當代網路社群中，以幽默為包裝的「客家迷因」常淪為傳遞族群偏見的載體。本研 究旨在探討Dcard 匿名留言區中，年輕世代如何透過數位物件進行客家意象的微歧視展演與偏見同化。在方法上，本研究採混合研究取徑，結合大數據文本探勘（TF-IDF、 KNN演算法）與質性紮根理論，針對數萬筆之Dcard 留言文本進行深度解剖。 研究發現：第一，客家迷因多採用「半真半假」的訊息結構，挪用真實的勤儉文化 符號，將其極端化為「摳門」的物質化與本質化污名。第二，透過KNN 演算法檢定證實，留言區呈現「機械式洗版（零變異）」與「惡意客製化（高變異）」的雙軌演化，前者用於集體消音，後者用於火力升級。第三，主流群體在面對局內人抗議時，熟練地運用五大「數位中和技術」——幽默偽裝、虛假權威、被害者敏感化、反向自嘲與比例原則挑戰，作為規避道德譴責的防衛修辭。最後，Dcard 的平臺演算法與喧嘩效應進一步催化了多數暴力。本研究不僅補足了數位微歧視與網路語用學的本土實證，更期盼能為當代網路平權與多元包容提供反思契機。;In contemporary online communities， "Hakka memes" disguised as humor frequently serve as vehicles for ethnic prejudice. This study aims to explore how younger generations on the anonymous platformDcardengage in digital microaggressions and biased assimilation through Hakka imagery. Methodologically， a mixed-methodsapproach is adopted， combining big data text mining(TF-IDF and KNN algorithms) with qualitative grounded theory to deeply analyze tens of thousands ofDcardcomments. The findings reveal: First， Hakka memes predominantly employ a "half-true" message structure， appropriating the real cultural symbol of frugality and polarizing it into materialized and biologically essentialized stigmas of "stinginess". Second， KNN algorithm tests confirm a dual-track evolution in the comment sections: "mechanical flooding"(zero mutation) used for collective silencing， and "malicious customization"(high mutation) used to escalate attacks. Third， when facing protests from cultural insiders， the dominant group skillfully utilizes five "digital techniques of neutralization"—humor disguise， false authority， victim sensitization， reverse self-deprecation， and proportionality challenges—as defensive rhetoric to evade moral condemnation. Finally， Dcard′s platform algorithms and the "cacophony effect" further catalyze this majority violence. This study not only fills the gap in local empirical research on digital microaggressions and online pragmatics but also aims to provide an opportunity to reflect on contemporary online equality and diversity.
