An option for less privacy friendly searching.

lootjes

I know a lot of people use Kagi as a privacy focused search engine. But personally I just use it because it gives better search results.
I would love an opt-in feature where it tries to learn from your search behavior to improve search results.

Most other search engines implement this feature and it works practically transparently to improve search results.

dreifach

Just want to chime in and say I am strongly against this. As soon as there is an alternative option for users (and for Kagi as the provider), incentive creep will ensue.

RoxyRoxyRoxy

In what ways are you envisioning this would learn from your history? We already have customisation options, through Lenses, domain filtering, Bangs, etc.

lootjes

For me, and I imagine moet people, domain filtering is far too manual to do with everything. Other search engines like Google and Bing learn from what you click on to give more tailored results. I can notice that logged out Google searches are far worse than kagi but logged in search is still superior.

Dustin

lootjes I would be very interested in some examples as this has not been my experience at all.

dreifach

lootjes I think this gets to the core of Kagi's philosophy, and it is something that may need to be publicly refined for its users. My own belief is as follows: algorithmic results that learn from your behaviour, intentions, and desires are a double-edged sword. While they can help in some instances, they do not deliver Kagi's mission to "humanise" the web. Personalising knowledge discovery risks Kagi Search/Assistant turning into individualised echo chambers for its users.

Manually adjusting what information one is exposed to is, I feel, fundamentally different to having it adjusted for you — once it is outside your control you cannot as easily, relatively speaking, control the information you are exposed to (I am, of course, slightly exaggerating here. Search results are nowhere near as bad as the kind of personalisation one may experience on TikTok or IG Reels).

Automatic adjustment is like going to a library or archive as a historian and being served the information that suits your inclinations. Manual adjustment is like going to a library or archive as a historian, speaking to the librarian or archivist, and then being pointed in directions that would help you with your research — while simultaneously acting as a check and exposing you to competing (yet not sensationalist) evidence. It is about a good faith process that you control rather than a bad faith process outside of your control. Importantly, it is also far more transparent.

empower5069

I agree with dreifach.

I don't just use Kagi to avoid Google using my search data to sell me things or train an AI, but also because the results are not tailored to an individuals algorithm.

Now, while I trust the Kagi team a significant amount more than Google to not be a bad actor and abuse the power that would give them, I would still be incredibly wary of using any personalised algorithm. It would be a step in a direction I would not wish to financially support Kagi with. I agree with the comments that it would be against their apparent philosophy.

If a feature like this was implemented I would probably cancel my subscription on principle.

blahblahblahkagi

I feel the downvoting and upvoting of specific websites kinda IS the personalized algorithm. I’d say maybe raising the limit on web filters could be a solution to something more personalized or adding new ones that can be switched on.