HUMAN CONTENT MODERATION: For Online Safety and Trust

1. What is a content moderator? and What is content moderation?

Content moderators are on the front line in the fight to help businesses maintain positive reputations and good relationships with their customers. In a highly competitive marketplace, the content that is hosted on a company’s platform can help them stand out, for better or worse. User-generated submissions, in the form of reviews, videos, social media posts, or forum discussions can be valuable to a brand. However, much like anything in this world, there is also a dark side. No platform is safe from those who seek to do harm by uploading unwanted subject matter such as spam, indecent photos, profanity, and illegal content. When content is submitted by a user to a website, that piece of content will go through a screening process (the content moderation process) to make sure that the content adheres to the regulations of the website. Based on predefined criteria, content moderators will then decide whether a particular submission can be used or not on that platform. Unacceptable content is removed based on its inappropriateness, legal status, or its potential to offend.
The content moderator’s role is, therefore, to moderate and filter the content to keep a company’s social media pages, blogs, and website safe for users.

2. Human content moderation provides better service to the community.

Online platforms are increasingly using automated procedures to take action against inappropriate (or illegal) material on their systems. There are some fantastic algorithms that have been instrumental in quickly locating and deleting obvious harmful content, such as hate speech, pornography, or violence. AI has done a lot to help automate content moderation and has allowed the process to move more quickly. 

This attempt to move toward more automation shouldn’t be a surprise. For years, tech companies have been pushing automated tools as a way to supplement their efforts to fight the offensive and dangerous content that can inhabit online platforms. However, it’s becoming clear that these algorithms are really not up to the task when the situation demands more nuance. AI is unable to understand differing contexts, such as satire, scams, intentional disinformation, or a genuine desire to inform rather than shock or offend. Most importantly, technology lacks the ability to take action, offer help or intervene, particularly when there is an immediate danger to the safety of a child or adult. Considering the young age of many internet users, the need for constant and vigilant moderation is apparent.

As a result, despite advances in technology, human-led content moderation is becoming more important, not less. The human content moderator is a highly skilled individual, often making decisions that can not only mean the difference between success and failure of a brand but also between life and death.

3. Case Study: Youtube, Facebook, Twitter, & Google experiment with automated content moderation

YouTube – the greater use of AI moderation had led to a significant increase in video removals and incorrect takedowns

As 2020 saw the unprecedented and rapid spread of coronavirus around the world, tech companies were faced with a dilemma: How to ensure their platforms would be safe to use and free of harmful and offensive content while also reducing the need for content moderators to come into the office. As a possible solution, many social media companies handed the reins over to AI content moderation to make decisions on content that might violate their policies.

YouTube announced these changes back in March 2020 and warned that offending content will begin to be removed automatically without human review, and therefore “users and creators may see increased video removals, including some videos that may not violate policies”.

However, even YouTube couldn’t anticipate just how overzealous the AI  content moderation turned out to be in its attempts to spot harmful content. By September 2020, YouTube reported that the greater use of AI moderation had led to a significant increase in video removals and incorrect takedowns. Between April and June 2020, approximately 11 million YouTube videos were removed, double the usual amount.

Since then, YouTube has brought back more human moderators to ensure more accuracy with its takedowns.

Facebook – a lack of human moderators who were able to discern what broke the platform’s rules

Facebook reported similar figures, with 22.5 million removals reported due to hate speech in the second quarter of 2020, more than double the amount compared to the first quarter, before they set the AI to work. In fact, just one day after Facebook launched its machine-learning content moderation, there was an upsurge in user complaints that the platform was making mistakes. Facebook’s algorithm began blocking a whole host of legitimate posts and links, including posts with news articles related to the coronavirus pandemic and flagging them as spam.

In Syria and other hotspots, social media is used by human rights campaigners, activists, and journalists to report and document potential war crimes. However, as AI struggles to understand context and intention, accounts were taken off Facebook due to the graphic content of their posts.

In contrast, the amount of disturbing content that was taken down portraying child exploitation and self-harm actually fell by at least 40% in the second quarter of 2020 because of a lack of human moderators who were able to discern what broke the platform’s rules.

In November 2020, Facebook called its human content moderators back into the office to ‘review its most sensitive content’.

Twitter – AI has a lot more to learn before it can match the consistency of human-lead moderation.

Twitter informed users early in 2020 that they would increase their reliance on machine learning content moderation to remove “abusive and manipulated content.” However, in a blog post from March 2020, they acknowledged that AI would be no replacement for human moderators.

And as expected, there were wildly inconsistent results. While Twitter happily reported to their shareholders that half of all tweets on the platform deemed to be abusive or in violation of policy had been removed by its automated content moderation tools, campaigners fighting against racism and anti-Semitism in France noticed a more than 40% increase in hate speech on Twitter. As less than 12% of those posts were removed, it’s obvious that AI has a lot more to learn before it can match the consistency of human-lead moderation.

Google Search- One of its content moderation AI tagged African Americans as gorillas.

Since long before the pandemic hit last year, Google has been experimenting with artificial intelligence to police online content. Results have been mixed at best. According to a paper published in 2019, a machine-learning tool used to scour comments sections on major online newspapers for hate speech had developed a racial bias. The algorithm began to highlight comments written in African-American vernacular as problematic, regardless of the actual content of the comment. This was not the first time Google experienced bias problems with its AI. Recently, one of its photo-search tools tagged African Americans as gorillas.

The findings shine a light on machine learning’s susceptibility to bias, in a large part due to certain groups of society being underrepresented in the data used to train machine-learning systems. It also highlights the limitations of machine learning as a tool to moderate content. AI too often fails at understanding context and as the study shows, needs a lot more work if it’s to avoid accusations of over-policing the voices of minority groups online.

4. Content Moderation: In-house or Outsourced?

In-house moderation

Running an in-house content moderation team can provide many benefits. The proximity of your moderation team allows you to have a more hands-on approach and enables you to have more control over your moderation operations every day. However, this means you are accountable for the results of your own moderation policies since there’s no third party doing the job for you. Another, not inconsiderable downside to in-house is cost. The expense of setting up a content moderation department can be quite daunting for many large companies, let alone small or medium-sized firms. You will need to invest a considerable amount of time and money in building content moderation tools, API, hiring and training your content moderation team. Forming a team of professional moderators takes time, from hiring and training to performance feedback and monitoring. Finding the right people to work with is not always easy, especially if you need people quickly to deal with a suddenly increased workload. Then there is the outlay on all the latest hardware and software to ensure your team can do the job as expected.

Outsourced moderation

For many companies, outsourcing is a preferable option over in-house moderation. What follows are the main reasons that companies choose to work with Leap Steam for their content moderation needs:

– Work with Experts

By outsourcing your content moderation, you can expect the assistance of experienced moderators. You can rely on Leap to have a roster of professional moderators who can provide you with the moderation support that you need for your business, based on your own guidelines and standards. 

– Easy Scalability

Outsourcing makes it easy for businesses to scale up when they need additional moderators quickly. Where it would normally take weeks, or even months, to attract, interview, hire, and then train a content moderator, Leap can take just a few days. With an already existing pool of experienced and well-trained personnel, it’s simply a matter of moving moderators from one campaign to the other when needed. 

– Keep Costs in Line

Maintaining an in-house team of moderators can be a drain on a business’s expenses. Outsourcing eliminates the need to provide expensive office space, desks, computers, and moderation software tools for staff.

– Focus on Core Activities

Core activities are generally defined as strategic tasks that improve customer value and drive profits. The core activities of any business should focus on providing superior quality goods and services to its customers. When businesses start to grow, however, other tasks can distract teams from their core goals. By outsourcing jobs such as content moderation, companies can allow their core management and decision-making teams to maintain their focus on their most important goals, objectives, and tasks. 

5. How to Outsource your content moderation to Leap Steam?

Every platform that handles user-generated content needs to have a moderation process in place. It protects your users, it protects your brand, and it protects your company from legal liabilities.