Pinterest Shares Insights Into Optimizing Social Platform Algorithms for Positive Engagement

Pinterest has shared a new overview of the “non-engagement signals” that it uses to refine its algorithms, and improve user experiences, in order to avoid going on direct engagement indicators alone.

Because by focusing on explicit engagement signals, like Likes, comments, etc., that can lead to negative outcomes.

As explained by Pinterest:

“User engagement is a critical signal used by Pinterest and other online platforms to determine which content to show users. However, it is widely known that optimizing purely for user engagement can surface content that is low-quality (e.g., “clickbait”), or even harmful. Our CEO, Bill Ready, explains that if we’re not careful, content ranking can surface the “car crash we can’t look away from”. On the other hand, “if you ask somebody after they saw the crash, ‘you want to see another one?’, the vast majority of people will say Goodness no,””

This is a key challenge for social platforms, as the incentives of straight engagement, which will increase usage, can often lead to platforms prioritizing the wrong content, or inadvertently driving more people to post borderline rule-breaking material, designed to lure clicks.

It’s actually something that can’t be solved fully, though every platform is trying.

Meta CEO Mark Zuckerberg wrote about this challenge back in 2018:

“One of the biggest issues social networks face is that, when left unchecked, people will engage disproportionately with more sensationalist and provocative content. This is not a new phenomenon. It is widespread on cable news today and has been a staple of tabloids for more than a century. At scale it can undermine the quality of public discourse and lead to polarization. In our case, it can also degrade the quality of our services.”

To help address this, Pinterest has partnered with UC Berkeley and the Integrity Institute to create a new “Field Guide to Non-Engagement Signals”, which provides an overview of how to build algorithmic approaches that are based on more beneficial signals than direct interaction alone.

And that, over time, will lead to more beneficial user experiences.

As per the guide:

“There is strong evidence that ranking by predicted engagement is effective in increasing user retention. However retention can be further increased by incorporating other signals, including item “quality” proxies and asking users what they want to see with “item-level” surveys. There is also evidence that “diverse engagement” is an effective quality signal.”

In order to address the first element, Pinterest runs user surveys to glean feedback into the user experience.

Most social apps run variations of the same, gathering expanded feedback, beyond straight engagement alone. These measures, while seemingly innocuous, can have a big impact on how each platform looks to rank content, because they provide additional indicators of what users want to see.

Using Meta as an example, back in 2021, Meta noted that:

“One of the top pieces of feedback we’re hearing from our community right now is that people don’t want politics and fighting to take over their experience on our services.”

That wouldn’t necessarily be reflected in straight engagement reports, which would have shown more interaction with political posts. But it’s direct feedback like this that can change the entire approach of social apps, by providing more color around what people like, and don’t like, beyond what the raw numbers suggest.

The guide also notes that manual labeling of content is another means of qualifying interests, though this takes significant labor to enact.

In alignment with this, Pinterest says that it puts a big focus on non-engagement signals to improve the user experience, while also using various measures to customize and refine each person’s feed.

“For example, our industry-leading inclusive product work has relied heavily on Non-Engagement Signals. When a user tells us the body type, hair pattern, or skin tone they want to prioritize in their feed, Pinterest can adjust what they see first.”

Though focusing on these elements does come with challenges.

For one, making non-engagement signals the priority will likely impact short-term retention and metrics.

Removing clickbait, for example, is a positive move, but less clickbait means, of course, fewer clicks, so platforms need to be able to weather changes in their data that may look negative in the initial phases.

Which many publicly listed companies, in particular, may struggle to do.

Pinterest also notes that tuning your algorithms “for emotional well-being”, which is the broader aim of this approach, takes time, and trial and error in many cases, which is much harder to stomach. But over time, Pinterest says, focusing on these elements will increase longer term retention.

I guess, some of this will also vary by platform, and Pinterest, which is focused on shopping, probably has an advantage in this respect over, say, X (formerly Twitter), which is driven by real-time news engagement.

But the impetus here is that social platforms can drive more beneficial user experiences by focusing on short and long-term goals, as opposed to driving the most possible interactions.

The guide also notes that generative AI “could be used to create better quality signals and enable new kinds of user controls” in future, refining these elements.

It’s an interesting overview, with a range of considerations for how we can improve social media interaction, and the elements of incentive that platforms use to demonstrate their utility.

Really, the bottom line is that focusing on engagement alone will drive more clicks, but the incentives that this drives will also, eventually, degrade a platform significantly.

Social platforms have a significant impact on how we engage and interact, and as such, there’s some responsibility there to focus on more positive, beneficial engagement signals.

You can read the “Field Guide to Non-Engagement Signals” here.