Hacker News Reader

Comment by dheera | Hacker News Reader

It would be awesome if Google had up/downvote buttons like Reddit and use them to eliminate all the fluff in search results, even if on a targeted community basis. For example if a recipe has no pictures, too many ads, not clear instructions, or no equivalent volume measurements (e.g. "4 grams sugar (about 1 tsp)") I'm likely to downvote it, and those properties can be learned from multiple downvotes.

View on HN

I would love for that to work but I'm skeptical that there would be a lot of signal in that sea of noise. We're talking about well over a billion monthly active users; for every HNizen annoyed with the state of the modern web, there are thousands of ordinary people that probably don't mind J. Peterman-esque recipes laden with product placement and somehow incorporating a listicle - as long as they get their recipe.

And once you open the floodgates of allowing user feedback to influence search results, the arms race begins, with every major news site checking the referrer for Google and begging you via popover to upvote their search result, and armadas of malware-controlled botnets being launched to generate authentic-looking feedback, and so on.

You could do a couple of things:

* Make your upvotes only influence your own search results, and other users like you (similar search history, similar geography, similar past upvote pattern). The fraudster upvoters would just improve SEO for their own fraudster circle. They would see their fraud-upvoted site as #1 within their own circle but it wouldn't change the ranking for me. The programmer circle gets programmer-vetted results on technical help. You know they are a programmer from their search history.

* Sites begging to upvote get a massive SEO hit of -10000 downvotes. Google's NLP should be good enough for this already.

What you are describing is a literal filter bubble, where very soon your only exposure to content will be seeing content you are already liking, ideas you are already adhering to, people you are already looking to. A bubble where ideas and people that you and people like yourself don't like automagically get hidden from you so that you never get exposed to them. A nightmare.

I think this can be done a little more intelligently. Up/downvotes should be based on the structure of the site and more so than the meaning of the content.

An downvote should not mean "I disagree with this author" but rather "this site was useless spam" or "I derived no information from this site because it threw 5 popups in my face". Or if I downvote a bunch of paywall sites it should stop showing me paywall sites.

We're looking for a spam filter, basically.

> An downvote should not mean "I disagree with this author" but rather "this site was useless spam"

How would you tell them apart? If you could identify abusive site formats mechanically anyway why even solicit votes?

It's still relative. I hate paywalls, some people may prefer to just pay. I hate popups with a passion, so much that I will leave the site, but Average Joe probably just clicks them away.

What you're proposing in this comment doesn't even work on Reddit/HN. People don't even agree what each star represents in 5-star rating systems.

That's because those rating systems naively apply the ratings across the entire population. The average of the star ratings should be weighted by the normalized dot product of each rater's style and your rating style, or the similarity between each rater and you. A 0-star rating from someone whose preferences are orthogonal to you should not influence the average rating you see.

Its a really nice idea, but I wonder if it would work in practice:

1. Would a significant amount of real humans who actually cared about the content cast a vote (up or down)?

2. Would it be even remotely possible to protect against fraudulent voting at Google scale? (remember they can't even prevent fake business listings on Google Maps, which is at least theoretically closer to verifiable from, for example, business records)

I don't think it will work in practice. Reddit and other voting platforms require a login to prevent spam voting and do people want Google to know which articles you like and don't? Reddit works because anonymity which Google will not provide.

But Reddit, along with most major sites now, is also rife with the same marketing manipulators and "influencers" the article is commenting on. Digg tried something similar before Reddit and was pretty much ruined by biased organized vote manipulation (brigading). Not that I agree with Google's "trust us plebes we know better than you" model of secrecy either. The SEO problem is the prime example of why this doesn't work also. Search engines have become a vital resource for the modern internet. We need an open, publicly funded non-profit search engine.

1. If the upvote button somehow stuck on later after navigating to the site, I probably would. I realize this would probably make Chrome even more tied to Google in a potentially slightly evil way though, but it might improve search quality drastically

2. I think Google could do it. They're pretty damn good at detecting bots on Google.com just by analyzing whether your behavior looks human or not, and they also have logic in place to detect fraudulent ad clicks. I think they could do a much better job with fake business listing detection but it's probably just not their focus. Search, on the other hand, is their lifeblood.

They A/B tested this as an experiment at least once. I recall there was another try too, but couldn't find info about it.

http://justinhileman.info/article/googles-edit-search-result...

I fear that would be worse as it would just get super-SEOd with paid upvotes.

This would be good if you could filter using only votes from people you know. Then you could mix in maintained lists like “all active HN users” from various communities.

But then a lucrative business of selling upvotes would be created for people willing to pay (usually businesses) to promote their content. And that's SEO all over again.

I initially believed the "O" in SEO was misleading, as this kind of promotion is not optimisation (at least not for everybody), It turns out it's another letter that's misleading: "S" should be for Sales, not Search.

They essentially do.

Bounce rate is a key ranking factor. Going back to the search results too quickly is a negative signal.

This already exists. If you navigate back to the search results then that counts as a downvote. If you don't navigate back, that's an upvote.

I know this is true, but my use of tabs means I've opted out of the system.

I open tabs for the top results and just close them when done. Often, I have to re-search with added "verbatim" filter or quotation marks because Google seems to ignore my most important keyword 90% of the time.

Hmm, I open result pages in a new tab. I don't "back", just close tabs

then SEO would turn towards having an army of bots upvoting your product and downvoting the competitor. And if you have something in between to detect bots, users wouldn't want to use it (who would solve a captcha to help optimize search results?) and SEO would find ways around it

Upvoted.

Reply on news.ycombinator.com