It is fairly clear that there was learning going on when the children observed aggressive behavior, but leaping immediately to the conclusion that what was learned was aggression, and not the specific behaviors exhibited by the adult models and repeated by the children, seems at least a little presumptive. If it can be assumed that the children in the experiment had never witnessed the specific behaviors of the models prior to the experiment (which would have been necessary to establish for the experiment itself to be valid), then the behavior might have been simply frightening to them, and acting out the behaviors might have been a method of familiarizing themselves with the behavior so as to understand it and make it less frightening. Though the end result would be the same desensitization to aggression, the possibility of this mechanism is important. This possibility, and the construction and results of the Bobo experiment as they now stand, also raise some serous ethical concerns. If the fear I theorized above were actually at work in the learning process, then simply subjecting the children to the behavior would create ethical problems by needlessly frightening the children. Of even greater concern, however, were the long-term effects of the Bobo experiment on the children involved in the experiment that witnessed the aggressive behavior of the adult models. Eight months after the time of the experiment, forty percent of these children still exhibited the same type of aggressive behavior that they had...

This suggests that even this single instance of exposure to aggressive behavior was enough to create a long-term pattern of behavior in the children. If the experiment truly caused long-term behavioral differences, especially negative ones, in the children involved, it was completely unethical in the first place.
Bandura himself made it clear that simple observation did not necessarily equal learning, and that even learning did not mean that a child or individual would then exhibit the learned behavior (Van Wagner 2009). His experiment shows, however, that such observation and learning can lead to behavioral changes, and this makes it unethical to engage in any further experiments with subjects that are unable to give informed consent (such as small children, for instance). His experiments also make it clear that the mechanism by which observational or social learning occurs needs to be more fully understood before any broader conclusions about the theory can be drawn. The theory provides an interesting way to look at the workings of the brain and behavior, but the conclusions are far from certain.

References

Isom, M. (1998). "Albert Bandura: The social learning theory." Florida State University. Accessed 29 August 2009. http://www.criminology.fsu.edu/crimtheory/bandura.htm

Van Wagner, K. (2009). "Social learning theory." Accessed 29 August 2009. http://psychology.about.com/od/developmentalpsychology/a/sociallearning.htm