Classifying a topic as/as-not viral dreck is, I posit, quite hard. Indulge me a bit here - I primarily come to HN to read about ML/AI/DNN. I mark the ones I like with the favorites button. During my work downtime, I go back to my favorites folder and re-read the topic. For some of the topics such as Word Embeddings, I must have re-read that paper a dozen times. I have actually used the material I learnt from my HN favorites list at my workplace, built it into products, gotten $$ off of it. So stuff like JIT compiler, LibHTTP, Bitcoin Bubble, Nintendo (all the topics trending right now) are viral dreck to me. I don't want this viral dreck. Get it off of HN!
You can see how quickly this sort of reasoning becomes vacuous.
I have an FB account mostly to interact with my extended family. Most of them are in their 70s and 80s. Every single topic they interact with on FB qualifies in my mind as viral dreck. However, by the nature of their comments, they seem to derive much pleasure in engaging with it. Given their age, who am I to deny them their cup of poison. Its not like they are going to get wise on word embeddings at the age of 75. Let them have their viral dreck.
..Indulge me a bit here - oh don’t encourage me. :)
On the “this is hard” point. Yes it is. I’m asserting that every business have a handful of core, hard problems. One of journalism’s is maintaining good quality standards, traditionally using subjective yet fairly well established definitions and methods, like an editor with good taste, formal fact checking, etc.. One of Apple’s hard problems is making consumer tech elegantly simple. The rhetorical click-wheel iPod. Again, there’s a lot of subjective elements that go into this.
How does a FB tackle this? Judging by your favourite HN article, I will forward a guess that you immediately jump to an ML or an algorithmic way of classifying things. On that front, I think FB could get some wins. Is dreck classification all that different from spam classification? Does PG’s “plan for spam” apply here? I would guess that it does for some dreck types. I actually think FB is currently deploying some spam-filter like system to deal with the most onerous “fake news.” Google’s starting point was page rank, a relevance ranking system. Fairly quickly, they added in “quality” ranking to supplement that. This is probably the least monolithic of their core systems. At the same time, SEO came into being and anti-spam became the third plank to google search. Taken together, this is quality classification. The inverse of dreck classification.
I don’t want to overemphasise these things though. FB is one of the best positioned to use “computer science-ish” solutions. But, I think the bigger choices are not at that level. They’re not a choice of technological approaches. It’s more mundane, values, priorities, product culture. Defining what they are and making good content a part of that definition.
HN already has quality built into it. When a feature is released (rarely, after all HN is not even a commercial product) someone makes a decision. Does nesting comments like this encourage dreck, in HN’s case flame wars and quips? Does changing something about moderation improve the quality of discussion? Are the really good, but slower, harder and more technical topics being buried? How do we unbury them? This speaks to your ideal HN.
No one has ever asked and answered such a question about youtube comments, we see the results. I don’t think FB has done enough of this, especially as relates to the high exposure. “viral” content.
The best articulation of the approach (IMO) is Spolsky’s “social UI^” and the early podcasts with him and Atwood. When they were making Stackoverflow, a lot of attention was paid to the type of content/behaviour encouraged by certain product/UI choices. They had an idea of what a good comment, question or answer looks like. They built the product to produce these. It wasn’t just about “more stuff.” They had an opinion about what stuff is valuable.
FB will need to be more programmable, but that’s easier if you start from the Spolsky mentality. Instead of “dreck classifier,” you just need to recognise flame wars, thought bubbles, divisive headlines. You need to decide that a comment by someone who has read an article is more valuable than a comment by someone who has not. The components become more achievable sounding, IMO.
Classifying a topic as/as-not viral dreck is, I posit, quite hard. Indulge me a bit here - I primarily come to HN to read about ML/AI/DNN. I mark the ones I like with the favorites button. During my work downtime, I go back to my favorites folder and re-read the topic. For some of the topics such as Word Embeddings, I must have re-read that paper a dozen times. I have actually used the material I learnt from my HN favorites list at my workplace, built it into products, gotten $$ off of it. So stuff like JIT compiler, LibHTTP, Bitcoin Bubble, Nintendo (all the topics trending right now) are viral dreck to me. I don't want this viral dreck. Get it off of HN!
You can see how quickly this sort of reasoning becomes vacuous.
I have an FB account mostly to interact with my extended family. Most of them are in their 70s and 80s. Every single topic they interact with on FB qualifies in my mind as viral dreck. However, by the nature of their comments, they seem to derive much pleasure in engaging with it. Given their age, who am I to deny them their cup of poison. Its not like they are going to get wise on word embeddings at the age of 75. Let them have their viral dreck.