 My talk will be on AI ethics and how it can be more than a lip service. So we're really in the gutter now in the sense that we know that AI and data-driven practices are increasingly shaping and changing our society. So it shouldn't come as a surprise that this same society wants these AI actors or AI professionals to take responsibility for these impacts, the impact they have with their applications. And with taking responsibility, we mean more than merely being compliant with the law. Obviously, we want our innovations to be compliant with the law. But sometimes this law lags behind or the technology goes too fast, depending on your take. And we need more guidance. And next to that, because the impact of AI can be so fundamental, we also want our AI actors and professionals to have more to exceed a checkbox mentality. I mean, we really want that AI is being developed in a way that it reflects the societal values that we uphold in our democratic society. And that there is reason for, next slide please, that there is reason for this AI ethics to be developed. We can see on a daily basis that we have discussions on deep fakes. So do we want to have deep fakes in our society? It's crucial that we can rely on the trustworthiness of, for instance, pictures or movies. So it just becomes questionable what does this do with our society. We had incidents, indeed, on facial recognition that we know that it doesn't perform well with black people, and especially not with black female females. We had controversies with AI applications that are used as a recruitment tool, which is super interesting, because actually a lot of these recruitment tools are introduced to get rid of human bias, but we found that they actually reestablished biases and, for instance, discriminate against women. So there is this need to rethink and ensure that we have AI applications that are worth wanting from a societal point of view. So AI ethics is a kind of, yeah, how to say it, a kind of development that is taking place, I think five years or so, it's becoming very popular, both in academia as well as in the professional and the public domain. So in academia, ethicists, sociologists are focusing on question of, what does it mean to have responsible AI? Or what does it mean to have a good explanation? But also in the public and the private domain, people are providing and developing tools in order to ensure this ethical AI. Next slide, please. So one important actor in the field in the public domain is the high-level expert group on artificial intelligence. So they have published ethical guidelines for trustworthy AI. And what they've done, so they formulated four ethical principles of respect for human autonomy, prevention of harm, fairness and explicability. And based on these four principles, they also developed requirements and they also developed an impact assessment. So all kinds of tools for people in the field, professionals in the field to ensure that their AI application takes into consideration these important values. Next slide, please. But it's not just in the public domain. So also the companies that are developing these applications are thinking about how to do this in a societal acceptable way. So I could have picked IBM Microsoft, I picked Google, Google AI. So they formulated seven objectives for their AI applications. For instance, being beneficial, socially beneficial, incorporate privacy design principles. And so they use these company values, so to say, to assess their own activities. Next slide, please. So next to public, private actors, we also see academia, but also NGOs formulating all sorts of guidelines and codes of conduct. This is one specifically for data science. And again, we see principles such as avoid harm, maintain accountability and oversight, and increase trustworthiness. Next slide, please. What's interesting is there was a kind of research into all these new ethical guidelines and principles. And they found three principles that reappear throughout all these codes of conduct and principles, namely accountability, fairness, and privacy. So these are the three focal points currently of AI ethics. And other principles that just didn't make it to the top three were justice and explainability, for instance. So if we think about this, what do we think about this actually, right? So is this good? Are we happy that we have all these people, actors engaging in making sure that our AI applications are ethically sound? So you might think, yes, this is brilliant, right? We want to have AI applications that adhere and represent these principles. However, next slide. There are also some, I believe, valid concerns with this approach. And one of the biggest concerns is referred to as ethics washing. So what we do see is that sometimes companies use actually or better even abuse these ethics principles to actually circumvent, to avoid more stricter regulation, legal regulation, for instance. So they say, well, don't bother regulating us. We have these principles. We'll be fine. So we don't need extra regulation. And this is of course not the goal of AI ethics. Next slide. And this is a valid concern also follows from research. Recently, there were 160 AI ethics guidelines analyzed. And it turned out that only 10 have proper enforcement mechanisms. So this means that in practice, when there is a nice opportunity coming up, which might be in conflict with the formulated ethical principles, it's actually quite easy to set aside these principles and go for the opportunity, so to say, because there is not a proper enforcement mechanism. So it's really important to understand that AI ethics is the kind of self regulatory strategy and therefore also surface from these kinds of weaknesses. Next slide. Oh, one slide back, please. Oh, one slide. Oh, we're missing one slide. Okay, it doesn't matter. Just let the slide be what it is. And the other point of critique is that as we saw when we looked at these AI ethics principles, they're all rather like high level and abstract. And what it doesn't really account for is that in real life AI practices, people are taking already responsibility in the sense that ethics is not just about big principles, but also about how professionals in their daily lives, in their daily professional lives already think about values and ethical issues. And I'm happy to say that Chris, who is the next speaker, is actually a moral exemplar of a professional who already shows that there is a lot of people on the floor who actually already take this into account and this kind of knowledge, this kind of bottom up ethical knowledge should be more recognized. And this actually brings me to this slide that's on the screen, the value tabs. A big question is whose values are... You have five minutes left Esther. Perfect. Whose values are we actually embedding? I already heard also in the in the previous stock AI for good. But who is defining good here? This is so important. So we have cultural values. So for instance, the European high level expert group obviously represents our European view. But what about the corporate values of Google? They are more aligned with American, US based values. Obviously they have capitalistic values. They have there's a company. And one domain of values or values that are missing oftentimes are community values, especially the communities that are actually affected by the AI applications. So there is also a call for more participatory AI design where we actually try to already include in the development phase the foreseeable communities that will be affected by the AI applications in order to ensure that also their values are part of the design process. This brings me again to the questions. Question is all lost now for AI ethics. So should we then just stop and do something else? No, I don't think so. First of all, I think it's important to understand that although AI ethics is kind of new, in practice companies, researchers, institutions, government agencies always have been engaged in this kind of decision making and formulating these kinds of values and principles. But it was never very open and public. So what at least AI ethics brings us is that this discussion now is open to more actors. So I think this alone is already very valid. However, the points of critique that I just mentioned are also important to take into account. So we do need to kind of rethink how we want to take the next step. It comes to AI ethics. Next slide, please. Okay, next slide, please. Yes. And so within ISEQ we already discussed a holistic approach to AI, but also within AI ethics. I think a holistic approach would be something to work on. So this is the kind of model that I am working on. It has kind of, it has three layers. And in order to, I believe to have an AI ethics that is more than just window dressing, we have to look at AI ethics on these three levels. So the first level is a technical level. So this is about which values get embedded and how can we ensure that they are not just the values of those in power, but also of those who are affected by the AI application. And I believe we can only do that by focusing as well on the people who are developing, making these AI applications. And especially for Tilburg, it's super important because we are, to a certain extent, educating these professionals. So we have to make sure that they not just develop their technical skills, but also their techno-moral skills and competences. And we do, so that's good. But we can always do it more. And finally, to ensure that it's not just lip service AI ethics, we also have to include the checks and balances on the organizational level. So we need to make sure that there are proper enforcement mechanisms being developed to ensure that throughout the whole process of the AI development, these values are embedded. So to stress, this is really not a pick and choose model. So you really have to take into account all three levels in order to develop a holistic approach to AI ethics. And I believe this could be a way to ensure that AI ethics is more than just lip service. Next slide, please. Thank you for your attention. And this is the literature if you would like to look into that.