This process is oriented to existing implemented agents and provides an accurate and compelling measure of the level of cognitive development. However, it takes much more time and effort to be performed.
First of all, we need the actual agent and a particular problem domain definition for testing. As mentioned above, rating is based on architectural components and cognitive skills. Architectural components of the agent are identified through internal inspection of the implementation. Cognitive skills present in the agent are assessed thanks to the definition and execution of specific cognitive tests adapted to the established problem domain. Once we have the list of architectural components and cognitive skills found in the agent, we can apply ConsScale metrics to obtain the nominal level of consciousness, the cognitive graphical profile and the CQS score (you can use the Online ConsScale Calculator to do this).
Note that comprehensive cognitive test/s have to be devised for each cognitive skill. These tests have to be designed in such a way that they validate the integrative and developmental inspiration of the scale. In other words, whenever possible, higher level cognitive tests will require the presence of all (or most) lower cognitive abilities in order to be passed.
See this paper for an example of the ConsScale Standard Evaluation Process in the domain of video game bots.
ConsScale - ConsScale Rating