’s compliance. However, ethics in AI is slightly different from following clear rules and standards. In fact, some legal versus ethical conundrums arise when it comes to AI. For example, it is legal to try and increase engagement with online content but there may be unethical tactics used to do so. As an example, addictive algorithms that feed content which plays on people’s psychological vulnerabilities, such as envy, guilt, and insecurity. Encouragingly, this manipulative design is starting to get noticed and addressed as per the recent European Commission’s ruling that Tik Tok’s addictive design breaches EU law.
Another example is the copyright of AI generated content. AIs are trained on a vast corpus of training data, a lot of which is aggregated from online content. The intellectual property rights of the generated output is often unclear. While it may be considered fair use in some countries, and therefore not illegal, there is an ethical consideration of whether the original works should be accredited or not. And whether human content creators are being deprived of earnings. This relates more to ethics and transparency, one of the key requirements in several AI governance frameworks.
In fact, without ethics, there is the risk that a law-abiding AI could carry out unanticipated harm because the law is absent in situations where we may want to guide behaviour. Laws can also be immoral if one looks at the past and how discrimination and inequality were legalised.
Ethical language can also be too ambiguous for legal interpretation, whereas AI ethicists are able to understand the zones of discretion. According to the World Economic Forum, the role of AI ethical officers is on the rise especially as ethical AI policies are becoming legislated and enacted. Therefore, ethical impact assessments should be included as part of evaluating whether AI systems are responsible and safe.
In the world of marketing, customer experience and product design, there are specific impacts that need to be factored in. For example, whether manipulative design results in impulsive and compulsive shopping, whether biased algorithms perpetuate inequality, whether customer data is used transparently and whether advertising content generated by AI is original. An ethical impact assessment specifically developed for these fields is able to identify the level of risk of certain practices and where campaigns are misaligned with ethical boundaries.
