She's 28, Loves God and Her Family, and Might Be the Reason You Can't Have Kids Emma Waters is leading a national fight against fertility tech--with her hot husband's permission, of course. By 5 am every morning, Emma Waters is on the sofa with a blanket, plotting America's baby-making future. For two hours, she stays off her phone, inspired by, the self-help bestseller. At 7:30, she does emails. Waters, 28 years old, is a monk of productivity who makes motherhood look like all-natural Adderall. At 8 am, it's time for breakfast and the Bible. Waters tries to live her life according to the mantra "Love God, get married, have babies." She asks that I not use the real names of her two daughters for their safety, even though they've been printed elsewhere. I'll call them Gertie and Ophelia, which are close enough. Ophelia, the baby, runs around in white tights, shirtless, clutching a pink purse; Gertie's doll won't stop crying. There isn't much to say about Jack, other than that he's hot, is studying to be a preacher, and has a big desk at home, much bigger than Waters', featuring a bust of Nietzsche. In the same room, Waters has a pull-out secretary where she takes meetings, not far from a portrait of a glowering President Trump . The only thing Waters doesn't really do is exercise; she hasn't found the time. Not many people outside of DC have heard of Waters, who mostly works from home in a dreary Pennsylvania suburb named after a long-offshored manufacturer. Yet Waters is a factory for policy papers and op-eds extolling the harms of unregulated reproductive technology, probing the moral issues that liberals don't want to touch with a 10-foot turkey baster.

Short- and long-term outcomes of an algorithm often differ, with damaging downstream effects. A known example is a click-bait algorithm, which may increase short-term clicks but damage long-term user engagement. A possible solution to estimate the long-term outcome is to run an online experiment or A/B test for the potential algorithms, but it takes months or even longer to observe the long-term outcomes of interest, making the algorithm selection process unacceptably slow. This work thus studies the problem of feasibly yet accurately estimating the long-term outcome of an algorithm using only historical and short-term experiment data. Existing approaches to this problem either need a restrictive assumption about the short-term outcomes called surrogacy or cannot effectively use short-term outcomes, which is inefficient. Therefore, we propose a new framework called Long-term Off-Policy Evaluation (LOPE), which is based on reward function decomposition. LOPE works under a more relaxed assumption than surrogacy and effectively leverages short-term rewards to substantially reduce the variance. Synthetic experiments show that LOPE outperforms existing approaches particularly when surrogacy is severely violated and the long-term reward is noisy. In addition, real-world experiments on large-scale A/B test data collected on a music streaming platform show that LOPE can estimate the long-term outcome of actual algorithms more accurately than existing feasible methods.