Ethical Aspects and Challenges for Interactive Task Learning

ions from low-level data. Moreover, being able to ask the human instructor for help (e.g., explanations or alternatives for action) does not exist in a data-driven approach; if the answer is not in the data, no statistical process in the world will be able to extract it. However, there are potential downsides to ITL as well, in particular for humans. Humans would have to engage with an autonomous machine over possibly extended periods of time in a way that would lead to successful knowledge acquisition for the machine. Some effects would be limited to the teaching interaction itself, whereas others could extend far beyond the teaching context. In the following, we examine three classes of ethical aspects as they arise in ITL: 1. What is being taught and what are the associated risks? 2. What are the dynamics of human-machine instruction? 3. What effects will ITL have on human instructors and society?