What is your internet search methodology ?

post by George (George3d6) · 2020-05-23T20:33:53.668Z · score: 15 (9 votes) · LW · GW · 1 comment

This is a question post.

Contents

  Answers
    7 quanticle
    1 Raven
    1 hybridhuman
    1 Yitz
    1 Sønderjye
None
1 comment

I feel like the heuristics we use to search random tat on the internet are an incredibly larger part of our lives.

I wouldn't make a claim as extended as "googling skills are on the largest determinants of success in life", but I wouldn't look very suspiciously at someone that tried to make it either.

So I'm rather curios what the "search methodology" of people here is.

Let me try to give an example by detailing mine, it's non-ideal since I apply it mainly subconsciously, but that's why I'm looking to improve it, also I'm probably missing some things ,but again, this is just an example not a claim that it's in any way "good.

a) Find a phrasing that is < 5-6 word long and type it into google
b) If a certain key word isn't showing up quote it
c) If am looking for a specific type of discussing type the name of a form together with it (e.g. LessWrong: Fobar falacy, SSC: edgy political thing, Reddit: Corporate sponsored product review, StackOverflow: copy paste of a tiny part of the debug log which I hope is unique enough for my problem)
d) If what I'm looking for seem high-falutin to be considered "scientific" consider google scholar instead
+ ublock origin with strict settings, disable 3rd party cookies, container addons for e.g. google and fb in order to somewhat limit amount of ads
+ sci-hub now for quicker paper reading (basically you just click it, it tires to find the doi of the abstract you are looking at and opens it in a scihub tag)

For me, at least, thinking of this requires trying to observe myself and query my subconscious, rather than thinking about it the same way I think about explicit actions/choices.

I'd wager it's the same for many people and thus a lot of critical components might be missing... but, there's not much that can be done about that.

Answers

answer by quanticle · 2020-05-24T22:45:11.150Z · score: 7 (3 votes) · LW(p) · GW(p)

Gwern has written extensively on how to use Google efficiently. Some highlights:

  • Use site: to search a particular site. For example, if I'm looking for the Ars Technica review of the Google Pixel 3A, I'll type: site:arstechnica.com Google Pixel 3A. Or, if you want get a link "Meditations on Moloch" quickly, site:slatestarcodex.com Meditations on Moloch
  • Don't be too specific -- people are bad at remembering specific words, so limit quoted phrases to two or three words
  • Learn the jargon of the field you're searching and use those phrases. For example, if the field uses "logistic regression" as a common approach, add that phrase to your search

In addition, I wouldn't bother trying to search sci-hub directly from Google. Instead, find the actual journal article you're looking for, copy its DOI number, and paste that into sci-hub.

comment by George (George3d6) · 2020-05-25T10:25:07.391Z · score: 1 (1 votes) · LW(p) · GW(p)

The Gwern article I was unaware of, I will check it out.

In addition, I wouldn't bother trying to search sci-hub directly from Google. Instead, find the actual journal article you're looking for, copy its DOI number, and paste that into sci-hub.

I was speaking of a sci-hub addon, which auto detects the DOI in a page you are reading and opens the article in scihub (i.e. to make find DOI -> open sci hub -> past DOI and search a single step of "click addon button")

answer by Raven · 2020-06-02T16:21:47.882Z · score: 1 (1 votes) · LW(p) · GW(p)

Mine is different depending on the question.

  • If it's an error message, I search the relevant piece of the error in quotes.
  • If it's a question I know the keywords for, then I search those keywords, and if I get lots of stuff/wrong stuff I add extra keywords to narrow. Ex: "flesh autoignition temperature" or "rdna 2"
  • If it's a general question that I don't have keywords for, then I phrase my query as a pseudo human-readable question. For example, "how far can you fall before dying" or "was leather armor used in the middle ages".
  • Sometimes I use the site: command. Most sites' built-in search is pretty crappy and it's easier to do it through google. The time range command also gets used a lot but I don't know the text command off hand so I do it through the GUI.
answer by hybridhuman · 2020-05-31T19:56:45.383Z · score: 1 (1 votes) · LW(p) · GW(p)

Off the top of my head, I can count three broadly-defined reasons I've used search engines, or objectives for which I've used them:

1. Looking for an answer to a question with a definitive/'finite'/narrow* answer, e.g. "What movies are showing in [cinema near me]?" or "where can dogs be sunburned?"

2. Seeking any and all information about a particular known topic, with a plan to collect and collate that information. My undergraduate dissertation was about three films released between 1999 and 2007, and I used Google a lot to visit the films' official sites and get a sense of what film journalists thought of them. This is also how I search places like JSTOR and Google Books.

3. Trying to name/define/delimit a topic or idea I don't know much about, almost always with the goal of moving that topic into box 2; e.g. naming a specific programme or advert I saw on TV, finding online forums or communities with a culture that appeals to me.

For 1, I generally search in full sentences without capitalisation or punctuation. In particular, I often use the "that film where X happens" or "that book with X character" formula, because I once read that Google deals with those questions very well: an observation borne out by my own searches, even for obscure media. This is definitely inefficient, but it's usually so easy to find results that the extra work is negligible. I also love going to the movies, and wonder if the "what films are showing in X" search is part of that ritual for me.

For 2, I usually search for two to four words (very rarely more), and will often go through several results in depth looking for useful information before going back to refine search terms. Say I wanted to read about Chaucer's dream visions - I'd type "chaucer dream poems" first, probably read one or two articles on general Chaucer poetry to check for useful links or small pieces of info I could grab, and then search something like "chaucer dream visions macrobius" to get more specific. Usually, I'm balancing breadth of potential results with specificity of topic, and trying to filter my keywords to get authoritative and/or trustworthy information. Usually the syntax in these searches goes broad topic --> subset --> subset of subset.

For 3, I will often be working off fragmented memories, and will use quotes a lot - find an identifying feature the object, idea or group I'm talking about must have to be the same as what I am looking for, and quote that in as many different wordings as I can think of. I will aggregate with more and more key words decreasing in perceived relevance as I continue to search, and will put in minuses if a result I don't want keeps coming up.

I will definitely be checking out Gwern's writing on this topic. Sorry this is ridiculously long.

*is there a term which precisely conveys this idea?

comment by Raven · 2020-06-02T16:26:58.331Z · score: 1 (1 votes) · LW(p) · GW(p)
*is there a term which precisely conveys this idea?

A factual question. I had to google to remember the phrase. First was term that means specific question and then when that didn't work I tried term that means question with a definite answer and the answer was the top expanded result.

comment by hybridhuman · 2020-06-03T10:58:36.830Z · score: 2 (2 votes) · LW(p) · GW(p)

Thank you!

answer by Yitz · 2020-05-26T03:13:30.725Z · score: 1 (1 votes) · LW(p) · GW(p)

There are many different reasons to be searching something online, but when specifically searching to find the answer to some specific question I have, in general I tend to search in two different patterns—searching with keywords based on how I expect other people to ask the question I want answered (which will likely return results sourced from discussion boards, Quora, etc.), or searching with keywords based on how I’d expect someone to talk about the topic if writing a personal blog post or something. In general, the latter tends to use industry-specific wording, and either assumes deep familiarity with the topic or none at all, which can be hard to read through, while the former tends to use more generalized wording, which can sometimes make it harder to find what I’m looking for, but is also normally from users with a similar knowledge base, which can make the solution easier to understand.

answer by Sønderjye · 2020-05-24T17:00:31.277Z · score: 1 (1 votes) · LW(p) · GW(p)

My first step starts out searching searching in Ecosia. I haven't looked deeply into the exact impact but there are offhand enough evidence to convince me that the positive impact is worth the cost of a slightly worse search engine than google and slightly less privacy than duckduckgo.

If I can't find a satisfying answer in Ecosia I switch to google.

comment by Roko · 2020-05-24T22:17:57.155Z · score: 4 (2 votes) · LW(p) · GW(p)

It's probably only worth using ecosia if you think your time is virtually worthless. Tree planting is mostly a useless activity in the grand scheme of things, though one gets into a somewhat complicated calculation to show this.

Google claims there are 3*10^12 trees in the world, ecosia claims to have planted about 10^8. Trees are in an equilibrium with a bunch of stuff like atmospheric CO2 as well. Your personal contribution to the number of trees via ecosia is likely ~ a few trees (not clear whether they reach maturity though).

Do anything else useful with your time before this.

comment by Sønderjye · 2020-05-26T15:15:03.503Z · score: 1 (1 votes) · LW(p) · GW(p)

On the effectiveness of treeplanting: Would you mind linking to those calculations? Trees remove ~130 giga ton CO2 annually and with your number ecosias trees remove ~4*10^6 tonnes CO2 which is something like 10% of the worlds annual CO2 emission from fossile fuel. Granted, its a lot more complicated and it obviously isn't a feasible permanent solution given saturation and whatnot but it seems like an intuitive delaying measure, at least insofar as their forests take a good while to reach maturations.

On Ecosia: Their April financial statement says they spend 14% of their income on green investment so it's not exclusively treeplanting.

comment by George (George3d6) · 2020-05-24T20:49:35.126Z · score: 2 (2 votes) · LW(p) · GW(p)

Can't you simply e.g. donate 200$ each year to offset this ? E.g. google charge (I think) ~1$/click for a US demographic (some exceptions, blah blah) and how many search engine ads do you click ? For me it's ~0, but let's say... 100 a year ? add to that like 1$ hundred impressions + 10,000 searches a year. Granted, this is a very rough number, but I'm being rather charitable with the profit here, I think, considering a large part of that is actually operational costs.

It seems like your search data is hardly worth more than that, and the advantages of using google are many in terms of time saving. Enough to be e.g. worth 200$.

I get why one wouldn't want to use google for ethical reasons, but at the eod all the search engines which use a centralized structure are equally bad, they just happen not to hold a monopoly (however, in that case, if you're just anti-monopoly, you might as well use e.g. Bing which seems closest to google in terms of quality)

comment by Sønderjye · 2020-05-26T17:56:35.737Z · score: 1 (1 votes) · LW(p) · GW(p)

Thanks for helping me realize that searching on Ecosia does squat when I'm not clicking adds. Looks like they make money per click and also per puchase through adds.

1 comment

Comments sorted by top scores.

comment by Zachary Robertson (zachary-robertson) · 2020-05-23T21:07:38.714Z · score: 4 (2 votes) · LW(p) · GW(p)

I asked a related question [LW · GW] and got some answers about finding things on the internet. Didn’t completely satisfy me, but my question was significantly more vague so it might help you!