How Algorithmic Confounding in Recommendation Systems Increases Homogeneity and Decreases Utility¶
Chaney, A. J. B., Stewart, B. M., & Engelhardt, B. E. (2018). How Algorithmic Confounding in Recommendation Systems Increases Homogeneity and Decreases Utility. ACM, 224-232.
Cited by¶
2 citations across 2 artifacts.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Selection-Visibility Gate
- Recommendation and marketplace systems: an item accrues feedback only where it is shown, and is shown on the strength of feedback it has already accrued, so the exposure policy determines which items ever become eligible to be judged good.
This sourceSimulates the exposure-feedback loop in which what a system shows determines what accrues evidence of being good.
Supported in partVerified against the work's full text
Simulates a recommender feedback loop in which past recommendations shape the data later used to train the system, but does not state the strict 'feedback only where shown' eligibility claim.
“the underlying recommendation model is trained using data that are confounded by algorithmic recommendations from a previously deployed system.”
- Recommendation and marketplace systems: an item accrues feedback only where it is shown, and is shown on the strength of feedback it has already accrued, so the exposure policy determines which items ever become eligible to be judged good.
Mechanisms¶
- Search, Ranking, or Matching Algorithm
- Its characteristic failure is a rich-get-richer dynamic: rank by popularity and the top counterparts accumulate ever more exposure, which earns them more ranking, until a few winners entrench and the rest of the supply starves — the Matthew effect playing out in a results page.
This sourceShows that recommendation feedback loops, including popularity ranking based on prior interactions, can amplify unequal item consumption and concentrate attention on a small set of items.
- Its characteristic failure is a rich-get-richer dynamic: rank by popularity and the top counterparts accumulate ever more exposure, which earns them more ranking, until a few winners entrench and the rest of the supply starves — the Matthew effect playing out in a results page.
Verification¶
Does it exist? Confirmed. This work's DOI resolves to a registered record, which fixes its identity. That is all it fixes.
Does it back the claim? Read against the text for 1 of 2 citations: 1 supported in part. Each verdict is shown under its citation below, with what in the work backs the sentence.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Registry ID ref:4e74e1eb567a · see in the full table