Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ask HN: What if AI was used to verify users reading articles before discussion?

news.ycombinator.com

Post reply on HN
Print view View original

31–38 of 38 posts

Re: Ask HN: What if AI was used to verify users reading articles before discussion?

#31

Post by l0b0 »

(Go ahead and downvote, but I think the tone in this answer is appropriate.)

Yes, let's sprinkle magic fairy dust on threads to fix human foibles. Let's make sure there are no more meta-discussions, like people not being able to read the article for any reason. Let's also avoid any posts like "Hey, if you like the author you should check out X". Or, conversely, "Hey, this person is a known crank[1][2][3], don't bother reading." Or anything where a one-paragraph tl;dr could've easily replaced the entire article, and answering based on the tl;dr is completely reasonable. Or any of a host of other types of comments which I couldn't think of off the top of my head.

Let's instead make sure everyone who wants to post needs to remember things from the article which are completely irrelevant to the actual content, and that they spend a few minutes trying to guess what the question means and which exact characters will appease the algorithm.

Let's not.

Re: Ask HN: What if AI was used to verify users reading articles before discussion?

#32

Post by version_five »

I find HN has a better SNR than any other user discussion I've read on the internet. Arguing against points addressed in the article seems like a good thing to me (maybe I misunderstand though and you mean people are missing objective facts). My biggest criticism of the discussions is that there usually isnt much diversity of opinion for controversial topics and most outside a narrow range get downvoted. I think diversity would get worse if there were more barriers to commenting.

Re: Ask HN: What if AI was used to verify users reading articles before discussion?

#33

Post by saternius »

You can check out our best attempt at a question generator at http://quickquiz.me, to get an idea of what type of questions a large language model (T5) can produce. Its the best model my team and I witnessed. We are still collecting data to generate more salient, and well structured questions. Its a work in progress.

Re: Ask HN: What if AI was used to verify users reading articles before discussion?

#34

Post by GOATS- »

> I then came upon the idea that perhaps something like GPT-3 could be used to have users answer 3 questions about the article's content before being able to comment.

The Norwegian equivalent of the BBC did this.

https://arstechnica.com/gaming/2018/03/how-a-norwegian-comme...

Re: Ask HN: What if AI was used to verify users reading articles before discussion?

#35

Post by antisthenes »

> I recently read a frustrating thread on HN where it was obvious several people were making arguments against points that were already addressed in the article.

Ummm, yes? Isn't that the point of a discussion thread? Just because the article makes some points doesn't exempt them from criticism.

Many articles posted on HN are outright garbage - e.g. very poorly constructed studies or some undergrad psych/nutrition juju that doesn't pass a basic critical thinking test, so it's quite rightful that people don't waste their time reading them (or at least reading them completely)

> The discussion in the thread was cluttered with people attacking straw men instead of having any fruitful discourse.

Typically, people attacking straw men will do so regardless of whether they read the article. The problem with those people is that they invent the straw man, attack it, win, and do all those things without ever caring if it relates to the topic at hand at all.

Re: Ask HN: What if AI was used to verify users reading articles before discussion?

#36

Post by brudgers »

morceauxdebois wrote: post #5
onion2k wrote: post #2

If I found my posts were blocked because some AI said I wasn't posting correctly I'd leave and never come back[1]. The chilling effect this would have on discussions here would have a much more detrimental impact than any benefit from preventing bad posts. [1] Some people might argue this is a benefit.

Your posts wouldnt be blocked, you would have to verify that you have read the article first before posting.

How do I prove I read a repository or a landing page?

Re: Ask HN: What if AI was used to verify users reading articles before discussion?

#37

Post by jgwil2 »

chroem- wrote: post #29
bjourne wrote: post #21

I have noticed that too and it is pretty fucking annoying (+). If the article is about US colleges then the first comment is always about over education and how not every job requires a college degree. If the article is about trains it is about how the US, unlike Europe, is too large for train networks. If it is about renewable energy, the first comment is about how the green movement is shooting itself in the foot b…

Many people would already consider HN to be a bay area-centric echo chamber. Based on the examples you listed, this seems to be an argument to explicitly enforce that echo chamber so that nonconforming posts are blocked outright. At that point, why even bother allowing users to discuss things at all, if posters are only allowed to agree with each other?

Nowhere does GP suggest anything of the sort. Only suggestion is to weigh upvotes less heavily relative to recency in determining comment order.

Re: Ask HN: What if AI was used to verify users reading articles before discussion?

#38

Post by jgwil2 »

version_five wrote: post #32

I find HN has a better SNR than any other user discussion I've read on the internet. Arguing against points addressed in the article seems like a good thing to me (maybe I misunderstand though and you mean people are missing objective facts). My biggest criticism of the discussions is that there usually isnt much diversity of opinion for controversial topics and most outside a narrow range get downvoted. I think dive…

I think the complaint is that people routinely rehash points that were already addressed in the article, because they didn't bother reading it in the first place.
Post reply on HN