Algorithm Adaptation Bias in Recommendation System Online Experiments
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916925944102912 |
|---|---|
| author | Zheng, Chen Zhao, Zhenyu |
| author_facet | Zheng, Chen Zhao, Zhenyu |
| contents | Online experiments (A/B tests) are widely regarded as the gold standard for evaluating recommender system variants and guiding launch decisions. However, a variety of biases can distort the results of the experiment and mislead decision-making. An underexplored but critical bias is algorithm adaptation effect. This bias arises from the flywheel dynamics among production models, user data, and training pipelines: new models are evaluated on user data whose distributions are shaped by the incumbent system or tested only in a small treatment group. As a result, the measured effect of a new product change in modeling and user experience in this constrained experimental setting can diverge substantially from its true impact in full deployment. In practice, the experiment results often favor the production variant with large traffic while underestimating the performance of the test variant with small traffic, which leads to missing opportunities to launch a true winning arm or underestimating the impact. This paper aims to raise awareness of algorithm adaptation bias, situate it within the broader landscape of RecSys evaluation biases, and motivate discussion of solutions that span experiment design, measurement, and adjustment. We detail the mechanisms of this bias, present empirical evidence from real-world experiments, and discuss potential methods for a more robust online evaluation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_00199 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Algorithm Adaptation Bias in Recommendation System Online Experiments Zheng, Chen Zhao, Zhenyu Information Retrieval Machine Learning Online experiments (A/B tests) are widely regarded as the gold standard for evaluating recommender system variants and guiding launch decisions. However, a variety of biases can distort the results of the experiment and mislead decision-making. An underexplored but critical bias is algorithm adaptation effect. This bias arises from the flywheel dynamics among production models, user data, and training pipelines: new models are evaluated on user data whose distributions are shaped by the incumbent system or tested only in a small treatment group. As a result, the measured effect of a new product change in modeling and user experience in this constrained experimental setting can diverge substantially from its true impact in full deployment. In practice, the experiment results often favor the production variant with large traffic while underestimating the performance of the test variant with small traffic, which leads to missing opportunities to launch a true winning arm or underestimating the impact. This paper aims to raise awareness of algorithm adaptation bias, situate it within the broader landscape of RecSys evaluation biases, and motivate discussion of solutions that span experiment design, measurement, and adjustment. We detail the mechanisms of this bias, present empirical evidence from real-world experiments, and discuss potential methods for a more robust online evaluation. |
| title | Algorithm Adaptation Bias in Recommendation System Online Experiments |
| topic | Information Retrieval Machine Learning |
| url | https://arxiv.org/abs/2509.00199 |