321–330 of 408 posts
Earlier quoted context omitted.
Do you have watch history off? I have never experienced anything like this, but I pay for Premium.
Yes, I have watch history turned off. I suspect that gets me a different selection of ads.
Earlier quoted context omitted.
> 1 or 2 upvotes makes me feel like maybe it doesn't belong. This is more or less artifact of HN algorithm, it's common to get single digit votes for majority of your posts. Whether something blows up feel almost random, you have to get pretty lucky to hit a time window when there's not that many posts or a lot of people look at new page and upvote the post at the same time to make it snowball. Many links are posted…
s/algorithm/dynamic/ HN doesn't have an algorithm, per se. There are voting mechanics, and some sites gain or lose a penalty based on content or type (most generic news sites, for example, are slightly penalised). There are keyword / topic penalties too for issues that are dominating the hivemind for a period. But mostly what you're seeing is simple mass-media power-law effects, along with early-action advantage: - V…
> A small amount of early activity (upvotes, flags, comments) tends to have an outsized effect
This is exactly the problem.
> what you're seeing is simple mass-media power-law effects
I would challenge that point. Power law comes from some feedback loop, which is partially from network effects but it can be massively amplified by the system, which is exactly what HN does. Not only it bakes the power law directly into the score eqaution, but it also shows the list sorted by score by default, which creates a positive feedback loop on votes.
Actually I'm a bit perplexed that it works this well, HN algorithm was one of the first that I implemented on our site and it was quite terrible even after a lot of tuning. I feel like it must be tuned for some volume of posts and people, otherwise it doesn't make much sense to me.
Earlier quoted context omitted.
Genuinely curious: why do I always get scammy ads in iOS apps and Google products but I don’t see anywhere near the same amount on Instagram or TikTok? Are Meta and TikTok better at filtering scammy ads out? Maybe their ML recommendation systems realize I’d never click on such an ad and the other platforms can’t figure that out.
IG, TikTok, Meta have a lot of data they can use to do good ad targeting that is harder to apply to Google/iOS apps. Also has to do with the ad format, placement, etc.
I encountered a similar problem with blog subscription services: I was using "follow.it", which would handle subscriptions and send out an email when I have a new article. It worked fine until follow.it went spammy a few months ago, switching to emails that were mostly objectionable junk ads. Does anyone have advice for handling blog subscriptions? I'm thinking of switching to Kit, which has a free tier that seems re…
RSS
Earlier quoted context omitted.
Yes, I have watch history turned off. I suspect that gets me a different selection of ads.
Maybe it's a dark pattern to punish you for turning off personalization.
Earlier quoted context omitted.
The ad industry likes to say that their industry is clean and the people who buy ads for scams are the problem, but the truth is the entire industry is complicit with the scamming, and stuff like this shows it. If the ad industry were merely hapless victims of the scammers, rather than willful participants in the scamming, they'd be eager to receive reports of scams.
I feel like there are almost two completely incompatible stories being sold by ad tech people. One is that "you shouldn't use an ad blocker, reputable websites have ads from Google and other reputable companies and they wouldn't be scams or malware". The other is that any time there's a story about malware and scams on reputable websites, they say: "there are so many ads being submitted, Google is doing their best bu…
For the bulk of legit advertisers, this won't affect the bottom line of a given campaign and will keep out the scammers, etc. Leading to a much higher quality ecosystem. It would also give Google the potential for a higher percentage share of profits in the system. It would also likely reduce the CPM pricing though, possibly to a larger extent than profits from the validation system itself.
But it would be a much healthier overall system.
Earlier quoted context omitted.
s/algorithm/dynamic/ HN doesn't have an algorithm, per se. There are voting mechanics, and some sites gain or lose a penalty based on content or type (most generic news sites, for example, are slightly penalised). There are keyword / topic penalties too for issues that are dominating the hivemind for a period. But mostly what you're seeing is simple mass-media power-law effects, along with early-action advantage: - V…
I don't understand the point of that first nitpick, this is an algorithm, at least in the normie sense of the word as a ranking system for list of posts. > A small amount of early activity (upvotes, flags, comments) tends to have an outsized effect This is exactly the problem. > what you're seeing is simple mass-media power-law effects I would challenge that point. Power law comes from some feedback loop, which is pa…
https://knightcolumbia.org/content/the-myth-of-the-algorithm...>
To the extent that HN does utilise specific procedural mechanisms to adjust the priority of content, it's virtually always away from the typical patterns of algorithmic amplification: less emotion, less outrage, fewer hot takes, less nationalism and relgious flamewars, and specifically toward "intellectual curiosity and thoughtful conversation": https://news.ycombinator.com/item?id=13108404>.
It would be possible, yes, though incredibly disingenuous, to argue that what HN is doing is itself amplification. Yes, any curation is an amplification of some content over other, but in a world where "algorithmic content" means clickbait, brain-crack, and stickyness, HN is quite clearly aiming for something else.
Another facile objection is that HN fails to achieve its stated goals. Well, yes, it does, and the mods freely admit this (see, e.g.: https://news.ycombinator.com/item?id=20188101>). Why does HN fall short? Because the problem is hard (see, e.g., https://news.ycombinator.com/item?id=16163743>).
If power-law dynamics were purely the result of manipulative algorithmic amplification, we'd see them only in online media subject to such amplification. And that's simply not the case. Power laws are fundamental to not only all of human communications and interactions (word and letter frequencies, for example, neither of which suggest a strong influence by algorithmic amplification), but to all manner of natural phenomena, including those entirely outside the realm of biological activity (e.g., frequency/magnitude plots of earthquakes, volcanoes, asteroid impacts, and stellar novae).
And in the realm of interpersonal online communications, HN's goals and interventions (mods, voting, and some programmed mechanisms) are desperately trying to swim upstream. As someone whose online tenure pre-dates the Web and extends to pre-Eternal September Usenet, HN has done remarkably well, and outlived many of its antecedents' and competitors' useful or entire lives (Usenet, Slashdot, Digg, Reddit, Google+, et cetera). Trust me, I'd love to see it do better (a view often voiced by mods as well). But in an ordinal ranking with what actually exists it's an exemplar.
This isn't a nitpick, it's a core and central point with (literally) universal applicability.