arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2608.26804stat.MLcs.LGstat.ME

基于因果模型的增量推荐

Incremental Recommendation via Causal Models

  • Hologen
  • Imperial College London(帝国理工学院)
  • Spotify
  • University College London(伦敦大学学院)

机构由 AI 辅助整理,请以论文原文为准。

Athanasios Vlontzos, David Gustafsson, Michael O'Riordan, Ciarán M. Gilligan-Lee

AI总结:

该研究针对推荐曝光资源有限问题,提出双阈值定向策略的因果模型,在Spotify用户A/B测试中减少7%推荐曝光且不降低内容消费,证明因果模型泛化性更优。

AI中文摘要:

推荐曝光是有限资源,若向本会自然发现内容的用户推送推荐,既无增量价值,还会挤占其他推荐位。我们通过将现有生产环境推荐模型扩展为因果架构来解决该问题,所用留存数据来自常规实验基础设施已收集的数据,无需新增数据采集。核心挑战在于处理组与对照组观测的归因窗口存在差异:处理组用户在短期直接响应窗口内归因于推送流,对照组用户则在多天窗口内归因于自然流,该不匹配使朴素处理效应减法失效。我们提出双阈值定向策略,仅在推送流概率高且自然流概率低时才推送推荐。在数百万 Spotify 用户的生产规模A/B测试中,该策略将推荐曝光减少7%,且未出现统计显著性的整体推荐内容消费下降。我们进一步表明,与生产基线相比,结合留存数据的联合训练提升了处理组头部的校准度,这可作为因果模型比仅基于观测数据训练的模型能学习到更具泛化性表示的证据。

英文摘要:

Recommendation impressions are a finite resource, hence delivering a recommendation to a user who would discover the content organically yields no incremental value and displaces other recommendations that could. We address this by extending an existing production recommendation model to a causal architecture using holdback data that is already collected as part of routine experimentation infrastructure, requiring no new data collection. A central challenge is that attribution windows differ between treated and holdback observations: treated users are attributed a stream within a short direct-response window, while holdback users are attributed organic streams over a multi-day window. This mismatch makes naive treatment-effect subtraction invalid. We resolve this with a dual-threshold targeting policy that delivers a recommendation only when the probability of a treated stream is high and the probability of organic stream is low. In a production-scale A/B test on millions of Spotify users, this policy reduces recommendation impressions by 7% with no statistically significant reduction in overall recommended content consumption. We further show that joint training with holdback data improves calibration of the treated head relative to the production baseline, and argue this can be taken as evidence that causal models learn more generalisable representations than models trained on observational data alone.

补充信息

↑