3 ms·

In their article “Independent and Unbiased Search Results” I believe Mojeek fundamentally misunderstands the goal of a search engine. There is no such thing as

In their article “Independent and Unbiased Search Results” I believe Mojeek fundamentally misunderstands the goal of a search engine. There is no such thing as an unbiased search engine. The entire purpose of a search engine is to be biased towards relevance. The opposite of biased isn’t fair, it is random. If you search for “cars” what should the engine show you? There are likely billions of pages containing the word. (And for some queries, the correct page won’t even contain an exact string match.) If you want to create a useful ranking, your first few results need to have identified the correct intent but also be sensitive to timeliness and location. https://blog.mojeek.com/2018/08/independent-and-unbiased-search-results.html https://blog.mojeek.com/2018/08/independent-and-unbiased-sea...

mellosouls 2y ago
The entire purpose of a search engine is to be biased towards relevance. The opposite of biased isn’t fair, it is random. The opposite of "biased" is "unbiased" - as in the title - meaning neutral or impartial here and I think you are possibly misinterpreting the context. The blog entry you link to is from 2018, at the height of the culture wars when the big tech players including Google were being criticised for pushing political agendas in what they displayed to the user. The (marketing) claim here is clearly suggesting Mojeek does not use an ideological filter.
- danpalmer 2y ago
  I don't think things are this simple though. Spaces that attempt to either have no moderation, end up over-representing those who hit up against moderation elsewhere. Having moderation requires having an opinion, and the opinion ultimately becomes the product.
  - mellosouls 2y ago
    Undoubtedly it's a difficult task, but I think there are degrees to which you can at least have a decent go at good faith moderation (eg see HN) and the article is appealing to a perspective that that good faith isn't there.
  - dartos 2y ago
    A search engine isn’t really a space though, it’s a tool. It’s not a social network. The job of a search engine is to return relevant (by some metric) results given a query string, not to show you posts you’ll agree with or which the authors of the search engine find tasteful. You might say that the relevance metric is a kind of bias, and you’re not wrong, but those biases can usually be empirically measured and understood, unlike the biases of unmoderated people in social networks.
    - danpalmer 2y ago
      I would indeed say that relevance measurement is the bias. You can build a bad search engine with simple rules like TFIDF, but pretty much anything more than that needs opinion and bias. For example, if there's a content farm that rips off content from another site, assuming you don't want to show both the original and the farmed content (which would be a bad search engine by most people's metrics), you need an opinion on which the original was. Sometimes you might know for sure, but often you won't. Even if you do know who the original was, you need the opinion that the original is better than the copy. When you have content farms providing summarised versions, now you need an opinion on whether summarised content has value instead of/alongside the original – summaries do provide value, but how much? None of this gets to politics or the legal requirements for handling things like hate speech or other content that is illegal in some/all regions, and you've already got essentially editorial control being exerted. Social media moderation and search engine results are fundamentally the same problem, it's just that the base interaction is a search rather than viewing a following feed or viewing a video.
      - janalsncm 2y ago
        Yes, exactly. If a group of anarchists puts up spammy websites optimized for TF-IDF, blocking or downranking them suddenly means you are biased against anarchists. The flip side is that history is littered with examples of rules which seem neutral on their face but which have very biased intentions. For example “literacy tests” for voting, or poll taxes.
- janalsncm 2y ago
  I understand that the term “unbiased” exists, and that this article was written promising a lack of political bias. I purposefully chose an anodyne example of “cars” to illustrate that even in a context that’s not politically charged, promising “unbiased” results is a fool’s errand. In particular, if you have any distribution of political opinion that is correlated with any other factor you consider in relevance calculations, there’s no way of avoiding accusations of bias. I’ll give two more examples. If I search for “best baseball team” (or if you like, “mlb game today”) how many results should be from Red Sox fans? If I’m in Boston? If I’m in LA? There is obviously a desirable geographic bias to the query. To take a more politically charged query, if I search for “Afghanistan war” there is a geographic bias (whether to show results for Soviet or American invasions depends somewhat on location), but also a temporal bias. Political support for American occupation decreased over time. So if your search engine favors recent results, it will have more negative sentiment than overall sentiment of all possible articles you may have shown. The final point is that these are not systems which are easily interpretable. Aside from explicit filtering, it’s entirely possible to start from reasonable rules like timeliness, text matching, and location matching, but still end up with skewed results and be accused of bias, because they are biased, just not in the way people think.
  - 8n4vidtmkvmk 2y ago
    I'm not a sports fan, but if I search "best baseball team" I would expect to find the team with the most wins, or the best current set of players based on their historical stats, or the team most likely to win the current or next season. So maybe a time bias (want to know what team is best now, not 10 years ago), but not a geography bias. I want to avoid echo chambers. I don't want to be told the team closest to me is best just because it might make me feel happy.
    - pickledoyster 2y ago
      Then why not search for "the baseball team with the most wins" or "best baseball team historically", "baseball team most likely to win in 2025", etc.? With a query like the one in your example, someone else might be completely justified in expecting to find the team with the best financials, the best little-league option for their child, etc. I believe such queries are a perfect case of GIGO, and the amount of effort spent on optimizing for them instead of giving users more agency is what's driving down both the experience (bad results) and usability (little to none operators and explicit filters) of search engines as tools.
photonthug 2y ago
> you want to create a useful ranking, your first few results need to have identified the correct intent but also be sensitive to timeliness and location. This fundamentally misunderstands the goal of a search engine too. This part is mainly for helping to sell ads and not rank by relevance. The correct way to rank results for a query like Rome is to return results related to a modern city, then ancient times, and keep the local pizza place at the bottom. Yelp/maps/search really don’t need to be all wrapped up in the same utility function, and honestly no one asked for this. It’s just been that way for so long that people have internalized the idea that it must be essential. Localization/trending content etc mainly serve the needs of advertisers, trackers, marketers.. not searchers. News and product search can/should probably be orthogonal, but instead we’re going the opposite direction and increasingly removing distinctions between maps/video/image/text search. This helps platforms to erode the distinction between user-directed organic discovery vs promoted content.
ColinHayhurst 2y ago
Information neutrality is preferred to unbiased (which we used in 2018) but a bit of a mouthful and not widely understood. For more recent and nuanced views on this key topic see: https://blog.mojeek.com/2022/05/freedom-to-seek-matters.html https://blog.mojeek.com/2022/05/freedom-to-seek-matters.html
pickledoyster 2y ago
There is a bias in who defines "relevancy". If the search engine implicitly sets the indexing and ranking algorithms, filters, etc., the results are biased. If the user can control the filters, set their ranking algorithm, then the relevancy (bias) is in the user's control. I can tell the search engine what it needs to show. There are billions of pages with the word "cars"? Alright, I want to see 20 pages that were indexed most recently, excluding sites that are flagged as spam/malware/SEO scams by a blocklist of my choice. I'd also add a contextual filter/category to exclude the animated movie and mechanical means of transport, as my intent is news about the band. No semantic engine, no synonyms this time, I only need exact matches, thanks. Let me do all this with a GUI or at least with search operators that are always respected. And allow me to save these settings as bookmarklets. Instead, there's way too much effort going into these black boxes of manure, which results in groupthink and uncritical repetition of claims such as "you want to create a useful ranking, your first few results need to have identified the correct intent but also be sensitive to timeliness and location" Here's the real bias: the company behind the search engine and its product managers are defining what is "useful", that it needs to "identify the correct intent" without giving the user any way to clearly state the intent, and that it needs to be "sensitive to timeliness and location" (what if it doesn't?). This is what Google does and SEOs keep parroting, while we look at the decrepit state of Google Search without wondering whether there might be a better way to build a search engine. Sure, a search engine like that would not be a mass market product, but it would not give its users results that fit the lowest common denominator of a population with a median reading comprehension level of an 8th grader.
e_i_pi_2 2y ago
Agreed with this, rewarding users doesn't make much sense in this context, I don't want to know what other users are publishing I want to know what's the best result across the internet. How much will "wikipedia" be using this search engine? They should probably be on the first page of a lot of queries but they wouldn't ever be using this. It's definitely a cool idea but I wouldn't call it a search engine, maybe something some "community search engine"? Definitely a good use for this similar to stackoverflow for teams where you can bias results towards your own network
mhitza 2y ago
> If you search for “cars” what should the engine show you? A definition of the word car, and a wikipedia link, should suffice. The more generic a search term is the fewer results it should surface, IMO.

mellosouls 2y ago

The entire purpose of a search engine is to be biased towards relevance. The opposite of biased isn’t fair, it is random. The opposite of "biased" is "unbiased" - as in the title - meaning neutral or impartial here and I think you are possibly misinterpreting the context. The blog entry you link to is from 2018, at the height of the culture wars when the big tech players including Google were being criticised for pushing political agendas in what they displayed to the user. The (marketing) claim here is clearly suggesting Mojeek does not use an ideological filter.

danpalmer 2y ago
I don't think things are this simple though. Spaces that attempt to either have no moderation, end up over-representing those who hit up against moderation elsewhere. Having moderation requires having an opinion, and the opinion ultimately becomes the product.
- mellosouls 2y ago
  Undoubtedly it's a difficult task, but I think there are degrees to which you can at least have a decent go at good faith moderation (eg see HN) and the article is appealing to a perspective that that good faith isn't there.
- dartos 2y ago
  A search engine isn’t really a space though, it’s a tool. It’s not a social network. The job of a search engine is to return relevant (by some metric) results given a query string, not to show you posts you’ll agree with or which the authors of the search engine find tasteful. You might say that the relevance metric is a kind of bias, and you’re not wrong, but those biases can usually be empirically measured and understood, unlike the biases of unmoderated people in social networks.
  - danpalmer 2y ago
    I would indeed say that relevance measurement is the bias. You can build a bad search engine with simple rules like TFIDF, but pretty much anything more than that needs opinion and bias. For example, if there's a content farm that rips off content from another site, assuming you don't want to show both the original and the farmed content (which would be a bad search engine by most people's metrics), you need an opinion on which the original was. Sometimes you might know for sure, but often you won't. Even if you do know who the original was, you need the opinion that the original is better than the copy. When you have content farms providing summarised versions, now you need an opinion on whether summarised content has value instead of/alongside the original – summaries do provide value, but how much? None of this gets to politics or the legal requirements for handling things like hate speech or other content that is illegal in some/all regions, and you've already got essentially editorial control being exerted. Social media moderation and search engine results are fundamentally the same problem, it's just that the base interaction is a search rather than viewing a following feed or viewing a video.
    - janalsncm 2y ago
      Yes, exactly. If a group of anarchists puts up spammy websites optimized for TF-IDF, blocking or downranking them suddenly means you are biased against anarchists. The flip side is that history is littered with examples of rules which seem neutral on their face but which have very biased intentions. For example “literacy tests” for voting, or poll taxes.
janalsncm 2y ago
I understand that the term “unbiased” exists, and that this article was written promising a lack of political bias. I purposefully chose an anodyne example of “cars” to illustrate that even in a context that’s not politically charged, promising “unbiased” results is a fool’s errand. In particular, if you have any distribution of political opinion that is correlated with any other factor you consider in relevance calculations, there’s no way of avoiding accusations of bias. I’ll give two more examples. If I search for “best baseball team” (or if you like, “mlb game today”) how many results should be from Red Sox fans? If I’m in Boston? If I’m in LA? There is obviously a desirable geographic bias to the query. To take a more politically charged query, if I search for “Afghanistan war” there is a geographic bias (whether to show results for Soviet or American invasions depends somewhat on location), but also a temporal bias. Political support for American occupation decreased over time. So if your search engine favors recent results, it will have more negative sentiment than overall sentiment of all possible articles you may have shown. The final point is that these are not systems which are easily interpretable. Aside from explicit filtering, it’s entirely possible to start from reasonable rules like timeliness, text matching, and location matching, but still end up with skewed results and be accused of bias, because they are biased, just not in the way people think.
- 8n4vidtmkvmk 2y ago
  I'm not a sports fan, but if I search "best baseball team" I would expect to find the team with the most wins, or the best current set of players based on their historical stats, or the team most likely to win the current or next season. So maybe a time bias (want to know what team is best now, not 10 years ago), but not a geography bias. I want to avoid echo chambers. I don't want to be told the team closest to me is best just because it might make me feel happy.
  - pickledoyster 2y ago
    Then why not search for "the baseball team with the most wins" or "best baseball team historically", "baseball team most likely to win in 2025", etc.? With a query like the one in your example, someone else might be completely justified in expecting to find the team with the best financials, the best little-league option for their child, etc. I believe such queries are a perfect case of GIGO, and the amount of effort spent on optimizing for them instead of giving users more agency is what's driving down both the experience (bad results) and usability (little to none operators and explicit filters) of search engines as tools.

photonthug 2y ago

> you want to create a useful ranking, your first few results need to have identified the correct intent but also be sensitive to timeliness and location. This fundamentally misunderstands the goal of a search engine too. This part is mainly for helping to sell ads and not rank by relevance. The correct way to rank results for a query like Rome is to return results related to a modern city, then ancient times, and keep the local pizza place at the bottom. Yelp/maps/search really don’t need to be all wrapped up in the same utility function, and honestly no one asked for this. It’s just been that way for so long that people have internalized the idea that it must be essential. Localization/trending content etc mainly serve the needs of advertisers, trackers, marketers.. not searchers. News and product search can/should probably be orthogonal, but instead we’re going the opposite direction and increasingly removing distinctions between maps/video/image/text search. This helps platforms to erode the distinction between user-directed organic discovery vs promoted content.

ColinHayhurst 2y ago

Information neutrality is preferred to unbiased (which we used in 2018) but a bit of a mouthful and not widely understood. For more recent and nuanced views on this key topic see: https://blog.mojeek.com/2022/05/freedom-to-seek-matters.html https://blog.mojeek.com/2022/05/freedom-to-seek-matters.html

pickledoyster 2y ago

There is a bias in who defines "relevancy". If the search engine implicitly sets the indexing and ranking algorithms, filters, etc., the results are biased. If the user can control the filters, set their ranking algorithm, then the relevancy (bias) is in the user's control. I can tell the search engine what it needs to show. There are billions of pages with the word "cars"? Alright, I want to see 20 pages that were indexed most recently, excluding sites that are flagged as spam/malware/SEO scams by a blocklist of my choice. I'd also add a contextual filter/category to exclude the animated movie and mechanical means of transport, as my intent is news about the band. No semantic engine, no synonyms this time, I only need exact matches, thanks. Let me do all this with a GUI or at least with search operators that are always respected. And allow me to save these settings as bookmarklets. Instead, there's way too much effort going into these black boxes of manure, which results in groupthink and uncritical repetition of claims such as "you want to create a useful ranking, your first few results need to have identified the correct intent but also be sensitive to timeliness and location" Here's the real bias: the company behind the search engine and its product managers are defining what is "useful", that it needs to "identify the correct intent" without giving the user any way to clearly state the intent, and that it needs to be "sensitive to timeliness and location" (what if it doesn't?). This is what Google does and SEOs keep parroting, while we look at the decrepit state of Google Search without wondering whether there might be a better way to build a search engine. Sure, a search engine like that would not be a mass market product, but it would not give its users results that fit the lowest common denominator of a population with a median reading comprehension level of an 8th grader.

e_i_pi_2 2y ago

Agreed with this, rewarding users doesn't make much sense in this context, I don't want to know what other users are publishing I want to know what's the best result across the internet. How much will "wikipedia" be using this search engine? They should probably be on the first page of a lot of queries but they wouldn't ever be using this. It's definitely a cool idea but I wouldn't call it a search engine, maybe something some "community search engine"? Definitely a good use for this similar to stackoverflow for teams where you can bias results towards your own network

mhitza 2y ago

> If you search for “cars” what should the engine show you? A definition of the word car, and a wikipedia link, should suffice. The more generic a search term is the fewer results it should surface, IMO.