ConsScale Abstract Architecture Definition

Using ConsScale, the level of development of an agent is also assessed in terms of its architectural components (i.e., ConsScale doesn’t focus exclusively on cognitive skills). The scale is based on both external observations (behavioral profiles) and internal observations (agent’s internal architecture). This prevents pre-programmed behaviors, which could be intended to fool the scale, to distort the final measure.

The following table summarizes the defined abstract components:

In order to characterize the architectural design of an agent in terms of the development of consciousness we need to formally define the basic components of an artificial situated agent. Such an agent interacts with the environment by retrieving information both from its own body and from its surroundings, processing it, and acting accordingly. We have identified a set of essential (abstract) architectural modules: sensors, sensorimotor coordination, internal state (including memory), and effectors (see the table above for a complete list). These modules implement the basis of the following processes: perception, reason, and action. Cognition and learning can develop in an agent as a result of the combination of the former processes during the interaction with the external world and their own inner state.

The main abstract architectural components are:

• Body (B). Embodiment is a key feature of a situated agent. Agent’s body can be physical or software simulated (as well as its environment). A boundary is established between agent’s body and its environment (E). The rest of components are usually located within this boundary.
   
• Sensory Machinery (S). Agent’s sensors are in charge of retrieving information from the environment (exteroceptive sensors or Sext) or from the agent’s own body (proprioceptive sensors or Sproprio).
   
• Action Machinery (A). In order to interact with the environment the agent uses its effectors. Agent’s behaviour is composed of the actions ultimately performed by this machinery.
   
• Sensorimotor Coordination Machinery (R). From purely reactive agents to deliberative ones, the sensorimotor coordination module is in charge of producing a concrete behaviour as a function of both external stimuli and internal agent’s state.
   
• Memory (M). Internal agent’s state is represented both by its own structure and stored information. Memory is the mean to store both perceived information and new generated knowledge. Even agents that do not maintain explicit state can be said to have a minimal state represented by its own structure, i.e. preprogrammed sensorimotor coordination rules.

  

 The former components refer to an abstract architecture; therefore, we are not considering here any particular agent implementation or concrete sensorimotor machinery. Using the presented abstract architecture allows us to define consciousness levels independently of particular implementations. Different classes of agents could be obtained depending on the concrete implementation of the abstract architecture. It is also important to note that no specific component of this architecture is responsible for the production of consciousness. Instead, we support the argument that consciousness could emerge from the interaction of the specialized processes present in the agent.

ConsScale - ConsScale Levels