Objectives

Despite popular notions of “filter bubbles” and “echo chambers” contributing to radicalization, little evidence exists to support these hypotheses. However, social structure social learning theory would suggest a hereto untested interaction effect.

Methodology

An RCT of new Twitter users in which participants were randomly assigned to a treatment of “filter bubble” (personalization algorithm) suppression. Ego-centric network and survey data were combined to test the effects on justification for suicide bombings.

Findings

Statistically significant interaction effects were found for two proxies of the echo chamber, the E-I index and modularity. For the treatment group, higher scores on both factors decreased the likelihood for radicalization, with opposing trends in the control group.

Conclusions

The echo chamber effect may be dependent on the filter bubble. More research is needed on online network structures in radicalization. While personalization algorithms can potentially be harmful, they may also be leveraged to facilitate interventions.

中文翻译：

---

检查过滤泡和回声室对自由基化的交互作用

目标

尽管“过滤气泡”和“回声室”有助于激进化的流行概念，但几乎没有证据支持这些假设。然而，社会结构社会学习理论会提出一种迄今未经检验的互动效应。

方法

一项针对新 Twitter 用户的 RCT，其中参与者被随机分配到“过滤气泡”（个性化算法）抑制的治疗中。以自我为中心的网络和调查数据相结合，以测试对自杀性爆炸理由的影响。

发现

对于回声室的两个代理，EI 指数和模块性，发现了统计上显着的交互效应。对于治疗组，两个因素的较高得分降低了激进化的可能性，而对照组的趋势相反。

结论

回声室效应可能取决于过滤泡。需要对激进化中的在线网络结构进行更多研究。虽然个性化算法可能有害，但它们也可能被用来促进干预。