I actually had an offer from my advisor to study for a PhD under him by basically taking on a major coding project to develop a non toxic format of social media, or at least to get as close to that as possible. I also remember it involving some non crypto uses of blockchain software potentially but my prof was more interested in the complex web development aspects of it and actually got my web design professor in on pitching the potential for it to me.
We had me enroll in a master's level web design class to trial run it, and it turned out I wasn't the webcode wunderkind they thought I might be given my good web design work for my courses so far. I'm creative but not a savant basically. We all agreed that my PhD potential was best sought after more design experience got under my belt.
I thought the whole deal with PhDs was that they were supposed to be new research that pushes the field forward. What field were you studying? What was the academic contribution the prof was expecting?
Software Engineering, and it was more intended as a build project than as an academic venture. We would basically have been setting out to strip out all the addiction features Social Media is built with and see what can be built with what's left and what new ideas can make it a good experience for the user without queuing them for ranked rage fest matches against whoever else is in the comment section they clicked on in the rage bait article.
Idea: enable downvotes, but they don't just lower the ranking of the comment you downvote, they lower the ranking of the top level post. And they have a much larger effect than upvotes on ranking.
Now if you get into a fight with someone in the comment section, nobody will see it because your downvotes on each other will push the whole post into obscurity. Only uncontroversial posts where everyone agrees succeed. Thus, people are incentivised to make posts that everyone can enjoy. If you post something racist and start a fight in the comments, your post gets no upvotes, and the fight is contained from spreading further.
Downside: this system doesn't discriminate between content thats controversial because it sucks and content that's controversial but good. For example, any kind of vegan content is going to be buried. You'll get a much stronger hivemind effect on this social media.
That's the same as creating a botnet to downvote a post directly, except now you need to plug it into an LLM instead of just running a script. It's way more work.