Creator feedback

Creator feedback addresses a gap in how feeds, which choose which posts to show each reader, have been built: LinkedIn Engineering said it historically considered the ranking problem primarily from the viewer's perspective1. Its 2018 post says the model did not consider how much the creator may appreciate receiving feedback from the viewer, and calls this a blind spot1. The value of a response differs by creator: for the top 1% of creators, one more like or comment from an unknown follower may go unnoticed, but for the average creator one more piece of feedback from a close colleague can be very meaningful and could lead to a conversation and maybe a career opportunity1. A LinkedIn research paper from June 2021 makes the same point, saying that feeds have mostly modeled the preferences of the people reading and that it is equally critical to prioritize creators' interests, because quality content brings in new readers and keeps existing ones2. The paper proposes predicting how feedback from readers encourages creators to post, then using that prediction to reshape how likes and comments are spread across creators2.

LinkedIn's feed combines the value of a response to a creator with viewer-focused predictions: LinkedIn Engineering says its second ranking step merges predictions from different models into one combined score that balances three aspects, namely viewer-side value (passive actions such as clicks and dwell time, and active ones such as comments and reshares), the impact of a viewer's actions on their network, and the upstream value to the creator of different viewer actions3. The 2018 post lists some of the inputs behind this: whether the viewer is likely to click, like, comment or reshare, how that feedback may spread through the viewer's network, and how much the creator would value feedback from that particular viewer, such as a close colleague1. The 2024 post says the models for the creator side have architectures similar to the viewer-side model, and that the weights combining the aspects are finely tuned, though it gives no values3. A Microsoft patent from September 2020 describes estimating how much more a creator would gain from one more response, by asking its model how likely a member is to post at several different amounts of feedback already received and assuming the gain falls as feedback piles up4. A Microsoft patent from December 2022, about suggesting people to follow, describes a score where a creator who gets little feedback may post more after a little extra, a creator who gets a lot may not, and the sensitivity may differ between two creators with the same amount of feedback5. Our read these patents probably show how the creator-side value could be estimated, because the 2024 post names the same upstream value to the creator and says its models are built like the viewer-side ones, but LinkedIn has not said the patents describe the live feed345.

Creator feedback matters most for creators who are not big: LinkedIn Engineering said the feed is designed to show their quality posts to people close to them, such as colleagues and connections, because the feed looks at how closely the viewer is connected to the creator1. Post quality is also an input, so low-quality or spammy posts do not get a boost just because a creator needs feedback1. LinkedIn has not said how many responses count as the first few or how large the effect on reach is13.