| > The only reason is because I'm the only developer and simply haven't had time to build it yet! I see you're using React on the frontend. And, you only have an iOS app but not (yet) an Android app. Are you not using React Native? In that case it's fairly easy to target both platforms. I know RN is a bit of a shitshow but not supporting the biggest mobile platform in the world is arguably worse > We gather "tag" and "description" metadata for articles even though it isn't currently displayed in the UI. If you're taking this metadata directly from the articles I'm skeptical about the accuracy/completeness. Anyway, it's better than nothing I guess > It's not a perfect filter by any means, but keep in mind that even if an article title looks like clickbait it will only rank highly if users are finishing it This is something, but still high-quality articles with clickbait titles will outcompete high-quality articles without clickbait titles. Incentive still being: write clickbait titles. Anyway, if someone is 0.25x as likely to finish a bad article having clicked it, but 10x more likely to click a clickbait article, you still have a major problem beyond "not perfect", imo I know it's a big ask for you to solve every single problem with online reading/discussion but this one is so tangled with the rest it's kinda hard to ignore. Today's AOTD: "If Everyone Else is Such an Idiot, How Come You're Not Rich?" From the past few weeks, some selections: - "Meet the social media echo chamber that is radicalizing you & your friends. - Alexa Rohn"
- "Racism Is Terrible. Blackness Is Not."
- "A White Woman, Racism and a Poodle"
- "The American Press Is Destroying Itself"
- "Dear Fuck Up: How Do I Figure Out What I Want in Life When Every Day Feels the Same?"
- "You Should Be Feeling Miserable"
- "Tom Cotton: Send In the Military"
- "The Sickness in Our Food Supply" I'm sure some of these are great, but be honest: did the title have anything to do with people clicking through? |
> Are you not using React Native?
Not even! Our iOS app just uses WKWebViews for the main UI so yes it would be pretty trivial to do the same thing on Android. Only excuse is that it's in our backlog with 100 other things that we also really want to build.
> If you're taking this metadata directly from the articles I'm skeptical about the accuracy/completeness.
Haha yes, it's an absolute clusterfuck that I'm currently trying to clean up enough to make useful. The descriptions are usually actually pretty solid, but there is so much noise in the tags/topics.
> This is something, but still high-quality articles with clickbait titles will outcompete high-quality articles without clickbait titles.
Yes, yes, yes! I think providing more context in the way of the description could help to cut down on this, but you're very right that there is no easy or complete fix (at least not that I can imagine!). Something else to think about would be looking at the ratio of clicks to completions instead of just the sum of completions alone. That way in your example the 10x likelihood of a click could be cancelled out by the low 0.25x completion rate.