Studies have emphasized that empathy is a learnable skill that can be developed through learning from experience. Applying this in the context of human-system interaction, an empathic system is one that automatically acquires an initial knowledge of empathy that is permitted to be incomplete, hence inaccurate and imperfect, but improves this knowledge each time it learns from its interactions with the user until its empathy knowledge matures over time. This paper posits a self-improving empathy learning for user-centric systems. Due to the nature of the problem that centers on incremental experiential learning, this problem becomes a machine learning application. We highlight the need for machine learning techniques that can address the challenge of an empathic system self-improving its empathy model efficiently online from incomplete knowledge through the intelligent use of its experiences. We suggest that the community position itself to investigate empathic reasoning in the context of a self-learning system with machine learning at the core of its AI.