Twin revolutions in wearable technologies and health interventions delivered by smartphones have greatly increased the accessibility of mobile health (mHealth) interventions. Micro-randomized trials (MRTs) are designed to assess the effectiveness of the mHealth intervention and introduce a novel class of causal estimands called "causal excursion effects." These estimands enable the evaluation of how intervention effects change over time and are influenced by individual characteristics or context. However, existing analysis methods for causal excursion effects require prespecified features of the observed high-dimensional history to build a working model for a critical nuisance parameter. Machine learning appears ideal for automatic feature construction, but their naive application can lead to bias under model misspecification. To address this issue, this paper revisits the estimation of causal excursion effects from a meta-learner perspective, where the analyst remains agnostic to the supervised learning algorithms used to estimate nuisance parameters. We present the bidirectional asymptotic properties of the proposed estimators and compare them both theoretically and through extensive simulations. The results show relative efficiency gains and support the suggestion of a doubly robust alternative to existing methods. Finally, the proposed methods' practical utilities are demonstrated by analyzing data from a multi-institution cohort of first-year medical residents in the United States (NeCamp et al., 2020).
View on arXiv