SecurityBrief New Zealand - Technology news for CISOs & cybersecurity decision-makers
Story image
TikTok's unveils plans to safeguard and diversify recommendations
Mon, 27th Dec 2021
FYI, this story is more than a year old

TikTok has unveiled new details about the platform's work to safeguard and diversify its recommendation system.

To keep people's For You feeds fresh – and interrupt repetitive patterns – TikTok's recommendation system works to intersperse diverse types of content along with those people already know they love.

"Our community tells us they love finding creators they wouldn't have known to follow, or learning about a new interest because it was recommended to them, along with enjoying content that already matches their taste," the platform says.

However, certain kinds of videos can sometimes inadvertently reinforce a negative personal experience for some viewers, like if someone who's recently ended a relationship comes across a breakup video.

"We want to share more about some of the work underway to address this and improve the experience for viewers on TikTok."

Diversifying recommendations

"At TikTok, we recognise that too much of anything, whether it's animals, fitness tips, or personal wellbeing journeys, doesn't fit with the diverse discovery experience we aim to create. That's why our recommendation system works to intersperse recommendations that might fall outside people's expressed preferences, offering an opportunity to discover new categories of content," the company says.

"For example, our systems won't recommend two videos in a row made by the same creator or with the same sound. Doing so enriches the viewing experience and can help promote exposure to a range of ideas and perspectives on our platform."

As the company continues to develop new strategies to interrupt repetitive patterns, it is looking at how its system can better vary the kinds of content that may be recommended in a sequence.

"That's why we're testing ways to avoid recommending a series of similar content – such as around extreme dieting or fitness, sadness, or breakups – to protect against viewing too much of a content category that may be fine as a single video but problematic if viewed in clusters," it says.

"We're also working to recognise if our system may inadvertently be recommending only very limited types of content that, though not violative of our policies, could have a negative effect if that's the majority of what someone watches, such as content about loneliness or weight loss. Our goal is for each person's For You feed to feature a breadth of content, creators, and topics."

This work is being informed by ongoing conversations with experts across medicine, clinical psychology, and AI ethics, members of our Content Advisory Council, and the TikTok community.

A new way to shape what you see

"As we build safeguards into TikTok by design, we also want to empower people with more choices to customise their experience to their own preferences and comfort," TikTok says.

For example, it is working on a feature that would let people choose words or hashtags associated with content they don't want to see in their For You feed.

TikTok already enables people to tap any video and select "Not interested". This new tool will offer another way to help people customise their feed – whether for a vegetarian who wants to see fewer meat recipes, or someone working on self-esteem who would rather see fewer beauty tutorials.

Building safety into recommendations

TikTok's For You feed is designed to help people discover original and entertaining content, and the company has a number of safeguards in place to support this aim.

"In addition to removing content that violates our Community Guidelines, we try not to recommend certain categories of content that may not be appropriate for a general audience," TikTok says.

"Our safety team takes additional precautions to review videos as they rise in popularity to reduce the likelihood of content that may not be appropriate for a general audience entering our recommended system."