The proposal should identify some open problem in the area, some novel behaviors to be demonstrated and studied, or some important phenomena to be explained. In addition, it should present a computational approach to addressing these problems that includes one or more proposed representations of learned knowledge, performance elements that use that knowledge, and learning methods that acquire this content. You should also describe plans for evaluating the proposed methods, whether this takes the form of experimental studies, comparisons to psychological data, or theoretical analyses. Finally, the text should discuss how the proposed approach differs from, and moves beyond, previous work in the area.
You need not present things in this order. For instance, you may want to review related research first to motivate the need for additional work. A good research proposal is very much like a good research paper, except that it has few results and spends more space on plans for future research. As in a research paper, one or more convincing examples are crucial. The proposal should convince readers that you understand the field, have good ideas about still needs to be done, and have plans for how to implement and evaluate those ideas.
For more information, send electronic mail to firstname.lastname@example.org