Common sense or censorship: How algorithmic moderators and message type influence perceptions of online content deletion

Abstract

Hateful content online is a concern for social media platforms, policymakers, and the public. This has led high-profile content platforms, such as Facebook, to adopt algorithmic content-moderation systems; however, the impact of algorithmic moderation on user perceptions is unclear. We experimentally test the extent to which the type of content being removed (profanity vs hate speech) and the explanation given for its removal (no explanation vs link to community guidelines vs specific explanation) influence user perceptions of human and algorithmic moderators. Our preregistered study encompasses representative samples ( N = 2870) from the United States, the Netherlands, and Portugal. Contrary to expectations, our findings suggest that algorithmic moderation is perceived as more transparent than human, especially when no explanation is given for content removal. In addition, sending users to community guidelines for further information on content deletion has negative effects on outcome fairness and trust.

On impersonal justice: libraries' neutrality as an act of change

“…That implies that librarians have the obligation to let any voice be heard. Censorship, online (Gonçalves et al, 2021) or in physical libraries (Kann-Christensen and Pors, 2004) only makes some information more attractive for groups who leverage it to gain consensus. However, the transfer of knowledge and free formation of opinions also entails communicating facts not rumours, opinions or misinformation and critical points of view against the risks of some ideologies, propaganda and harmful information practices.…”

“…That implies that librarians have the obligation to let any voice be heard. Censorship, online (Gonçalves et al. , 2021) or in physical libraries (Kann-Christensen and Pors, 2004) only makes some information more attractive for groups who leverage it to gain consensus.…”

Does distrust in humans predict greater trust in AI? Role of individual differences in user responses to content moderation

“…These features must be at a very granular level to be implemented for machine detection, which can be challenging in the context of content moderation (Molina et al, 2021). 2 When users think that AI is unable to detect subtleties of human language, they tend to perceive AI as unfit to classify usergenerated content, because incorrect classification can lead to content being taken down, an issue many believe to be a violation of freedom of their speech, tantamount to censorship (Gollatz et al, 2018;West, 2018).…”

“…Similarly, Jhaver et al's (2018) analysis revealed that one of the biggest challenges of human classification is that moderators have "different viewpoints and tolerance level, and what might offend one person may be perfectly reasonable to another" (p. 22). In other words, although humans do indeed have the capacity to contextualize and empathize with a post (Gollatz et al, 2018), users acknowledge that humans have their own biases and experiences that can also influence their decision-making (Jhaver et al, 2018). In fact, in one set of studies, participants expressed more acceptance of AI compared with humans for content moderation because decisions made by AI are at least based on the same constant rules being applied (Binns et al, 2018;Jhaver et al, 2018), and, thus, are "statistically fair" (Binns et al, 2018: 9).…”

Moral disciplining: The cognitive and evolutionary foundations of puritanical morality

“…Other forms of action delegations may imply that we let machines speak for us, in order to maintain decency of speech (Hancock, Naaman, & Levy, 2020). Puritan norms would require people to discipline themselves into suppressing emotions like anger or infatuation, so that their speech be free of hostility or innuendo; machine puritanism would give people leave to feel whatever they feel, in exchange for letting machines rewrite their emails, text messages, and social media posts to eliminate every trace of inappropriate speech (Gonçalves et al, 2021). In a more extreme form of this norm, people may be expected to let a machine block their communications if the machine detects that they are in too emotionally aroused a state.…”