Emprical Side of Responsible AI – Crash course
In the context of the course Introduction to Focus Areas we created teaching and learning materials to introduce the topic of Responsible AI. The course consists of a two hour lecture to introduce fundamentals, followed by a two hour session that focuses on applying the newly acquired knowledge.
This crash course introduces the fundamentals of empirical research methods, which are used to develop, evaluate, and assess how users interact with AI systems, to lead to a Responsible AI design. In addition to learning about the different steps of experimental research, including defining the research question and hypotheses, specifying variables, and conducting statistical analyses, examples of current human-AI research are discussed. In the crash course participants are learn about human-centered approaches that take into account user needs and evaluate how users perceive additional information are necessary. A key focus is a developed crowdsourcing experiment as part of ENKIS to illustrate the fundamentals that students need to consider when conducting a small online experiment via a crowdsourcing platform.
For the second practical part, we provide three scenarios focusing on how to responsibly design human-AI collaboration. Each scenario involved potential stakeholders facing a challenge that required the help of an expert knowledgeable in empirical research and data science. Our slide instruct how much time is needed for each step, i.e., discussing and understanding the scenario in groups and group discussions.
