Predicting the user's navigational behavior in the exploration process of a hypermedia device is a quite difficult task due to the high number of cognitive and technical factors involved. To measure the quality of a navigational model, testers must expose it to real users by mean of usability testing. Although powerful, classic usability techniques result too complex and expensive for the kind of data required conducting a navigability test. Usability testing also misses the effect that the performance of the user's machine has on navigation task and is unable to record user spontaneous behavior.

In order to avoid these problems, we have developed ANTS, an Automatic Navigability Testing System based on data gathering agents. These agents are sent to the user's machine, where they record the user's behavior discreetly, measuring the performance of the user's computer too. This approach not only reduces the data gathering cost but also avoids the volunteer's sensation of being observed also, since the physical presence of human testers is not required.