Opt-in collaborative shared result ranking

Luis

ranok thank you for the interesting suggestion! I'll mark this thread for review to assess community support for this feature.

bobby-tables

If I want to be tracked, I'll just use Google directly. 😉

This is a slippery slope that I would encourage Kagi NOT to get started on.

I also suspect that in practical terms it delivers little benefit as it implicitly makes assumptions about user behaviour.

I also suspect the technical issues you highlight (tabbed browsing) are only the tip of the iceberg. So you're opening yourself up to keeping track of all sorts of edge-cases which will eat up unnecessary time and effort to debug,

I would much rather Kagi devoted their time and effort to delivering quality of the results without needing to snoop on users. There are still many things Kagi can do in this area, including one I highlighted myself today ... de-ranking discussion links that have no replies.

awefdev

bobby-tables

I mean, I'd like to point you to the Kagi Privacy Settings page, https://kagi.com/settings?p=privacy

Currently this option can not be turned on. Kagi does not save any searches by default. In the future we may add features that will utilize your search history and then we will allow you to enable this.

Clearly this is a path that Kagi is already considering going down. Sometimes things are better when they have lots of data points, and this is something that could be done while maintaining anonymity. As long as the Kagi team is straightforward about what they are collecting, how they are collecting it, how they are using it, and make it opt-in... I'd be interested in them running a pilot for it at the very least.

... A gigantic "anonymous mode" toggle would also go a long way.

lukvol

It might be helpful if we could mark specific search results in some way. As an example, if I'm doing research, I often do many search requests which are related, but not the same. This leads to certain search results popping up over and over again. Often, I forgot which page was helpful and which not. Thus, it would be nice if we could mark them in some way. I had the following in mind:

Allow to add tags to search results. This could be nice in combination with lenses (e.g., show only websites which match a search term and do have tags x, y but not z)
Allow for adding comments to a search result
Allow for adding an expiration on tags. I feel like often, these things would only be needed for a few hours or a few days. It would be nice if they would be removed automatically afterwards.

What also would be nice is if this could be shared with a group of other kagi users (probably also with the possibility for setting an expiration on how long this should be shared)

Vlad

lukvol Good idea, we plan to have a more sophisticated tool for colelctig bookmarks and searching them

Vlad

Making this sticky to get more thoughts.

bobby-tables

awefdev

awefdev Sometimes things are better when they have lots of data points

IMHO that's the of-cited lazy excuse of Silicon Valley ... i.e. "we need MOAR DATA dammit"

The only thing that happens is you end up with a stupid big-ass petabyte-sized database that you then waste money paying some cloud provider so store for you.

And then, inevitably, you've got all that juicy data, so some smart-alec from management decides it should be monetised.....

I genuinely think it is possible to do a lot of things without harvesting vasts amount of data.

You might say "well, we'll have an anonymous tickbox". Sure. But we have no way of verifying Kagi's implementation. Plus what if Kagi decides to store the data in identifiable form for $time before "anonymising" it.

As I said. Slippery slope. A pandora's box.

Recast

Google already uses linger data to rank results, and I believe Kagi uses Google's results/ranking as a source. So I expect this data is already implicitly part of Kagi's search results, and Kagi does not need to try to re-implement this. Especially since drawing from Google means drawing from a much larger set of linger data than the data Kagi could ever collect.

We also have the community domain rankings page, which is a more explicit & intentional way of accomplishing the same thing of downranking bad sites.

I would prefer if Kagi spent development time on other features.

silvenga

Recast Another thought is that Google's current linger data tends to promote larger popular sites - and the content farms that are designed to keep users on a site. There's also some evidence that's the linger data is only from Chrome users (and it's already well known that websites change based on browsers e.g. Safari users can be charged more for airplane tickets). Personally, most of the traffic to my blog is Firefox users (definitely, a type of user).

So all said, many users might consider the linger data from Google to be poor.

On the other hand, if this was opt-in e.g. users explicitly vote, that could reduce that bias of more popular sites, biasing to just really good sites (e.g. smaller blogs that have the exact answer you need).

slamor

A small feedback button can be added to the search results. When this button is pressed, the type and nature of the issue can be specified. These feedbacks are kept publicly, separately for each site, and the ranking is adjusted according to these feedbacks.

slamor

There could be a voting feature where each search result has a like and dislike button. After one month, sites with the highest dislike ratio would drop down in the rankings, while those with the highest like ratio would move up.

Recast

silvenga On the other hand, if this was opt-in e.g. users explicitly vote, that could reduce that bias of more popular sites, biasing to just really good sites (e.g. smaller blogs that have the exact answer you need).

I agree. Kagi already has this feature as domain rankings, and I think it's really good.

shooda

I see the utility in the idea but I don't want it. I don't want my results to be aggregated to others. My value of a one result may be lower or higher than anyone else. It seems it could be tragedy of the commons for Kagi search. Give me the ability to have the results without crowd sourced value because I hold no value for the crowd. It's also the beginning of tracking which will suck as that's not the value prop I signed up for.

slamor

shooda How will offering straight Google, Bing... results exactly the same make Kagi an independent search engine? It should produce original solutions for ranking issues and search quality. Offering the sorting ability to the user and saying "get it and handle it yourself" is not the way to go. As for tracking, tracking is prevented by anonymous voting per account for search results, and only voting information is recorded.

tarq

I’d like this without the collaborative aspect. If Kagi could suggest uprankings/downrankings based on my behavior for my own searches and let me decide if I want to apply them, I’d be all about that.

Thibaultmol

I'm also down. I think with something like this, each submitted report to the collaborated list should require an x amount of character reasoning for why you ranked it like that. So it forces people to put in some effort on explaining why

Jesal

I agree with most of the points that @bobby-tables made. I think that Kagi already has significantly better search results (and rankings) than other search engines without the need for this type of telemetry. I think the additional telemetry mentioned in the original proposal would not be worth a marginal improvement to result ranking, if it even provides that.
If necessary, I would prefer a solution where we explicitly leave positive or negative feedback on results and then that data could help improve result rankings. Personally, I don't like the idea of other people's opinions directly affecting the ranking of my search results. This is one of the reasons that I really like Kagi's already existing domain leaderboard since it gives the benefit of crowd sourced information while allowing me to rank based on what fits my personal context.