Research in Public #10: What makes a good post? +Long form content is increasing on SN1
Just a quick update on some interesting results. I earlier posted about the determinants of post quality, but the model was pretty opaque. Here is a table distilling the essence of that in an easier to understand form:
\bar{z} is the average zaps received in the first 48 hours of posting; \bar{c} is the average number of comments received in the first 48 hours; and N is the number of such posts.The table shows that:
- Link posts with no text are the most common type of post, but the least rewarded.
- Long-form discussion posts with images or links in the post body are the most rewarded.
- Link posts are less zapped and less commented on in general compared to discussion posts.
- Images or links in the post body seem to enhance longer posts, but do not necessarily enhance shorter posts.
One of the main questions of the research project is whether SN's incentives lead to higher quality content. Let's define a "high quality" post as a post that is either:
>50words discussion post>50words link post with images or links in the post body. (Images or links in the post body of a link post suggest that the post content isn't just a copy/paste of the link itself. Also less likely to be AI generated.)
The graph below shows that the proportion of high quality posts is growing quite linearly over time on SN.
So, the dynamics on Stacker News indeed seem to be incentivizing the types of posts that attract more zaps and engagement. That seems like a good thing! Some people worry about the growth of users on SN, which has indeed been slow, but the quality of the discussion and engagement seems to have been going up over time.