I'd like there to be a "Report blogspam / SEO spam / plagiarism" button in DDG.
Step 1: Click the result...
Step 2: Ugh, swipe back
Step 3: Click "Report poor result" > "SEO spam"
Step 4: Click next relevant link
back
1 comments
How would this not be gamed by bots? I see how what you're suggesting would work if there were humans going through and moderating, but I can imagine they would be deluged with millions of reports. Doing it algorithmically is just giving the bad actors a mechanism to sink their competitor's results.
This is the thing people don't get about search. Not only is it incredibly difficult to accurately predict what content will give people exactly what they want, but there is an entire class of bad actors dedicated to making the results worse.
I think we need to stop using search so much. Curated indexes of information are often better for finding what you need. They organize bodies of knowledge that enable us to do more than just play detective and scavenger hunt.
Take a library for example. To find out how to repair a stuck brake caliper, you don't go to the library and tell the librarian "stuck caliper fix". You say "Can I have a book on auto repair, or on repairing car brakes?" You get the book, and it's an index of organized knowledge. In that book you'll learn how to fix the stuck caliper, but also see what causes stuck calipers and how to prevent them - because it's not blog spam designed to fix one error, it's organized knowledge.
Wikipedia is a great curated index. It is literally a giant list of links, with paragraphs summarizing what you can find in different pieces of the links. (it's also an entire semantic database, but ignore that) If you suggested starting Wikipedia to people today, they'd probably say it was crazy and nobody would use it because of how difficult it would be and that the results wouldn't be useful or accurate. And in spite of all that being true, people today probably can barely imagine life without Wikipedia.
> This is the thing people don't get about search. Not only is it incredibly difficult to accurately predict what content will give people exactly what they want, but there is an entire class of bad actors dedicated to making the results worse.
Most of my problems with Google isn't about spam sites.
Most of my problems with Google are because half the pages they show me doesn't include my search terms.
This isn't a hard problem and I'm tired of hearing people defend it.
Yes, years ago people would keyword stuff pages with white text on white background. I haven't seen this in years.
I hadn't realized it but I already do this for some things. I build things with Ruby on Rails infrequently enough that I need to reference docs a lot. I've been trained to not bother searching "the web", and instead go straight to the official guides[1]. I'm also Firefox user, so I've set up a dozen or so "keyword" searches[2]. "wiki " brings me to Wikipedia results for , similar for yf- yahoo finance, az- amazon, etc.. For me, this has been the killer feature for Firefox that would prevent me from using Chrome even if there weren't privacy/consolidation of power issues.
[1]: https://guides.rubyonrails.org/ [2]: https://support.mozilla.org/en-US/kb/how-search-from-address...
By all means continue to use your browser of choice, but Chrome does offer the same functionality for as many search engines as you'd like [1]. By default, the keyword for search engines you've used before is just the domain name of the website where the search engine is (which the omnibar is pretty good at filling in for you) but you can customize it to be whatever you want.
[1]: https://support.google.com/chrome/answer/95426?hl=en&co=GENI...
good to know!
I'm pretty sure that DDG also offers this type of filtering through their "bang" shortcuts. [0]
Enable personal blacklists.
If a site is found in more than x blacklists, manually take a look at it and decide either to remove it or keep it.
I'm pretty sure I could easily bot this though - I'm not an expect web scraper/automator by any means, my experience with PhantomJS is probably quite out of date, but what's stopping me from running x,000's of little scripts on a x,000 different VPS's (or a botnet, if I'm a bit evil), setup a unique looking User Agent and client environment in case of any finger-printing, and adding my competitor to a blacklist x,000 times? I'm pretty sure I could make it all look human enough. The manual check doesn't stop me from flooding the moderation queue at the very least, and perhaps the mod gets genuinely fooled, or convinced that my competitor must be up to no good.
What if that option only exists for paid DDG members, though?
Sure, you could game the system. But the drag on your finances would be that you would have to manage thousands of paid DDG accounts.
> but what's stopping me from running x,000's of little scripts on a x,000 different VPS's (or a botnet, if I'm a bit evil)
Price. You have to be seriously dedicated or have serious economic muscles to pull this off.
Unfortunately, you only have to compete with real users using that specific feature, and there aren't that many when measured against the power of even cheap attacks.
Then again:
1. even enabling personal blacklists would be a huge benefit to power users even if they aren't aggregated.
2. the user I replied to above suggested running n 1000 vps-es or using a criminal botnet. Both of these are significant hurdles.
Also while running n 1000 vps-es isn't a significant hurdle in todays clouds if it is a one time job, it at least becomes expensive if the blacklists are generated based on active users at random times around the clock forcing would-be abusers to not only fire up their vps-es, running a couple of queries and shut them down but also to keep them running around the clock, or rather in 8 hour shifts to not trigger abuse detection.
Oh that it was my expectation. A human would probably need to validate this. They might not be able to validate ALL submissions, but prioritize the ones most people report and it will still improve the results A LOT.
Build usage profiles on each client you encounter, calculate reputation scores, and weight the reports accordingly?
"Build usage profiles" is pretty much on the opposite pole to DuckDuckGo's whole 'privacy, simplified' deal.
Become Google then?
Google isn't Google anymore so maybe they should.
Verified volunteers.
In the case of plagiarism, require proof, make the reporter do the work. Ask them for the url for the alleged original source.
Simple. Enable Captcha! /s
Per account reputation. A bot will either perfect the reports in which case they are working in your favor,or be completely random, in which case you shadow ban/ignore.
Now you’re tracking users, undermining one of their USPs.
how about a blacklist browser extension, and like pi-hole, simply share lists of bad domains to omit from search returns.
If that is being done on the client side, then server sends results but extension blocks all of them resulting in an empty results page for viewer.
This needs to be done server side so that blank result pages are avoided. It would improve service for all as it'd be a better product all around.