Psychological Reactance to the Algorithmic Management of Online Expressions

As algorithmic management of online expressions becomes increasingly common on digital platforms, it can create tensions between autonomy and control. Autonomy reflects the value individuals place on the freedom to create content, whereas control represents the perceived threat posed by bot moderation to this freedom. To investigate this dynamic, we develop a novel theoretical framework rooted in psychological reactance theory. Through this framework, we examine how individuals respond to bot moderation and under what conditions psychological reactance manifests. Using Wikipedia as an empirical setting, we analyze more than 70,000 U.S. political articles and 17 million article versions contributed by 2.89 million Wikipedia contributors from January 2001 to April 2010. We find that bot moderation on content within a contributor’s domain of attention is associated with psychological reactance, manifested as more slanted political expression. The strength of this effect depends on two factors: (a) the degree to which contributors value their freedom of expression and (b) the extent to which moderation is perceived as threatening. On the one hand, moderation of content outside a contributor’s domain of attention is associated with reduced psychological reactance, whereas contributors with stronger political biases exhibit heightened reactance. On the other hand, repeated bot moderation is associated with stronger reactance, whereas contributors with greater experience in politically sensitive topics display reduced reactance. These findings highlight the need for governance strategies in the algorithmic management of online expressions that balance the enforcement of content standards with the preservation of contributor autonomy. Our study contributes to the understanding of human-bot interdependence and provides insights into designing governance structures that sustain neutrality while fostering constructive individual engagement.

History: Ahmed Abbasi, Senior Editor; Xitong Li, Associate Editor.

Funding: This work was supported by the Hong Kong Research Grants Council [Grant 16504622].

Supplemental Material: The online appendices are available at https://doi.org/10.1287/isre.2022.0446.