Human Hardware: Foraging with Search

In the last column, I looked at how Pirolli and Card theorized that we humans adapted our ancestral foraging strategies to retrieve information in a hypertext environment. The theory was first proposed in 1994, in a pre-Google era (although search engines were beginning to make their presence felt). If we look at the basic tenets of information foraging, the advance of web search introduces an interesting new wrinkle.

The whole concept of information foraging assumes that information is patchy and that there are costs involved in moving from patch to patch. The idea is to minimize the expenditure required to gain the information we’re looking for. It comes from the golden rule in optimal foraging: don’t waste more calories finding food than the food provides.

Calorie counting on Google

Obviously, caloric efficiency is not our primary concern in web interactions. It’s not like we burn a lot of calories exercising our mouse fingers. But old habits die hard. When we’re faced with a new situation we have to rely on old strategies, and when it came to navigating the web to find what we were looking for, we went back to the ancestral mechanisms our forebears used to keep from starving. In the process, we swapped information for food and time for physical energy as the resource to be conserved. This all happened subconsciously, below the level of our awareness.

So here we humans were, foraging through the online landscape looking for information, moving from patch to patch, when along came search engines, specifically Google. The reason web search was such a hit was that it took much of the pain out of information foraging. Suddenly we didn’t have to search for the patches. Through search, we could bring them to us. In fact, we could use search to create our own hyper-patch, made up of information scent cues from all the best patches available.

Time is precious

Although web search made information foraging much simpler, it introduced some interesting wrinkles that we have to bear in mind when looking at search interactions. Remember, time becomes the proxy for physical energy in online foraging. We subconsciously allocate a time resource that depends on a number of criteria: the value of the end prey (expected utility), our anticipation of how prevalent that prey might be (scarcity), our definition of diet (how specific is the information we’re looking for), how personally engaged we are with our information quest (a number of factors can influence this) and how pressed we are for the resource (how much time do we have available). Using these factors, we make a quick heuristic judgment on how much time we’re willing to allocate. Remember, this is almost never a conscious decision. It’s a gut feeling, a gnawing anxiety if resources are scarce (pressed for time) or a relaxed serendipity if resources are plentiful (casual browsing and killing time).

So, with each search, depending on intent, we assemble a new collection of information patches to consider. We scan the results, making judgments about the quality of the patches that lie on the far side of the blue hyperlinks. We know a click is a relatively small expenditure of time, but we also know that each click adds up. So the time we spend on the page and exploring the sites that come up in the results depends on the time we have allocated to the quest as a whole. If we expect the information to be easy to find, we’ll pass judgment quickly based on a cursory scan of the Golden Triangle. If we believe that the information will be harder to find, we’ll be willing to dig deeper, going below the fold and even to the second or third page of results. Our patience with information scent varies depending on our heuristic judgments about the odds of success for our quest and how relevant our searches are.

Foraging through results

In observing search behavior, we’ve seen how foraging behavior plays out on the results page. If we expect the information to be easy to find, we’ll decide how relevant entire blocks of results are based on the first few words we scan. If those first clues don’t pass the “sniff” test for scent, we’ll abandon the whole patch and move on. As we click through to landing pages, our quest for scent continues. We judge the relative relevance and the richness of information. We decide whether the patch matches the diet we had in mind, which we quickly defined with our query. And if not, we click back to the results page, there to move on to the next listing or relaunch our query.

There is no standard search behavior, because our strategies can vary from search to search. How far we scan, how quickly we click, where we spend the majority of time all vary depending on our intent. We do tend to take the same path through the results (the F shaped scan that forms the Golden Triangle) but our interaction with the actual results could look significantly different from search to search. For example, our search for scent will look different if we’re looking for consumer reviews on radial tires than if we’re looking for the address of a recommended restaurant and different still if we’re looking for a source for a hard to find item.

I’m not dumb, I’m efficient!

This also explains our query behavior. Since the beginning of search, the assumption has been that broad general queries are signs of laziness or ineptness on the part of the searcher. I’ve never agreed with this assumption. I think it’s our foraging strategies playing themselves out in a digital information environment. The goal is to find information as quickly and with as little effort as possible. Remember, although we’re never quite sure of the quality of the patch we’ll get when we launch a search, creating another patch for consideration is quick and easy. So, we begin by defining a broad diet, knowing that we can be more specific in our tastes if required with little effort. The search process is a heuristic give and take, with a quick structuring of queries balanced against quick judgments of the results. Being evolved omnivores, we have strategies for a broad and varied diet. Why should our tastes in information be any different? It makes sense to caste a wide net, start broad and spend more energy to refine our search only if necessary. You’re more likely to find something to match your taste if you start with an open mind.

Search allowed us to venture further and become more efficient in our foraging activities. It reduced the cost of abandoning lower quality information patches by bring a rich selection of alternative patches to our fingertips. The challenge for search at this point is to try to get an idea of our ideal diet, especially if we’re not sure. Another challenge is how to provide the best representation of information scent from the target sites. We’ve been conditioned to look for scent by repeated occurrences of our query in bold type in the title and description of the results. We supplement this explicit scent by implicit reinforcements we pick up from words as we scan listings. In our first eye tracking study we called this semantic mapping. These are words that connect to the implicit concept that surrounds our explicit query. If we search for Seattle restaurants, the semantic reinforcements could be the type of food that tickles our fancy, a location close to us, a chef’s name we recognize or even a price point. Although we didn’t include these qualifiers in our query (because it would require too much energy to do so) that doesn’t make them any less interesting to us if they should appear.

Gord Hotchkiss is CEO of Enquiro, a search marketing firm that produces search engine user eye tracking studies and other research. The Just Behave column appears Fridays at Search Engine Land.