Information scent.  In the  Information foraging theory an information scent is obtained when seeing an element that gives us an idea of the information source that produced it. For example a link can give us cues about the information it refers to. In the picture, the "trail of information footprints" represents the scent from the information source.  Source: representation by the author. Click on the image to enlarge it.

The search for and digestion of information on the web is one of the daily tasks to which we devote more time in Internet and it’s, probably, among the ones that cause most difficulties too.

One of the most interesting topics is, consequently, which strategies may be more appropriate in order to improve that search and how to implement them in the systems that serve and gather said information.

There exists a very interesting approach from the conceptual standpoint: the Information Foraging theory. This theory has its origins in the early 90s when Peter Pirolli and Stuart Card were studying human behaviour when doing intensive information search. 

Pirolli knew the foraging theory used by anthropologists and ethologists to explain the foraging strategies of animals and he recognised quickly that the human behaviour when searching for information was similar to that of the hunter-gatherers and animals in search for food. 

In 1997 G.W. Furnas introduced the idea that the representation of an object (for example a hyperlink) holds a “residue” that makes it possible to perceive what lies beneath it. 

The same year Pirolli converted the concept in the idea of “Information Scent”. According to Card this “scent” can be assimilated to the user’s “(imperfect) perception of the value, cost, or access path of information sources obtained from the proximal cues, such as WWW links”.

In this sense the proximal (nearby) information scent would give us an idea of the existence and, maybe, a vague localisation of the distal source (far away in terms of access) of the same.

Biologists created the foraging theory in the 70s, trying to explain some aspects of animal behaviour. Particularly they wonder what makes the predator choose between an easy to catch, but low energy contents prey ( for example a calf) and one of bigger energy return but more difficult to catch (an adult animal).

Another of the key questions is when to move to another region richer in food (information in our case), since the more exploited a region, the less food is available.

It appears that animals choose those strategies that maximise the benefit per cost unit. In other words they try those strategies in which a minimum consumption of energy gets the maximum energy in the form of food. Pirolli and Card define the profitability of an information source as “the value of information gained per unit cost of processing the source”

On the other hand it turns out that the optimal strategy to change foraging area within a region is by doing that when the return that is obtained in a particular area is lower that the average of all the region. This constitutes the so called “Marginal Value Theorem” developed by Charnov in 1976. It has served to model behaviour in fields as different as ecology and land tenure.

According to Pirolli, Card and other authors, it turns out that the behaviour of web users fits quite well with that of our hunter-gatherer ancestors regarding information seeking. It appears rather logical since thanks to those skills we were able to survive. More specifically, web surfers are able to quickly find the optimal path to the desired information. The problem is that in many web sites this path is considerably long and winding, if it exists at all.

So the idea behind this is 

• to create representations of information with a strong information scent, 
  
  
• optimising the information architecture to produce short, direct, optimal paths and
  
  
• to create informational regions where the information supply balances the consumption ratio (relevant information already acquired is no longer relevant, although it was still “there”).

According to Card and others* the features that have developed spontaneously on the server side are:

• Indexes like those of Lycos that create an inverted index trying to visit every page by following all its links beginning with one in particular. The user then issues a query and gets a set of links, that could be interpreted as a “foraging region”
  
  
• Category trees that group links, like the Open Directory Project that produces foraging paths and regions. 
  
  
• Web pages created by many users about specific topics, which gather lists of links with a common information scent.

InfoVis.net is an example of the latter case, where we gather “informational food” of a same type.

As we’ll see in the next issue some systems that take advantage of this model have already been constructed. Companies like Amazon.com or Google.com take advantage of different aspects of the same to favour information foraging.

Nevertheless, researchers face several challenges:

• It’ isn’t so easy to calculate the cost of an action on the web as it is in the natural world. We need an appropriate definition of “energy”. Information value is difficult to calculate.
  
  
• The initial foraging theory has been increasingly refined by including more and more factors beyond the initial ones that provided a simplistic model. The same happens with its informational counterpart: there’s still a lot to discover about human information foraging. 
  
  
• Information ecologies are eminently dynamic, while foraging theories tend to assume that the environments change little with time.  

Information visualisation takes advantage of the interdisciplinary approach. Maybe we’ll have to begin thinking about the creation of an Information Agriculture.

See also

• * "The WebBook and the WebForager. An Information Workspace for the World Wide Web". by  Stuart K. Card, George G. Robertson and William York.   
• "Surf like a Bushman", by Rachel Chalmers, New Scientist magazine, 11 November 2000.  
• "Information Foraging Theory Applied", by Tanya Rabourn.  

Links of this issue: