Intuitiveness in Active Teaching

While machine learning (ML) gives rise to astonishing results in automated systems, it is usually at the cost of large data requirements. This makes many successful algorithms from ML unsuitable for human-machine interaction, where must learn a small number training samples that can be provided by user within reasonable time frame. Fortunately, tailor they create as useful possible, severely limiting its necessary size—as long know about machine’s requirements and limitations. Of course, acquiring this knowledge turn cumbersome costly. raises question how easy are interact with. In work, we address issue analyzing intuitiveness certain when actively taught users. After developing theoretical framework property algorithms, introduce an active teaching paradigm involving prototypical two-dimensional spatial task method judge efficacy interactions. Finally, present discuss large-scale study into performance strategies 800 users interacting with two prominent our system, providing first evidence role intuition important factor impacting interaction.