3 ms·

I've thought a bit about this recently, since HN and Reddit in particular are prone to have the posts sorted by how much it appeals to the average user. Other

I've thought a bit about this recently, since HN and Reddit in particular are prone to have the posts sorted by how much it appeals to the average user. Other systems such as Twitter and Facebook limit you extremely in the fact that you only converse with likeminded people, and when you don't, you'll most likely never end up in a constructive discussion, just in a flamewar since everyone is dead set on their opinion already, and because of the limitations in the commentning system discourages it. What about a system based on NLP / ML where users can flag the truthfulness or speculation? (For the UI part, I'm thinking something similar to Medium, where you can select a favorite passage from a blog-post) If you rank posts by how objectively true they are, and the amount of truth/false-votes they get, you should be able to bootstrap posts in, while also dynamically adjust to the popularity of it (number of votes received). I know this is still a stretch with current machine learning technology, but I think we (the internet) really need a place where we can more easily filter out the bullshit, and discuss the things that really matter with people who don't already agree.

egjerlow 8y ago
I think we would need more than the 'truth' dimension to make this work. Writings can have great value even if what they say is not directly 'true' - think poetry, motivational writings etc.
- burgerdev 8y ago
  ... and the first 10k digits of pi might be 'true', but are neither educational nor entertaining.
pjc50 8y ago
This is an absolutely terrible idea, because truth is not something that can be inferred from the text alone; epistemology is hard enough for humans, let alone for machines. The approach of decoding human text into statements in propositional logic has been tried, it was the first 70s "AI wave", but ultimately it broke down on the complexity of the problem. Modern "ML" can only give you a stochastic truthiness. The only people who haven't given up on the approach of a massive symbolic reasoning database to work out what's true and false are Cyc. https://en.wikipedia.org/wiki/Cyc https://en.wikipedia.org/wiki/Cyc You might be able to have a system which auto-flags known bullshit statements, but people will get outraged by that.
- digi_owl 8y ago
  Yeah any kind of community "tagging" system is fertile ground for trolls and whatsnot. Just check out the mess that is steam tags, where soft-core visual novels gets mixed in with just about anything thanks to troll applied tags. We are simply asking for another racist AI or gorilla moment.
- NegatioN 8y ago
  Hi pjc, you are probably correct! :) However, I think although this system might not be able to infer truth, it would at least be possible to infer "effort put into the post" or "well structured post" vs the obvious low effort things we see every day. Which might be enough for a better system. Any suggestions for what you would train on to get a model which could rank more objectively than an upvote/downvote system can? I think we should be vary of being bound by everything that happened in the 70s due to the differences in resources for solving the problem, although I'm not in any way saying this is easily solvable or even solvable in our time. I do think it's not provable that we can't achieve this at some point.
  - pjc50 8y ago
    Some sort of minimum-effort or relevance filtering may be possible; after all, that's what spambayes achieves. You could probably feed it a human-approved list of "good" and "bad" posts and get a first cut at removing simple spam. But really you have to define "good" "objective" and "structured" first. Or even "effort". Is the system good if it promotes high-effort well-structured calls for genocide?

egjerlow 8y ago

I think we would need more than the 'truth' dimension to make this work. Writings can have great value even if what they say is not directly 'true' - think poetry, motivational writings etc.

burgerdev 8y ago
... and the first 10k digits of pi might be 'true', but are neither educational nor entertaining.

pjc50 8y ago

This is an absolutely terrible idea, because truth is not something that can be inferred from the text alone; epistemology is hard enough for humans, let alone for machines. The approach of decoding human text into statements in propositional logic has been tried, it was the first 70s "AI wave", but ultimately it broke down on the complexity of the problem. Modern "ML" can only give you a stochastic truthiness. The only people who haven't given up on the approach of a massive symbolic reasoning database to work out what's true and false are Cyc. https://en.wikipedia.org/wiki/Cyc https://en.wikipedia.org/wiki/Cyc You might be able to have a system which auto-flags known bullshit statements, but people will get outraged by that.

digi_owl 8y ago
Yeah any kind of community "tagging" system is fertile ground for trolls and whatsnot. Just check out the mess that is steam tags, where soft-core visual novels gets mixed in with just about anything thanks to troll applied tags. We are simply asking for another racist AI or gorilla moment.
NegatioN 8y ago
Hi pjc, you are probably correct! :) However, I think although this system might not be able to infer truth, it would at least be possible to infer "effort put into the post" or "well structured post" vs the obvious low effort things we see every day. Which might be enough for a better system. Any suggestions for what you would train on to get a model which could rank more objectively than an upvote/downvote system can? I think we should be vary of being bound by everything that happened in the 70s due to the differences in resources for solving the problem, although I'm not in any way saying this is easily solvable or even solvable in our time. I do think it's not provable that we can't achieve this at some point.
- pjc50 8y ago
  Some sort of minimum-effort or relevance filtering may be possible; after all, that's what spambayes achieves. You could probably feed it a human-approved list of "good" and "bad" posts and get a first cut at removing simple spam. But really you have to define "good" "objective" and "structured" first. Or even "effort". Is the system good if it promotes high-effort well-structured calls for genocide?