When working as Engineer in Data Science & Artificial Intelligence, you will identify and solve societal and technical problems. The problems you face will present themselves not in a clearly defined technical manner – as taught in the technical trajectories. Rather, you will encounter them in the context of an existing organization or system where specific stakeholders such as citizens, workers, customers, patients, farm animals, or wildlife face a problem. Your obligation as an engineer is to understand this context, it’s requirements and objectives and the different aspects and objectives that define what desirable and undesirable solutions are. More often than not, you will see that the problem and its objectives are different than initially presented to you by stakeholders in the light of available data, that data is incomplete or insufficient. You will also see that objectives from different stakeholders and context constraints contradict each other, both in being achievable through technical solutions, but also in their impact on the stakeholder and the context: a solution that works seemingly well according to various technical quality criteria may have potentially disastrous consequences for users, or people affected by this solution.
As an Engineer in Data Science & Artificial Intelligence, you must be aware of all the requirements and objectives of the problem you are aiming to solve, taking the full context into account.
The following two core courses expose you to this fundamental trait of being an Engineer in Data Science and Artificial Intelligence:
- 0LM190 - hilosophy & Ethics of Artificial intelligence,
- 2AMI90 - Data Intelligence Challenge.
Next to those we also have 2IMP40 - Empirical Methods in Software Engineering in this track.
The central topic of Philosophy and Ethics of Artificial Intelligence is the very concept of “intelligence” and how AI and Data Science impact society, and how can we make sure that this impact is positive and perceived as positive. You will learn how to evaluate the claim that certain kinds of machines can actually create, understand, and think, as well as to ensure that people affected by these machines have the appropriate control over them. In this way, you will understand more deeply the possibilities and limitations of data science and artificial intelligence and learn how to evaluate and mitigate its risks.
The Data Intelligence Challenge is a challenge-based course you take towards the end of your studies, to integrate the technical skills you learned before in a real-life challenge. You receive an open-ended problem from a real-world context affecting various stakeholders. Working in teams of 6-8 students you learn all aspects of working as an Engineer in Data Science & Artificial Intelligence that sit “between” the technical skills: identifying stakeholder objectives; working with complex, inconsistent and incomplete data; defining the right technical problem and approach; evaluating and communicating your work in terms of the original problem context and its impact on various stakeholders. You will specifically learn ways of reflecting on your work regarding assumptions, adequacy, and impact on the problem context and stakeholders. The main topic of the course revolves around fundamental concepts or Reinforcement Learning.
The Empirical Methods in Software Engineering brings data scientists together with computer scientists. After taking this class, you should be able to design and conduct an empirical study, i.e., identify appropriate research questions, select research methods that can be used to answer these research questions and justify the choice, apply these research methods, e.g., design interview protocols and user surveys, mine data from online repositories, and perform appropriate qualitative and quantitative analyses. Finally, you should be able to draw conclusions from empirical data and discuss issues that might have threatened their validity. In this way you can see how the techniques you have learned in data science courses can be applied and should be adjusted in the software engineering context.