Realistic or graphic avatars?
One of the most important books on dissecting sequential art – “Understanding Comics”, by Scott McCloud – contained a chapter that discussed the continuum between realism and abstraction in art. Specifically, it discussed representations of people in comics, with regard to how realistic or not the art chooses to portray them. A photograph would be the most realistic representation, as it is the closest approximation to a real, individual human face, with no abstraction; while a smiley face would be the polar opposite, the most abstract representation of a person.
The way an audience experiences a face on one of these two extremes is that the more realistic an image of a person is, the more the audience will recognize it as a specific person, while the more abstract, the more the audience will project themselves onto the figure.
These principles can also be used in elearning. When choosing avatars or other characters to place in your elearning, consider what you are trying to convey with those characters. If you are using photographic characters, it might be because you want your learners to see the characters as specific people, who have personal stories they are conveying. This is great if you are building an elearning that simulates the learner going through the material with a mentor, or scenarios where they have to interact with other people.
Abstract or iconic characters work as stand-ins for your learner, who can project themselves into the abstract character and picture themselves doing what is being demonstrated. This is important if you wish your learner to pay attention to a process being shown, and not who is doing this process.
A character in between can be used in either situation, and is also useful for representing a type of person or a role your learner needs to understand.