Agent Experimentation

To understand why the agent works and under what conditions, I will run experiments with the agent under different internal conditions. For example, in an ablation experiment, a function of representation is removed from the program. The agent's behavior is then observed to help understand the effect of that missing piece. Exactly what experiments to run will be clearer as I develop the theory and agent.

Through experimentation I plan to evaluate these more specific hypotheses:

1. Visual instantiation will prove to be useful for two kinds of symbolic mismatches that can occur in non-visual representations.

2. Visual instantiations can provide an intermediate representation through which otherwise dissimilar entities and manipulations can be aligned.

3. Visual representations alone cannot enable evaluation of the solution.