How to Be a Better Web Searcher: Secrets from Google Scientists

Researchers who study how we use search engines share common mistakes, misperceptions and advic
By Daniel M. Russell and Mario Callegaro 

Credit: Getty Images

In a cheery, sunshine-filled fourth-grade classroom in California, the teacher explained the assignment: write a short report about the history of the Belgian Congo at the end of the 19th century, when Belgium colonized this region of Africa. One of us (Russell) was there to help the students with their online research methods.

I watched in dismay as a young student slowly typed her query into a smartphone. This was not going to end well. She was trying to find out which city was the capital of the Belgian Congo during this time period. She reasonably searched [ capital Belgian Congo ] and in less than a second she discovered that the capital of the Democratic Republic of Congo is Kinshasa, a port town on the Congo River. She happily copied the answer into her worksheet.

But the student did not realize that the Democratic Republic of Congo is a completely different country than the Belgian Congo, which used to occupy the same area. The capital of that former country was Boma until 1926, when it was moved to Léopoldville (which was later renamed Kinshasa). Knowing which city was the capital during which time period is complicated in the Congo, so I was not terribly surprised by the girl’s mistake.

The deep problem here is that she blindly accepted the answer offered by the search engine as correct. She did not realize that there is a deeper history here.

We Google researchers know this is what many students do—they enter the first query that pops into their heads and run with the answer. Double checking and going deeper are skills that come only with a great deal of practice—and perhaps a bunch of answers marked wrong on important exams. Students often do not have a great deal of background knowledge to flag a result as potentially incorrect, so they are especially susceptible to misguided search results like this.

In fact, a 2016 report by Stanford University education researchers showed that most students are woefully unprepared to assess content they find on the web. For instance, the scientists found that 80 percent of students at U.S. universities are not able to determine if a given web site contains  credible information. And it is not just students; many adults share these difficulties.

If she had clicked through to the linked page, the girl probably would have started reading about the history of the Belgian Congo, and found out that it has had a few hundred years of wars, corruption, changes in rulers and shifts in governance. The name of the country changed at least six times in a century, but she never realized that because she only read the answer presented on the search engine results page.

Asking a question of a search engine is something people do several billion times each day. It is the way we find the phone number of the local pharmacy, check on sports scores, read the latest scholarly papers, look for news articles, find pieces of code, and shop. And although searchers look for true answers to their questions, the search engine returns results that are attuned to the query, rather than some external sense of what is true or not. So a search for proof of wrongdoing by a political candidate can return sites that purport to have this information, whether or not the sites or the information are credible. You really do get what you search for.

In many ways, search engines make our metacognitive skills come to the foreground. It is easy to do a search that plays into your confirmation bias—your tendency to think new information supports views you already hold. So good searchers actively seek out information that may conflict with their preconceived notions. They look for secondary sources of support, doing a second or third query to gain other perspectives on their topic. They are constantly aware of what their cognitive biases are, and greet whatever responses they receive from a search engine with healthy skepticism.

For the vast majority of us, most searches are successful. Search engines are powerful tools that can be incredibly helpful, but they also require a bit of understanding to find the information you are actually seeking. Small changes in how you search can go a long way toward finding better answers.

It is not surprising or uncommon that a short query may not accurately reflect what a searcher really wants to know. What is actually remarkable is how often a simple, brief query like [ nets ] or [ giants ] will give the right results. After all, both of those words have multiple meanings, and a search engine might conclude that searchers were looking for information on tools to catch butterflies, in the first case, or larger-than-life people in the second. Yet most users who type those words are seeking basketball- and football-related sites, and the first search results for those terms provide just that. Even the difference between a query like [the who]versus [a who] is striking. The first set of results are about a classic English rock band, whereas the second query returns references to a popular Dr. Seuss book.

But search engines sometimes seem to give the illusion that you can ask anything about anything and get the right answer. Just like the student in that example, however most searchers overestimate the accuracy of search engines and their own searching skills. In fact, when Americans were asked to self-rate their searching ability by the Pew Research Center in 2012, 56 percent rated themselves as very confident in their ability to use a search engine to answer a question.

Not surprisingly, the highest confidence scores were for searchers with some college degrees (64 percent were “very confident”—by contrast, 45 percent of those who did not have a college degree describes themselves that way). Age affects this judgment as well, with 64 percent of those under 50 describing themselves as “very confident,” as opposed to only 40 percent older than 50. When talking about how successful they are in their searches, 29 percent reported that they can always find what they are looking for, and 62 percent said they are able to find an answer to their questions most of the time. In surveys, most people tell us that everything they want is online, and conversely, if they cannot find something via a quick search, then it must not exist, it might be out of date, or it might not be of much value.

These are the most recent published results, but we have seen in surveys done at Google in 2018 that these insights from Pew are still true and transcend the years. What was true in 2012 is still exactly the same now: People have great confidence in their ability to search. The only significant change is in their success rates, which have crept up to 35 percent can "always find" what they're looking for, while 73 percent say they can find what they seek "most of the time." This increase is largely due to improvements in the search engines, which improve their data coverage and algorithms every year."

What Good Searchers Do

As long as information needs are easy, simple searches work reasonably well. Most people actually do less than one search per day, and most of those searches are short and commonplace. The average query length on Google during 2016 was 2.3 words. Queries are often brief descriptions like: [ quiche recipe ] or [ calories in chocolate ] or [ parking Tulsa ].

And somewhat surprisingly, most searches have been done before. In an average day, less than 12 percent of all searches are completely novel—that is, most queries have already been entered by another searcher in the past day. By design, search engines have learned to associate short queries with the targets of those searches by tracking pages that are visited as a result of the query, making the results returned both faster and more accurate than they otherwise would have been.

A large fraction of queries are searches for another website (called navigational queries, which make up as much as 25 percent of all queries), or for a short factual piece of information (called informational queries, which are around 40 percent of all queries). However, complex search tasks often need more than a single query to find a satisfactory answer. So how can you do better searches? 

First, you can modify your query by changing a term in your search phrase, generally to make it more precise or by adding additional terms to reduce the number of off-topic results. Very experienced searchers often open multiple browser tabs or windows to pursue different avenues of research, usually investigating slightly different variations of the original query in parallel.

You can see good searchers rapidly trying different search queries in a row, rather than just being satisfied with what they get with the first search. This is especially true for searches that involve very ambiguous terms—a query like [animal food] has many possible interpretations. Good searchers modify the query to get to what they need quickly, such as [pet food] or [animal nutrition], depending on the underlying goal.

Choosing the best way to phrase your query means adding terms that:

• are central to the topic (avoid peripheral terms that are off-topic)
• you know the definition of (do not guess at a term if you are not certain)
• leave common terms together in order ( [ chow pet ] is very different than [ pet chow ])
• keep the query fairly short (you usually do not need more than two to five terms)

You can make your query more precise by limiting the scope of a search with special operators. The most powerful operators are things such as double-quote marks (as in the query [ “exponential growth occurs when” ], which finds only documents containing that phrase in that specific order. Two other commonly used search operators are site: and filetype: These let you search within only one web site (such as [site:ScientificAmerican.com ]) or for a particular filetype, such as a PDF file (example: [ filetype:pdf coral bleaching ])

Second, try to understand the range of possible search options. Recently, search engines added the capability of searching for images that are similar to given photo that you can upload. A searcher who knows this can find photos online that have features that resemble those in the original. By clicking through the similar images, a searcher can often find information about the object (or place) in the image. Searching for matches of my favorite fish photo can tell me not just what kind of fish it is, but then provide links to other fishing locations and ichthyological descriptions of this fish species.        

Overall, expert searchers use all of the resources of the search engine and their browsers to search both deeply (by making query variations) and broadly (by having multiple tabs or windows open). Effective searchers also know how to limit a search to a particular website or to a particular kind of document, find a phrase (by using quote marks to delimit the phrase), and find text on a page (by using a text-find tool).

Third, learn some cool tricks. One is the find-text-on-page skill (that is, Command-F on Mac, Control-F on PC), which is unfamiliar to around 90 percent of the English-speaking, Internet-using population in the US. In our surveys of thousands of web users, the large majority have to do a slow (and errorful) visual scan for a string of text on a web site. Knowing how to use text-finding commands speeds up your overall search time by about 12 percent (and is a skill that transfers to almost every other computer application).

Fourth, use your critical-thinking skills.  In one case study, we found that searchers looking for the number of teachers in New York state would often do a query for [number of teachers New York ], and then take the first result as their answer—never realizing that they were reading about the teacher population of New York City, not New York State. In another study we asked searchers to find the maximum weight a particular model of baby stroller could hold. How big could that baby be?

The answers we got back varied from two pounds to 250 pounds. At both ends of the spectrum, the answers make no sense (few babies in strollers weigh less than five pounds or more than 60 pounds), but inexperienced searchers just assumed that whatever numbers they found correctly answered their search questions. They did not read the context of the results with much care.  

Search engines are amazingly powerful tools that have transformed the way we think of research, but they can hurt more than help when we lack the skills to use them appropriately and evaluate what they tell us. Skilled searchers know that the ranking of results from a search engine is not a statement about objective truth, but about the best matching of the search query, term frequency, and the connectedness of web pages. Whether or not those results answer the searchers’ questions is still up for them to determine.

Source: blogs.scientificamerican.com