In this work, we analyze how human gaze during reading comprehension is\nconditioned on the given reading comprehension question, and whether this\nsignal can be beneficial for machine reading comprehension. To this end, we\ncollect a new eye-tracking dataset with a large number of participants engaging\nin a multiple choice reading comprehension task. Our analysis of this data\nreveals increased fixation times over parts of the text that are most relevant\nfor answering the question. Motivated by this finding, we propose making\nautomated reading comprehension more human-like by mimicking human\ninformation-seeking reading behavior during reading comprehension. We\ndemonstrate that this approach leads to performance gains on multiple choice\nquestion answering in English for a state-of-the-art reading comprehension\nmodel.\n