159.算法的六副面孔：它是如何从处理数据，变成定义我们是谁的

Summary

This [[QizhulouYanBinke|起朱楼宴宾客]] episode by [[DavidWeng|大卫翁]] opens a year-long algorithm-and-media series by reading Tarleton Gillespie’s [[TheRelevanceOfAlgorithms|The Relevance of Algorithms]] alongside Kyle Chayka’s Filterworld. The source treats algorithms as media and power structures, not just software: they decide what enters the data field, predict what users want, rank what counts as relevant, perform objectivity, reshape user and creator practice, and calculate publics that can become identity categories. Its practical conclusion is not total algorithm rejection, but clearer coexistence through [[AlgorithmicDiversityDividend|multi-platform diversity]], source mixing, and more deliberate Feed Curation.

Key Claims

• The episode frames algorithm sociology as a missing layer in Chinese public discussion: technical progress is visible, but the social, political, philosophical, and identity effects of algorithmic media are less developed.
• [[PublicRelevanceAlgorithms|Algorithms of public relevance]] matter because they do not only process knowledge; they also certify which knowledge becomes visible, credible, searchable, or socially discussable.
• Algorithmic Inclusion Patterns / 算法包含模式 decide which people, topics, books, bodies, and experiences enter the searchable or recommendable field; the Amazon Fail example shows that classification and ranking can hide exclusion without formal deletion.
• [[AlgorithmicPredictionLoop|Prediction loops]] build a behavioral “digital twin” and train both model and user: likes, skips, watch time, and “not interested” controls convert complex people into measurable, predictable profiles.
• Algorithmic Relevance Assessment / 算法相关性评估 is necessarily partly opaque and dynamic. The source argues that ranking weights, A/B tests, anti-manipulation rules, and PageRank-style weighting make digital democracy closer to unequal voting than simple one-person-one-like ranking.
• [[AlgorithmicObjectivityPromise|The promise of algorithmic objectivity]] gives platforms legitimacy through mathematical neutrality, trend labels, best-match language, and a visible but incomplete backstage.
• Algorithmic Entanglement / 算法与实践纠缠 means users and creators change how they speak, tag, title, time, and package themselves so platforms can recognize them; creators can become traffic optimizers before they notice it.
• Calculated Publics / 计算出的公众 are publics assembled by model inference rather than voluntary association, making “people like you” a route by which platforms feed users group identity, taste, and worldview.
• The source qualifies simple [[InformationCocoon|information cocoon]] panic: algorithmic polarization is real, but many users mix Douyin, Xiaohongshu, YouTube, podcasts, and other sources, producing a partial Algorithmic Diversity Dividend / 算法多样性红利.

Key Quotes

“数字双胞胎” — the episode’s image for the algorithmic user profile.

“选择的消失” — the travel and recommendation-loop problem.

“隐形室友” — the closing metaphor for unavoidable algorithmic coexistence.

Connections

• Tarleton Gillespie, The Relevance of Algorithms, Kyle Chayka, and Filterworld — main intellectual sources used by the episode.
• Marshall McLuhan / 麦克卢汉 and Walter Benjamin / 本雅明 — media-theory references used to explain algorithms as perception-shaping media.
• Public Relevance Algorithms / 公共相关性的算法, Algorithmic Inclusion Patterns / 算法包含模式, Algorithmic Prediction Loop / 算法预判循环, Algorithmic Relevance Assessment / 算法相关性评估, Algorithmic Objectivity Promise / 算法客观性承诺, Algorithmic Entanglement / 算法与实践纠缠, Calculated Publics / 计算出的公众, and Algorithmic Diversity Dividend / 算法多样性红利 — central concepts created from the episode’s six-face framework and practical countermeasure.
• PageRank Search Relevance, Recommendation System Productization, Recommendation Distribution Advantage, and AI Ranking Reinforcement — existing ranking and recommendation concepts extended by the episode’s sociological frame.
• Information Cocoon / 信息茧房, Group Polarization / 群体极化, Attention Industrialization, Algorithmic Labeling, Algorithmic Desire Preemption / 算法欲望预支, and Personalization As Social Identity — existing social and psychological concepts clarified by the source.
• Amazon, Google, Spotify, TikTok, Instagram, Douyin, and Xiaohongshu — platform examples or adjacent algorithmic surfaces named or implicated by the discussion.

Contradictions

• None identified. The source extends the wiki’s algorithm and attention branch while qualifying its own strongest danger story: algorithmic cocoons and polarization matter, but multi-platform use means the cocoon is often leaky rather than total.