This is one vision of what a search engine's job is, but I'm not sure it's what most users are looking for. Even barebones pagerank goes far beyond just finding similar text - it uses structure of the web's link graph to estimate the quality of a page. Now obviously pagerank is not directly trying to ascertain which pages are truthful and which are not, but it is arguably using connectedness as a proxy - the assumption is that people tend to link to reliable, useful pages and tend not to link to incorrect, harmful ones.
Does vanilla pagerank also violate your boundaries for what a search engine should do?
I’m confident this is factually incorrect for a very very large subset of searches. Probably a majority.
No-click searches continue to grow in share while google continues to adopt design that presents information outside a traditional link, like a weather widget or the fact you were looking for presented at the top in large text outside the scope of a link.
This suggests a growing share of searches weren’t looking for links in the first place, they were looking for information and when the search engine provided that information, the user left without clicking a link
This is how google ended up with their modern design that discourages website clicks, they did research into their users and realized most people searching “Abraham Lincoln’s birthday” want to see the top result “February 12, 1809” and not the top result “https://www.loc.gov/item/today-in-history/february-12”
Obviously I’m aware that not all data is uncontroversial and agreed upon, but it’s a fact that a large percentage of searches are quite literally treating search engines as oracles of truth
It does not. I used the moon thing from last year as an example. Here[1] is a very noncontroversial query that still yields factually incorrect results in the box at the top. Yes if you read the results carefully and think about it you'll realize it's wrong but then you're doing the same thing you would be doing otherwise.
[1] https://www.google.com/search?sxsrf=APq-WBv0RNgr6Q3zUOjDc1lM...
In the "Abraham Lincoln’s birthday" example that the person gave, it absolutely does work that way in regards to the user's motivations.
When a user googles for "Abraham Lincoln’s birthday" They are almost certainly attempting to use a search engine as an oracle of truth.
> that still yields factually incorrect results
Search engines aren't perfect. That does not change the fact that the oracle of truth model, is a good model to describe a user's motivations.
Such as when they google "Abraham Lincoln’s birthday".
>This is how google ended up with their modern design that discourages website clicks, they did research into their users and realized most people searching “Abraham Lincoln’s birthday” want to see the top result “February 12, 1809” and not the top result “https://www.loc.gov/item/today-in-history/february-12”
But this is a false dichotomy. There are more options than either an infobox that says "February 12, 1809" and a "February 12 in history" page. I think you can agree that a link to the Wikipedia article on Abraham Lincoln will be a fine search result. At most the search engine needs to be smart enough to match "birthday" in the search query against variations of "born on" that are present in the article.
I am not saying that there is only a singular way of displaying information.
Instead the point here is that Yes, users who search that are actually trying to use Google as an Oracle of Truth here, and if the top results are anything other than what his actual birthday is, then that is a problem.
So in other words, yes it is being used as an oracle of truth, even if there are different ways of precisely displaying that.
> I think you can agree that a link to the Wikipedia article on Abraham Lincoln will be a fine search result
Only if the article links is actually the truth, and not some other number. So yes the search engine is being used as an oracle of truth, to find out what the actual answer is to what his actual birthday is.
Something doesn't need to be perfect to work well.
I would use a search engine that makes some attempt to rank higher quality, more accurate information higher than others.
The reason google removes factually incorrect results is because this is actually exactly what the vast majority of people use search engines for.
Yes, your users aren't using your search engine as an oracle for the truth, they're using it to find pages with similar text. It's their job (and specifically not yours) to determine how well the text they find corresponds with reality. Perhaps they're even aware the site is factually incorrect and enjoy reading it for the novelty, perhaps they want to cite it to warn others that sometimes things online are wrong. You have no idea what the people are doing with the information.
Google has started censoring their results because they decided people should use their search engine as a truth oracle (with the results you would expect, lets not forget the "when did George Washington go to the moon" thing from a year or so ago.) I think that's a mistake and is a large part of the reason I use DuckDuckGo.
You may have caught this one instance but there are likely an unlimited number of others. You're not going to catch everything and you make your search engine worse with every exception you make to the algorithm.