Suicide results should probably have a "don't do that" widget like google

Vlad

The advertising mentions "strict content filters to ensure children are not exposed to harmful content" (src) which seems almost like Kagi is taking a moral stance that children need to be protected more than adults?

These are controlled by parents and can be turned off. Sane defaults apply (safe search on).

Kantu

Vlad Note regarding the concept of "Precedent"

Different law systems handle it differently. In CommonLaw vs CivilLaw. In CommonLaw it is handled as an absolute, therefore all laws and thinking needs to be thought of in the concept of what came before.

In CivilLaw, it is left to the jurisprudence of the Judge and Jury. Different cases which might share overwhelming similarities can reach completely different outcomes because the Judge overseeing the case considers it to be the right way to approach the situation therefore and hopefully achieving justice, rather than being hamstrung by what other judges in different cases and different situations decided.

I say this because I think that even when holding precedence, and ideally consistency in the direction of Kagi as a worthy goal. precedence in on itself is at least from my perspective, "something to keep in mind", and not a goal on itself. After all when metrics become targets they stop being useful metrics.

Anyhow. Even when I agree that it is generally a bad idea to meddle with wherever things the user is searching, even for a pre-determined index result. I do also believe that having your users kill themselves because they are on a temporary bad place when you could have made a nudging difference on avoiding it... Is generally "overly formal/bad practice" to your presuppositions of what "Kagi ought be"

Either way. Regarding scalability;

1.- Reduce scope. It doesnt need to be a expansive warning. Or have a phone number at all. It can be a simple personaly written message: "We at Kagi do not condone suicide, and do not recommend absolute decisions with absolute outcomes based on temporary issues." then machine translated to varied languages.
2.- Keep tied only to Suicide. Suicide is an overwhelmingly bad outcome. So just decide to break precedent and have wherever message you decide only appear when a search for suicide appears rather than a generalized thing.

But yeah, these are my perspectives on the issue. I read most of the thread and I kept seeing different people mention things and getting tangled on the concept of Precedent and then the issues of Scalability pop up, I believe that this comment addresses both points.

Carnival0852

Kantu

Suicide is an overwhelmingly bad outcome.

I believe in respecting the dignity of an individual's own end of life decisions. And I would cite the prominence of the Canadian government's MAiD program that provided over 13,000 people in 2022 alone with medical assistance in dying as strong evidence that this is not a fringe view.

Euthanasia and/or assisted suicide is legal or officially condoned in countries throughout the world including Belgium, (aforementioned) Canada, The Netherlands, New Zealand, Switzerland, The United States, and others.

People seeking to make their own, well considered, responsible end of life decisions should not be condescended to by a website they are paying specifically to give them unadulterated search results.

You can learn more about responsible end of life decision making at https://www.dyingwithdignity.ca/

briskroad1698

sbrl While I am a (PhD Student) Researcher, suicide research / prevention is not my field.

If you don't have expertise in this field, why did you consider it important to mention that you are a PhD student? Was it done in the hope of gaining more credibility despite having earned none in the matter at hand?

robertoftheedwards

This discussion is depressing. Here's my attempt to build on the initial (in my view extremely meritorious) suggestion and what people have already discussed:

Kagi already enables safe search and safe image search by default (at least that's what it seems like it does to me). Kagi thus already has a precedent for how it curates its results: it omits NSFW results by default to make the experience for users ostensibly more pleasant.

Instead of playing widget whack-a-mole, I'd suggest adding some settings to customize what "safe search" means for users. Right now, I presume it omits just adult material. Kagi could add two additional sub-toggles(?) that would both be on by default:

Sub-toggle 1: Omit adult NSFW material of a sexual nature.
Sub-toggle 2: Omit harmful content that could lead to loss of life or injury.

When a user searches for "how to sacrifice squirrels to Odin", Kagi wouldn't list "Top 5 ways to kill squirrels in the name of the Norse gods". If a user wants to uncheck either toggle, Kagi could perhaps surface a dialog that warns the user about what they might see in their future search results, and how just because customers are using Kagi for questionable ends, that doesn't mean Kagi endorses or condones them. Maybe also prominently surface suicide help resources the dialog for when the user unchecks the second toggle.

I can see this being a much more scalable way to ensure search results don't include questionable content that I don't think anyone should be seriously reading or consuming. It'd let the product operate on a more rules based approach rather than making individual widgets and experiences for individual topics. But Kagi would still be letting users configure their experience to be more unrestricted if they wished.

KagiForMe

Lurker here. I actually agree with @Vlad's scepticism (for the above mentioned reasons which I will not repeat).

But some people want this 'feature'. So why not implement @robertoftheedwards's proposal?

robertoftheedwards Instead of playing widget whack-a-mole, I'd suggest adding some settings to customize what "safe search" means for users. Right now, I presume it omits just adult material. Kagi could add two additional sub-toggles(?) that would both be on by default:

Sub-toggle 1: Omit adult NSFW material of a sexual nature.
Sub-toggle 2: Omit harmful content that could lead to loss of life or injury.

Combined witha sensible default, this seems like a good compromise.

Thoughts?

Carnival0852

robertoftheedwards

Sub-toggle 2: Omit harmful content that could lead to loss of life or injury.

Just out of curiosity, would you include, say, joe rogan discussing ivermectin as a covid treatment in that category? What about joe biden saying the "if you get the vaccine you can't get covid"? Who decides what is "harmful" or what the threshold for "potentially lead to loss of life" is?

I don't have the family plan, so I don't know exactly how it works, but wouldn't a much better feature be the option to toggle customizable 'lenses' for your kids that could exclude a user-definable list of keywords and sites? Then parents could each decide for themselves what they consider to be harmful for each of their own kids.

Whom

What Vlad seems to miss here in my view is that including 0 "nudges" requires just as much defense and is just as much of a position that must be justified as including 1, 2, or so on. Slippery slopes do exist, but no one in this thread has given a reason why including 1 message like this means they will be compelled to do 2 or 10 or 10000. For the most part I just want Kagi to get out of my way and do not want its commentary on my searches despite recognizing that search is a task which is inherently a matter of balancing biases.

But like, what's the practical reason not to include this? The very worst it can do is nothing, and its goal of saving lives is uncontroversial. If the only reason not to do it is a fear of the slippery slope, then please tell me what makes the slope so slippery. Do you really think someone will see that one warning message and think "Oh, this gives me ground to argue there should be more warning messages about other stuff! I'll go to their forums now and annoy them!"? Of course they won't, you'll get requests for adding warnings for whatever pet issue a person has regardless. And whether you have 0 or 1 or 9999, your choices are to ignore it or to argue for your position. There's nothing saved by choosing 0, no additional consistency in your principles which you cannot get by making a different choice.

andrea

Vlad

Well I ended up paying for a subscription so so much for voting with my wallet.

I do want you to know that I read the entire thread at the time that I made my comment.
I'll try to address the items you pulled out now.

Interesting, where do you draw a line there for providing a 'help widget'? What if you want to rob a bank? Kill an animal? Commit fraud? Hack someones computer? When should we decide that we should "intervene" and produce something the user did not search for? If we do, there is endless ways more can be requested, annoying people who are genuinly curious.

Since I proposed a widget that appears outside of the actual search results, the intervention is minimal and still does produce the desired results, but simply offers an alternative "off ramp". Expanding to other intervention areas doesn't seem very useful, but could still be possible and disable-able in the same way that I mention for hardcore self-harm/crime users.

The second is more philosophical but I think the unresolved portion is

Thus the best option for us is to simply refuse to make the first precedent no matter what the pressure is and stick to search being search. Perhaps one day Kagi may become your 'assistent' with personalised biases, but for now it is just a search engine.

In which case I am suggesting that this harm reduction strategy is, indeed, a bias towards trying to dissuade suicide. And that's ok. And if people think it's not ok, they can turn it off, and never see a little widget above their search results for suicide.

And because I also read the replies afterwards I want to discuss Whom's message

But like, what's the practical reason not to include this? The very worst it can do is nothing, and its goal of saving lives is uncontroversial. If the only reason not to do it is a fear of the slippery slope, then please tell me what makes the slope so slippery. Do you really think someone will see that one warning message and think "Oh, this gives me ground to argue there should be more warning messages about other stuff! I'll go to their forums now and annoy them!"? Of course they won't, you'll get requests for adding warnings for whatever pet issue a person has regardless. And whether you have 0 or 1 or 9999, your choices are to ignore it or to argue for your position. There's nothing saved by choosing 0, no additional consistency in your principles which you cannot get by making a different choice.

Yes, this certainly seems like the best one-time exception, never to be repeated.

Carnival0852

Whom

including 0 "nudges" requires just as much defense and is just as much of a position that must be justified as including 1, 2, or so on.

Deciding between zero, and one, two, 9999, etc, nag screens isn't, in fact, arbitrary. There is an asymmetry between zero and non-zero; It takes manpower and money (my money!) to build maintain, and support each non-zero nag. It takes none of my money to have zero of them, and allows those developer resources to be used on actual search features.

What would Kagi spending my money to build this feature for you provide you that you installing the Ripple suicide prevention browser plugin wouldn't? The browser plugin is objectively a more comprehensive and robust solution as it works across the whole web instead of just one domain.

Vlad

KagiForMe If they are opt-in, that means the person 'in need' would need to know to enable them first, which kind of defeats the purpose.

This has been a long lasting thread around a complex issue and I thank everyone for their input so far. I am, as many of you are, emotionally moved by this question on a personal level. At the same time I do not want to allow the business that is meant to serve users globally be led by mine or any employee's personal emotions and biases and this is why for Kagi we have principles of operation. One of them is that Kagi is focusing on the business of search and providing the best search results in the world, avoiding precedents that would take it off that path.

In general I would want these three things addressed:

Evidence that Kagi results for suicide queries are currently not adequate. When I check them I conclude that situation is completely opposite and that Kagi has the most pro-life results (in context of suicide) of all search engines in the world already. The reason this is the case is that we surface many more results from personal/blogs sites (as a part of our humanize the web mission and Small Web initiative) and many of them have personal stories of attempt/regret etc. I'd argue therefore, that even without the widget results are more humane than any other search engine. Being able to get real stories from real people in results is a much better way to handle this in terms of desired outcomes we all want to see, than any sort of generic widget (and this is all happening completely organically by Kagi's design).
Evidence or research that suicide hotline widgets actually help in meangiful way. (if they are as effective, such evidence should be possible to produce)
If Kagi is to cross the line here and interfere with results for moral reasons, what principle would we use to handle further matters regarding queries about for example homocide, torture, abuse, animal cruelty and other queries that are sure to get scrutiny from another group of users down the road, asking why the cause they care about does not have a warning widget when we have one for suicide, and they feel much stronger about their thing. How do you deem one set of those morally acceptable to be left out unattended and intervene with others? This is the slippery slope question.

robertoftheedwards

Vlad Asking for evidence is fair. I have a couple ideas about that, but I think they'd be a bit off topic for this thread.

Re: the principle, I've personally found a lot of apps and services don't appreciate enough what it means to have some feature be default on or not be meaningfully configurable for every single person, globally. Suicide is easy to say it's bad across cultures. Other topics - say religion - can be a lot more polarizing. I've always liked how many ad blockers like uBlock Origin and others let users easily define their own rules in addition to whatever defaults the service comes with out of the box. Many times users can throw those same defaults out the window if they don't like them.

If and when Kagi wants to wade further in (more than having safe search) to the murky waters of curating search results based off the subject matter of the results, I'd really suggest you do so in a way that lets people easily define the rules that curate the results in such a way that they can feel that Kagi is reflecting their morals and ethics, and that Kagi isn't imposing its own on them.

In a possible future for such a feature, if I were initially guided through that feature and told "this rule omits results about harming people or animals, you can turn it off though", I'd probably keep it on. I'd still really appreciate in addition being told "you can fine-tune this later as well if you find it's interfering with your experience using Kagi."

Carnival0852

[Removed for excessive snark]

Vlad

Carnival0852 Let's try to keep a language of constructive discussion.

Having said that browser extension is not a bad idea in general, one that could offer necessary support following the user around the web.

Carnival0852

Vlad

Hey look, a quick Kagi search found exactly this thing:

https://www.internetmatters.org/hub/resource/ripple-suicide-prevention-tool/

WoundedMoose

Carnival0852 "Death with dignity" is very different from straight-up suicide.

Kantu

Carnival0852 your personal feelings should dictate what search results I have access to.

The results to Suicide would not be tampered with. What's being argued for is if there ought be a warmly written phrase indicating that "Suicide is generally not a good idea" at the top of the results

The fact that you are mixing both cases is highly indicative to me that you are not arguing from a point of honesty on the topic

Carnival0852

The Canadian government providing Assisted Suicide is not the same as Kagi as a private company needing to either condone suicide and its outcomes or tacitly assist/not dissuade its users from committing it

Carnival0852

WoundedMoose

"Death with dignity" is very different from straight-up suicide.

How so? Especially under regimes that ban medical professionals from providing assistance or even information to patients about end-of-life planning. Responsible people making their own well considered end-of-life decisions may need to resort to more home-brewed approaches.

acuteaura

Vlad,

At the same time I do not want to allow the business that is meant to serve users globally be led by mine or any employee's personal emotions and biases and this is why for Kagi we have principles of operation.

You are already interfering with the small web initiative (as you've said prior). I personally love this for my own search results, but it does skew towards content where the authors have the technical means and conviction to sideline the typical writing tools people use for either ease of accessibility or reach (e.g. Medium, Substack). To be blunt, you built a technocracy web initiative.

Evidence or research that suicide hotline widgets actually help in meaningful way.

That's a very utilitarian view. Morally, you may just not want to be involved in facilitating suicide. It was irritating when we exchanged emails and it is irritating to me now that you do not seem to grasp the difference between suicide and assisted dying. They have vastly different motivations and suicide, being the much more common, is usually linked to an acute state of crisis. Even though I'd generally be in favor of assisted suicide being available, there's still an ongoing discussion about the possibility of induced demand for such in Canada. Neutrality is not a neutral moral stance.

How do you deem one set of those morally acceptable to be left out unattended and intervene with others? This is the slippery slope question.

I think this is the hardest question in the room, but the absence of absolute answers does not mean you can avoid the question. Saying the line is fuzzy and concluding the line ends right below the wall of one side of the room because you are uncomfortable by the existence of the problem are entirely different things. I don't think you can claim to "humanize the web" and not answer this question for yourself and for your users when you act like every other large corporation and do the legal minimum.

Frankly, I personally don't consider interference with any of these queries a problem as long as it's reasonably established that's actually what someone is searching for (including intent, which is hard, and why most of these wouldn't work) and you are transparent about the request having restricted or specially curated results, because lack of disclosure (as common pretty much everywhere else; even up to using such info for marketing) is much more dangerous. I also think places like that suicide cult forum (from the article I sent you) or Josh's doxing and bullying forum should indeed be removed from the index, with a note that unsuitable results have been removed (similar to Google's DMCA removal notice).

Vlad

acuteaura

To be blunt, you built a technocracy web initiative.

That would be unfair and wrong characterization of what Kagi Small Web is. There are many technical sites submitted, but I'd say on average most are not. Here is what the first few links I just got in Kagi Small Web (your mileage may vary, links are random).

https://thebluemoment.com/2024/01/16/written-in-air/
https://crossexaminingcrime.wordpress.com/2024/01/18/mr-monk-and-the-two-assistants-2008-by-lee-goldberg/
https://tamino.wordpress.com/2024/01/18/polar-vortex/
https://antonyloewenstein.com/trt-world-interview-on-netanyahus-refusal-to-accept-palestinian-self-determination/
https://ultimatemaitai.com/2024/01/19/the-last-bottles-of-denizen-3/
https://jonworth.eu/the-future-of-long-distance-train-services-through-the-channel-tunnel/
https://bookmaniac.org/2024/01/18/weird-sf-and-a-kids-book-binge/

Does it lean geeky/nerdy? Perhaps. But that is the nature of the content independently written and maintained on the web these days. Most non-geeky people do not have a website and post on Facebook/Instagram instead.

That is exactly what Small Web is set to change by giving everyone a platform to surface their content and everyone can submit their blog or friend's blog. The only bias we are demonstrating here is willingness to give these humans a voice, regardless where they are coming from, something no other search engine does and this aligns with our mission to humanize the web.

Carnival0852

I personally don't consider interference with any of these queries a problem... I also think places like that suicide cult forum (from the article I sent you) or Josh's doxing and bullying forum should indeed be removed from the index, with a note that unsuitable results have been removed (similar to Google's DMCA removal notice).
acuteaura

I don't care what queries you personally don't think are a problem to interfere with, and I find it irritating that you think your personal feelings should dictate what search results I have access to. Please be respectful of other users who personally feel differently than you do.

acuteaura

The only bias we are demonstrating here is willingness to give these humans a voice,

The bias is that it is definitely no longer normal to understand the web this way. We live in a world where students no longer know what a folder in a filesystem is, and where explaining the Fediverse with "it's like email" yields confusion because email and gmail have been the same thing to these people and the notion that the internet is not a series of large monolithic services owned by megacorporations you sign up for is foreign. That's on one hand depressing, and on the other unfortunate reality.

I appreciate the wordpress.org link in there though.

Vlad

acuteaura

That's on one hand depressing, and on the other unfortunate reality.

I agree, and I hope you can agree that it is not Kagi to blame for this situation and that it is Kagi that is trying to do something about it, against all odds.

acuteaura

Vlad

I hope you can agree that it is not Kagi to blame for this situation and that it is Kagi that is trying to do something positive about it.

Oh, I have no doubt. That wasn't the point. As I said, I like this for my own feed.

But I also hope you can agree that it is still a bias, and maybe not the exact bias intended for. And that is already is a shallow bit of the slippery slope you are afraid of.

« Previous Page Next Page »