Just-in-Time Adaptive Interventions (JITAIs) are AI-based personal informatics systems that provide the users with personalized feedback based on their self-tracked data. The performance of these systems is influenced by how users engage with them. Therefore, this position paper aims to facilitate discussion about addressing user interactions that may lead to missing or incomplete data and, subsequently, impact the personalization aspect of JITAIs. We posit open questions about two such interactions: episodic abandonment and retrospective tracking. As one approach to answering these questions, we suggest that the algorithms in these systems require active human engagement for more accurate information about the users' current state and their desired personalization.