Information Foraging: A Theory of How People Navigate on the Web

Will people scroll on a webpage? How do they decide to click on a link? When do they leave a webpage? When do they prefer to search and when do they browse? How do they decide to search for information in a mobile app or on the web?

These and many other questions about web-user behaviors can be answered by the information-foraging theory.

In this article we present a broad overview of the theory and review some of its implications for web design.

Analogy with Animal Foraging Behavior

Information foraging was developed at PARC (former XEROX PARC) by Peter Pirolli and Stuart Card in the late 1990s and was inspired by animal behavior theories about how animals forage for food (hence the name). Thus, not surprisingly, animal-foraging and information-foraging theories share common terminology, as listed in the table below.

What Is Information Foraging?

Information foraging is the fundamental theory of how people navigate on the web to satisfy an information need. It essentially says that, when users have a certain information goal, they assess the information that they can extract from any candidate source of information relative to the cost involved in extracting that information and choose one or several candidate sources so that they maximize the ratio:

Rate of gain = Information value / Cost associated with obtaining that information

In other words, if people have a question, they will decide which webpage to go to based on (1) how likely it is that the page will provide an answer to their question, and (2) how long it’s going to take to get the answer if they go to that page.

(Animal-behavior science shows that a similar type of optimization holds true for animal foraging — hence, the optimal foraging theory that served as source of inspiration for Pirolli and Card. Basically, an animal needs to eat more calories than it expends, or it will starve and ultimately die without offspring. Across many generations, animals have evolved highly optimized food-foraging strategies.)

In layman terms, information foraging explains why people don’t scroll mindlessly or click on every single link on the page: because they attempt to maximize the rate of gain and get as much relevant information in as little time as possible. Scrolling or clicking a lot more would probably gain the user more information, but in the user’s estimation, the rate-of-gain ratio would decrease, because the numerator (the information value) would increase too little compared to the increase in the denominator (the interaction cost associated to getting the information).

You may wonder how people can be so rational as to always take the action that maximizes their gain. After all, we have countless examples of people behaving irrationally, against their own interest. In fact, human behavior can be well described by what Nobel-prize winner Herbert Simon called bounded rationality. Whereas the choices that people make attempt to maximize benefit and minimize cost, humans have a hard time precisely estimating benefit and cost and thus use satisficing and other imperfect heuristics to pick the most promising choices. The overall effect is still that of aiming for the highest rate of gain, even if the achieved number may not be the theoretical optimum in every case.

Specifically, people have no way of knowing in advance (1) how much information a patch contains; (2) how much time it will take them to extract that information. For a given task, they do know however how much real time they spent so far and how much relevant content they were able to get. In order to choose what source to look at next, they estimate how the rate of gain will change if they choose to explore a specific information source. Of course, the judgment won’t be perfect — it will simply be based on whatever external cues the information source emits.

Information Scent

How do people estimate how much information they will get from a page before visiting it? That’s where the concept of information scent comes into play.

Animals decide which patch to forage on based on (among other things) the scent that they catch from the environment: if the scent signals the food that the animal is interested in, it may decide to pursue it.

Similarly, as a user searches for information on the web, she judges the webpages she encounters based on how well suited they are for her goal. Each source of information thus emits a “scent” — a signal that tells the forager how likely it is that it contains what she needs.

Note that, like in the animal-foraging world where different animals eat different foods, the scent is highly depending on the type of information that the user is interested in. Thus, a carnivore may be attracted by the scent of blood, whereas an herbivore may be completely insensitive to it and maybe prefer the scent of ripe fruit or fresh grass. Similarly, the same source of information can have a great scent for one user because it holds exactly what he’s interested in, but zero scent for a different user with a different information need.

What makes the scent of a webpage? When a person lands on that very page, the scent is given by the title, images, and the information that is easily visible above the fold. If the user is searching for dish towels and lands on a site with pictures of strawberries, beer, and candy, she may assume that this page is unlikely to contain what she needs simply because the scent points into a different direction.

When a person looks at a link to a page, the scent is given by all the words and images associated with that link. Thus, the same person looking for dish towels may be strongly attracted by a link called Kitchen linens next to an apron, a kitchen glove, and several dish towels.

Costs of Addressing an Information Need

Remember that the rate of gain equation above includes two different variables: the information value and the cost of obtaining the information. There are two types of costs associated with obtaining information: (1) the actual time and effort involved in extracting the information from the various information sources, and (2) the opportunity cost — resulting from foregoing the benefits of exploring other documents in favor of the chosen ones.

(In a for-pay system, there will be a third cost in terms of the monetary price of each document, but since most websites are free to access, we won’t discuss payments further in this article. The important point is that users’ time and hassle are real costs, even if they don’t come with dollar signs attached.)

Opportunity Costs

Whenever users decide to inspect a webpage, they potentially lose the opportunity of looking at something else. That’s why it makes sense that, once users have landed on a page, they quickly extract the gist without spending time delving into the details. Scrolling down and reading every single word on a page is counterproductive — the utility derived from such an action is likely to be too small, whereas moving on and surface-scanning a different page may prove way more profitable.

Time Costs: Between-Patch and Within-Patch

The time reflects the effort involved in gathering the information that the user needs.

Looking up information on the web or on a webpage usually involves two types of user actions:

1. Between-patch activities: Gathering information sources (i.e., patches)
2. Within-patch activities: Inspecting each patch to extract information from it

For example, if you want to research treatments for the common cold, you may start by running a search on a search engine. The search-results page will offer a set of links that could be potential sources of information. At this stage, you’re working between patches, gathering information sources that you will explore later.

Once you have your search results, you will next click on the most promising links and read through their content to extract relevant information. At this stage you are working within the individual patches, inspecting them and extracting the information that is relevant to you.

Both between-patch and within-patch activities contribute to the overall time to satisfy an information need and both can impact the user experience in many ways. Users can create adaptations that allow them to minimize the time spent between patches or within patches. These adaptations are called enrichments.

Enrichments

Both the between-patch time and the within-patch time can be minimized by enrichments.

An enrichment refers to a user interaction, behavior, or strategy that aims to maximize the

utility of the information foraging. It can happen either between patches or within patches.

Think of enrichments as an extra tool that users may use in order to forage more efficiently. The tool may be something that they already have in their pocket (such as a learned behavior) or could be built on the spot and tailored to the specific patch (in which case the user must spend time creating it).

Behavior enrichments are the tools that users already have acquired and that help them extract information efficiently. These behaviors are adaptations that evolved over time and that proved to be successful in many situations in the past.

For example, over many interactions with the web, users have devised behaviors that minimize the time spent gathering patches or estimating information value of a patch. Thus, in order to avoid context-switching costs, millennials use page parking to clearly separate the information-extraction from the information-gathering components of the foraging. Or, in order to save time, users use F-pattern scanning on web-search results to be able to quickly assess scent without reading the entire test associated with a search result.

At the patch level, people have devised other scanning patterns (e.g., layer cake) to quickly locate the relevant content. They also tend to ignore banners or the right rail — another adaptive behavior in service of maximizing the information gain relative to the time spent on the page.

Interaction enrichments require extra effort from the user — these are the tools that the user needs to build on the spot, in order to maximize the efficiency of the information-foraging task.

For example, the user can spend time to think of specific keywords that best describe her query, hoping to increase the likelihood of relevant search results. Or she can set a lot of filters. Both of these actions are between-patch enrichments.

Within patches, people can also use strategies such as the use of within-page search to quickly locate content that is relevant to them.

Good User Experience Can Make Enrichments Unnecessary

Enrichments are risky for the user for two reasons: not only it is the case that some enrichments require users to pay an extra interaction cost or take upon themselves a larger cognitive load, but there is a chance that they won’t be right for the task. For example, a user using an F-pattern to scan a webpage may miss important concepts that don’t appear at the beginning of a paragraph. Or, a person searching for her first lawn mower may not have enough domain knowledge to come up with a set of relevant keywords to search for.

Now, imagine that the between-patch and the within-patch environment was one step ahead of users and adapted themselves to user needs so that the user didn’t need to create any special enrichments: these environments were already laid out to maximize the efficiency of information foraging. That’s what a good user experience is.

A good user experience involves web pages that are designed so that the user can get the maximum relevant information in the minimum amount of time.

Some optimization is already in place. Indeed, today’s search engines use sophisticated ways to rank the search results so that the most relevant ones appear first. They use common searches and personalize their results so that even if the user doesn’t take the time to refine her query, there is a high chance to get what she wants because the search engine “knows” her or because that’s what most people mean when they run a similar query. Autosuggestions for searches also reduce the time of thinking about these enrichments.

At the page level, the way in which information is laid out and presented on a well-designed page is suggestive of what’s most likely to be relevant on a page. For example, designers use scanning-friendly formatting such as bulleted lists, bolded keywords, descriptive headlines to help users find the information that is relevant to them.

Of course, the difficult problem is: each user is looking for a different type of food. She has a certain, very specific information goal in mind that may be different than that of your next user. How do you optimize your page layout and the information scent that your page emits so that it works for all users and for all goals?

The answer is: you don’t. You optimize for the top task that your page or site is supposed to address. That will take care of the many users who will be doing that task. It will also take care of the many users attempting to solve a different task, for whom your page is irrelevant — they won’t be tempted to click on your page.

Conclusion

When people look for information on the web, they attempt to maximize the rate of information gain over time: they want to get as much information as possible in the minimum amount of time. To reduce time, they estimate the information value of a page based on its information scent and use enrichments such as learned behaviors or interactions to improve the chance of quickly getting what they need in their information foraging.

Reference

P. Pirolli, S. Card. 1999. Information foraging. Psychological Review 106, pp. 643-675.