Human-in-the-loop machine learning: Reconceptualizing the role of the user in interactive approaches

Abstract

The rise of intelligent systems and smart spaces has opened up new opportunities for human–machine collaborations. Interactive Machine Learning (IML) contribute to fostering such collaborations. Nonetheless, IML solutions tend to overlook critical factors such as the timing, frequency and workload that drive this interaction and are vital to adapting these systems to users’ goals and engagement. To address this gap, this work explores users’ expectations towards IML solutions in the context of an interactive hydration monitoring system for the workplace, which represents a challenging environment to implement intelligent solutions that can collaborate with individuals. The proposed system involves users in the learning process by providing feedback on the success of detecting their drinking gestures and enabling them to contribute with additional examples of their data. A qualitative study was conducted to evaluate this use case, where participants completed specific tasks with varying levels of involvement. This study provides promising insights into the potential of placing the Human-in-the-Loop (HitL) to adapt and reconceptualize the users’ role in interactive solutions, highlighting the importance of considering human factors in designing more effective and flexible collaborative systems between humans and machines.