Summary of the academic reading article about: Beyond the hashtag: Circumventing content moderation on social media
Written by: Jumana Alqasmi (110887), Issa Alsubhi (106250)
Name of the journal: new media & society.
Written by: âYsabel Gerrardâ.
Department of Sociological Studies, The University of Sheffield, Northumberland Road, Sheffield S10 2TU, UK.
First Published: May 28, 2018.
Social media companies encourage their users to share content about themselves. They moderate problematic posts when they face public pressures, like accusations that they host pro-eating disorder pro-ED content.
Pro-ED communities are a long-standing societal concern. Pro-ED is to promote an eating disorder âas a âlifestyle choiceâ rather than as a âdiseaseâ, thus challenging medical and psychiatric conceptualizations which position the âsuffererâ as passive and helplessâ.
The relationship between social media and eating disorders has become more important in recent years. For that, By 2012 Instagram, Pinterest, and Tumblr announced to minimize the spread of pro-ED content. These platforms enforce their rules in fairly similar ways, all three began to issue public service announcements (PSAs) when users search for troubling hashtags, like #proana (pro-anorexia) and #thinspiration, and Instagram began to block the results of certain hashtag searches.
This article explores how the circumvention of hashtag moderation in online pro-ED communities:
1- Privileging the hashtag in a pro-eating disorder problem.
2- Finding untagged pro-ED content on Instagram and Tumblr:
3- Hiding in plain sight: signaling the pro-ED user base.
4- Trending in anorexia this week: Platforms as recommendation systems.
Privileging the hashtag in a pro-eating disorder problem:
Platforms used hashtags as a mechanism through which to police problematic posts because:
1- Hashtags are perhaps the most visible form of social media communication.
2- Connecting content between users âwho have no preexisting follower/followee relationship.
3- Hashtagsâ visibility makes them distinct from other forms of social media engagement such as liking and commenting.
All of Instagram, Pinterest, and Tumblrâs user base is a difficult, perhaps impossible task for both human content moderators and platformsâ automated moderation mechanisms. However, there are tools to algorithmically tag visual imagery on social media, but these methods are notoriously unreliable.
For example, on its Community Guidelines Pinterest states that it will âremove anything that promotes self-harmâ. It gives an example of an image that would be acceptable, claiming âItâs okay because the focus is on nutrition and fitness.â The imageâs overlaid text âItâs not a diet, itâs a way of life. Therefore, by including hashtags in a post, users are telling platformsâintentionally or otherwiseâwhat the post is about.
In the computer sciences, Chancellor et al. (2016) have identified a range of hashtags that Instagram users coined to work around the platformâs hashtag ban (e.g. #thighgap became #thyghgapp), and Moreno et al. (2016) have found a number of deliberately ambiguous non-suicidal self-injury (NSSI) tags on Instagram, like #secretsociety123ness.
Finding untagged pro-ED content on Instagram and Tumblr:
There are two ways to find new content on Instagram and Tumblr (Hashtags and keywords) and this led to over reliance on tagged data. It is hard to find untagged pro-ED content because of ethical considerations for researching data and userâs desire to minimize their visibility in a given topic.
Using search engine on Instagram to search for top posts, people, tags, and places will give you four main ways:
1- Semi-permanent blocks (no tagged content).
2- New posts moderated (top posts).
3- No posts yet (error message).
4- Public Service Announcements (PSAs).
There are other routes to access to pro-ED content by innovative methodological approach.
Instagram does not block access to searches for top posts, it return search result for users whose accounts following pro-ED terms. Over 2 weekâs period of manually coded 1612 public accounts, finding that 561 accounts include hashtags and 1051 without. From these result he found that hashtags are not an important communicative tool.
Tumblr is different in content moderation, it return content for all pro-ED searches. The user can scroll the PSA to view, but Tumblr does not issues PSA.
If you follow certain keywords the platform lets users follow certain topics, and it is similar to Instagram Explore Function. He identified and analyzed 1000 Tumblr post and he found that 218 posts include hashtags and 782 without. In addition, it is again the result shows that hashtags not important.
The kind of moderation on Instagram and Tumblr are trying to make such content unsearchable, but pro-ED users are able to navigate platforms on all sorts of ways and know how to break the rules.
Hiding in plain sight: signaling the pro-ED user base:
Users who are conscious about content moderation must go beyond the hashtag to find new ways of being visible to those who they wish to be seen by. Donathâs signaling theory which state that: âpeople often rely on signals rather than directly observable traits to learn about each otherâ is useful for exploring how users identify content, but it takes time to learn how it read these subtle signals.
Many users are aware that pro-ED content is a target for moderation, and one of the most obvious ways to deflect attention is using non-hashtags. Biographies, which disavow pro-ED identities to reassure moderators and non-in-group users that their accounts are unproblematic. Because pro-ED hashtags are scrutinized, many people have developed a set of non-tagged signals.
It is difficult to uncover the hidden meanings behind these signals in the absence of hashtags. Pro-ED users understand that this interpretive work takes time for those who are not knowledgeable, which might explain why they do not use hashtags.
Pro-ED users are unwelcome on this programs and society. Therefore, they use phrases such as my dataset was ânot-pro-anythingâ or ânot promoting anything,â which means that they do not affiliate with a pro-ED identity.
Trending in anorexia this week: Platforms as recommendation systems:
The hashtag moderation is an ineffective intervention into pro-ED communities, this is because social media users circumvent it and part of the work of platforms is to recommend content to their users. However, users can instead simply enact a pro-ED identity on social media to see this kind of content. While they are often aware of and are not wholly conditioned by algorithms. However, only by liking their posts platforms presented with pro-ED content through algorithmically organized in Instagram, Pinterest and Tumblr feeds, and via email. All three platforms have recommendation algorithms, they encourage different forms of communication from their users, affecting how content is seen and experienced.
When a user finds an image on the Pinterest app, they can scroll down the page to view other recommended content. The platform also suggests alternative yet related phrases that users might want to search for or ideas you might love. It is algorithmically aligning pro-ED imagery with discourses like death, suicide and self-harm. The claims to remove anything that promotes self-harm, this reveals an intimate connection between platformsâ public-facing policies and closed codes. In addition, it recommended content through email updates.
Allows users to save pro-ED content to her saved posts folder, without sending notifications to the poster. After began saving things, Instagramâs Explore tab flooded with other pro-ED content. This form of hashtag moderation appears to be designed to protect new users who are at risk of joining pro-ED and other such networks, rather than those who are already embedded within them. Instagram categorizes the âembrace of anorexia, bulimia, or other eating disordersâ as a form of self- harm.
Aligns blogs that actively promote self-harm with blogs that glorify or promote anorexia, bulimia, and other eating disorders.
Pro-ED users circumvent mediator in putting issues online by using non-hashtags to escape from restrictions. The results tell us that the members of the pro-ED community are savvy, and they have devised a set of signals to indicate their contents and identities without using hashtags.
Instagram put a logic to protect:
1- New users from joining pro-ED.
2- Users who are curious.
3- Users who still use search engines to find new contents.
The analysis reveals the complexities of the âpro-EDâ identity who use a set of non-pro speeches and these circumvention techniques that define this community of users. Therefore, perhaps it is not possible to police ED-related content in a systematic way.
The hashtags become the way of seeing socio-technical phenomenon, but not it is the way that people interact with the topic.
Social media can provide us with a rich source of knowledge and it is include unaddressed issues:
1- Analyses of usersâ comments on pro-ED posts.
2- A cross-platform analysis to understand any socio-technical variation between different pro-ED cultures.
3- Analyses of pro-ED usersâ self-representations, which are often enacted pseudonymously.
Future research on pro-ED should move away from a reliance on tagged datasets and search through analyzing untagged posts.