The area of animated agents used as an instructional technique is a well established research area, which receives contributions from fields as Educational Research, Computer Science, Artificial Intelligence, and Computing Machinery, among others.

The use of animated agents in computer-based training has been the object of research for a number of years. Koda & Maes (1996) have compared the effect of different agent features: geometric-shaped agent vs. human-like agent, caricature face vs. realistic face, text-and-sound agent speech vs. text-only speech. Similar studies have been carried out by Sproull, Subramani, Kiesler, Walker & Waters (1996), and King & Ohya (1996).

In the same line, Moreno (2001) has conducted a series of experiments, where she compared the role of the agent (social agency vs. no social agency), visual interface (fictional agent vs. video image of a real human being), and communication style (monologue vs. conversation).

Regarding the pedagogical style used by the agent, Baylor (2001) has carried out studies comparing two agents, one based on a predominantly instructivist philosophy, and the other one showing a predominantly constructivist approach. Students in this experiment would recognize the different styles, and they would choose one or the other depending on the kind of help they wanted (e.g., more structure in the instructivist version, "fuzzier" in the constructivist one).

Other studies have examined different agent personal characteristics (which affect their perceived pedagogical style). Hietala & Niemirepo (1998), for example, compare four agents with different genders and different skill level. Users would tend to prefer one or the other depending on their own personality type (introvert or extrovert), as well as the difficulty of the task.

Dehn & Van Mulken (2000), in their review of the most salient experimental studies conducted on the use of animated agents, have compared existing literature by classifying studies according to the kind of agents used (e.g., physical appearance, voice, etc.), the variables examined (both on the cognitive and the affective domains), and the results obtained.

This comparison of current literature shows that agents can be engaging (Koda & Maes, 1996), but also create an uncomfortable feeling (Sproul et al., 1996); they can be helpful (Van Mulken et al., 1998), but to some extent distracting (Koda & Maes, 1996); and users can interact with them (Takeuchi & Naito,1995), or they can ignore agents completely (Lester et al., 1997).

Dehn & Van Mulken's conclude that, given the dissimilar results shown by current literature, it seems to be that the effectiveness of animated agents is directly affected by the combined elements of physical features of the agent, target audience, domain in which agent is used, and type of task to perform.