X's algorithm prioritizes engagement, often showing users content they dislike, impacting Democrats more. Replying to posts has a significant effect on the algorithm, and the former head of product confirmed past algorithmic preferences for replies.
Infact verdict: Mostly True (69/100).
The claims about X's algorithm prioritizing engagement and serving more ragebait to Democrats are supported by a study published in the Proceedings of the National Academy of Sciences. This study indicates that the algorithm prioritizes engagement, often by showing users content they dislike, which aligns with the claim. The claim that replying has an outsized impact on the algorithm is supported by evidence showing that replying significantly boosts engagement. The statement about Democrat users confronting value-misaligned content is corroborated by the study's findings on feedback loops. Lastly, the claim about the former head of product confirming past algorithmic preferences for replies is supported by statements from Nikita Bier, although he notes this is no longer true.
X's algorithm prioritizes engagement above all else when generating a user's For You Page.
The claim is supported by a study published in the Proceedings of the National Academy of Sciences, which found that X's algorithm prioritizes engagement above all else. This is corroborated by multiple sources, including Slashdot and Election Law Blog, which discuss the study's findings.
Fact Check ScoreNone
Fact Check Weight0
Web Consensus Score85
Web Consensus Weight40
Source Quality Score80
Source Quality Weight20
Llm Reasoning Score75
Llm Reasoning Weight40
Llm Reasoning Score Raw75
Weighted Total88
Evidence SummaryStudy in PNAS supports claim; corroborated by multiple sources.
X serves more ragebait to people who say they are Democrats.
The claim is supported by the same study in the Proceedings of the National Academy of Sciences, which found that X serves more ragebait to Democrats. This is corroborated by multiple sources, including Slashdot and Election Law Blog, although the exact reason for this behavior is unclear.
Fact Check ScoreNone
Fact Check Weight0
Web Consensus Score75
Web Consensus Weight40
Source Quality Score70
Source Quality Weight20
Llm Reasoning Score65
Llm Reasoning Weight40
Llm Reasoning Score Raw65
Weighted Total66
Evidence SummaryStudy in PNAS supports claim; corroborated by multiple sources.
Replying is only a fraction of engagement but has an outsized impact on the algorithm.
The claim is supported by evidence indicating that replying significantly boosts engagement, which is a key factor in social media algorithms. This is corroborated by sources discussing the impact of replying on engagement metrics.
Fact Check ScoreNone
Fact Check Weight0
Web Consensus Score70
Web Consensus Weight40
Source Quality Score75
Source Quality Weight20
Llm Reasoning Score75
Llm Reasoning Weight40
Llm Reasoning Score Raw75
Weighted Total69
Evidence SummaryEvidence supports claim; corroborated by multiple sources.
Democrat users confront value-misaligned content by replying to it, which the algorithm learns from.
The claim is supported by the study in the Proceedings of the National Academy of Sciences, which describes a feedback loop where Democrat users reply to value-misaligned content, and the algorithm amplifies this behavior. This is corroborated by multiple sources.
Fact Check ScoreNone
Fact Check Weight0
Web Consensus Score70
Web Consensus Weight40
Source Quality Score70
Source Quality Weight20
Llm Reasoning Score70
Llm Reasoning Weight40
Llm Reasoning Score Raw70
Weighted Total66
Evidence SummaryStudy in PNAS supports claim; corroborated by multiple sources.
X's former head of product confirmed the algorithm favored replies but this is no longer true.
The claim is supported by statements from Nikita Bier, X's former head of product, who confirmed that the algorithm once favored replies but no longer does. This is corroborated by multiple sources, including Slashdot and Engadget.
Fact Check ScoreNone
Fact Check Weight0
Web Consensus Score60
Web Consensus Weight40
Source Quality Score60
Source Quality Weight20
Llm Reasoning Score60
Llm Reasoning Weight40
Llm Reasoning Score Raw60
Weighted Total58
Evidence SummaryStatements from former head of product support claim; corroborated by multiple sources.