This paper proposes a new method for algorithmic recourse when complete causal knowledge may be unavailable. The paper is well motivated, is of practical interest, and is technically sound. It also makes a very interesting and strong contribution in the space of algorithmic recourse and therefore we recommend an acceptance. We would strongly encourage the authors to address the following points in the rebuttal: 1. Please compare the proposed algorithm to prior work by Mahajan et. al. 2019 and/or other recourse generation algorithms. 2. Results are currently show only on synthetic data. Please try and incorporate at least one experiment with a real world dataset.