Inclusive AI – Newsfeeds and Social Media: How Data Annotators Create a More Inclusive Social Media Experienc

As part of our Inclusive AI series, we’re excited to highlight the impact of your work as data annotators. In our first article, we covered the crucial role you play in artificial intelligence (AI): you offer accurate data labeling to ensure models perform as intended plus diverse perspectives to prevent biased models. In the next part of our series, we’re narrowing down how your contributions help in different industries. Today, we’re focusing on newsfeeds and social media, and how your work directly contributes to a more inclusive online experience.

Why Social Media Needs Diverse Annotators

When you log into your favorite social media platform and scroll through your feed, you expect to see content that reflects your interests and that’s generated by the people you value in your network. With more than half of the world’s population considered active users of social media around the globe, you can imagine the work social media companies must do to ensure that the content you see is personalized to just you.

How do social media companies achieve this? Behind every newsfeed is a machine learning algorithm that relies on training data to understand what content different groups of people are interested in seeing. For example, the algorithm may learn that people interested in camping are often also interested in hiking and start showing related content to those users. But to start making these connections, the algorithm needs a lot of data on what its users want to see and what they don’t. That’s where data annotators come in.

At Appen, we have our global crowd of annotators who can support social media AI needs. Social media companies value the range of demographics, languages, and geographies that our annotators offer because these often reflect the diversity of their users. If annotators can share what they want to see on their newsfeeds, then the algorithm will come to learn what a user with a similar profile to the annotator may want to see as well. Because social media is so global, there’s a much larger user base to represent, so annotators from diverse demographics and geographies are absolutely essential to building an inclusive social media algorithm.

Inclusive AI - Newsfeeds and Social Media

Social Media Annotation: Real-life Examples

Appen’s annotators have participated in many social media projects for our clients. These have contributed to a more representative, inclusive social media experience for countless users around the world. Here are a few standout case studies that demonstrate the full power of our global contributors:

Improved Search Functionality

A social network provider was looking to improve their search engine. Their users used the search engine to find people, posts, news, and more. The social network provider wanted to train their search engine on new, labeled data to improve its accuracy across these categories. When they first began working with Appen, they started with a pilot of just 80 annotators in our crowd based in the U.S. Our annotators focused on rating and measuring the site’s search functionality for news items, identifying areas where improvement was needed.

Impressed with how fast our annotators were able to ramp up and deliver quality results on the project, the client moved forward with expanding the work to cover other types of search, including social, video, image, and trending topics. Eventually, the client extended the project to four markets around the globe to meet the growing demands of their users.

What stood out to the client, even when the project was still only in the U.S. was how diverse the rating feedback was. They appreciated the variety of perspectives our annotators had to offer, as these perspectives were an excellent representation of the client’s diverse user base. Without the global crowd, this client wouldn’t have been able to expand to other markets as well.

For more details, read the full case study.

Better Understanding of User Content

When a social media company needed large amounts of training data to improve its tool for understanding user content, they looked to Appen and our global crowd. Their tool intended to identify the intent, sentiment (think attitude, whether positive or negative), and entities (think people, places, and things) in user-generated content. The purpose was to use this knowledge to improve the help center, videos, and other features of the platform.

The training model needed data in the form of thousands of phrases that covered all the ways a user might input requests. Appen’s crowd was there to help: in just two months, over one million data samples were collected from annotators around the world using a variety of languages and slang. Thanks to the geographic and demographic diversity of our annotators, Appen was able to provide the client with the wide range of data samples needed to make a high-performing model.

For more details, read the full case study.

More Personalized Newsfeeds

Users of a leading social media company were demanding more relevant content in their newsfeed. The company reached out to Appen for help in collecting data from individuals that represented the profile of their users. Appen’s annotators provided personalized input, following strict requirements and a complex quality system. They rated newsfeed items on key measurements, like importance and impact of the content.

The resulting data was fed into the client’s algorithm to improve its ability to customize newsfeed content. Because Appen’s annotators came from such different backgrounds, experiences, and geographies, the client could deliver on this goal and provide a newsfeed that was more reflective of each of their user’s unique interests.

For more details, read the full case study.

Looking Ahead

We highlight the above examples simply to show how important it is for our annotators to reflect diversity in all possible ways. As a data annotator, you offer tremendous value to social media platforms who want to personalize their user experience. As technology companies expand their user bases across the globe, your work will continue to be a crucial piece of AI efforts.

If you’re interested in working on a social media project, apply for our Social Media Evaluator position or any one of our flexible, remote jobs.

Confidence to Deploy AI with World-Class Training Data
Website for deploying AI with world class training data