Principia Cybernetica Web

What is consciousness?

In the evolutionary cybernetic approach, consciousness is not some kind of mysterious entity that evades all forms of scientific analysis. Consciousness is rather a subtle and complex form of organization characterizing cybernetic systems or "agents". Consciousness allows these systems to interact in a purposeful, intelligent and sensitive way with their complex and changing environment. Because consciousness in its every-day sense is such a vague and ambiguous concept, we will try to explain it by subdividing it in its different aspects, starting from its most simple and universal properties, that are shared by all cybernetic agents, and building up towards its most advanced features, that can as yet only be found in humans, who occupy the provisionally highest level of the metasystem hierarchy.

A cybernetic agent is defined as a control system that interacts with its environment in such a way as to maximally achieve its goals or values. For natural systems, these goals are all derived from the overriding goal of maximizing fitness, i.e. survival, growth and reproduction. A cybernetic agent steers towards it goal by executing the appropriate actions, taking into account the feedback it gets through its senses.


The first, and most primitive, form of consciousness may be called sensation. To achieve its goals, or more specifically to survive, a cybernetic agent must be be able to perceive or sense its situation. The situation is determined by the state of environment and of the agent with respect to the agent's goals, and in particular by the deviation between the present state and the desired or goal state. Sensation is achieved through the system's sensors, which translate phenomena in the environment into internal information that makes sense with respect to the system's goals. A sensed deviation automatically triggers a corresponding action that would compensate for the deviation, in what may be called a "simple reflex".

The most rudimentary example of such a system is the thermostat, which senses the temperature in a room, determines the difference between the sensed temperature and the desired temperature, and activates the heating element if the difference is too great. The capabilities of sensation of a thermostat are the simplest one imaginable: it can sense only one variable, temperature, and only two values for that variable (i.e. one bit), "temperature too low" and "temperature high enough". More complicated systems may be able to sense many variables independently (e.g. temperature and humidity) and many different values for each variable, each triggering an appropriate action. Still, we would not yet say that such sensing agents are "aware" of their environment.


At the next level of complexity, which we call "complex reflexes" in the theory of metasystem transitions, separate sensations do not automatically lead to separate actions. Different sensations are rather integrated into an overall representation of the situation, which is compared with an overall representation of the system's goals. Different sensations and goals interact inside the cybernetic agent's "nervous system", affecting the internal, "mental" state of the system. The action that the agent eventually takes is determined by its internal state, which is the result of all previous sensations and goals, and the present perception. There is no longer an immediate connection between sensation and action. Rather the agent is affected by the whole of all previous and present sensations. Therefore, we may interpret the agent's mental state as not only as embodying not only a sensation of the present situation, but a global feeling or awareness determined by goals, past and present sensations.


At the next metasystem level, which we call "learning" or "associating", the agent's decision about which action to take is no longer determined directly by its state of awareness. The decision-making mechanism will now adapt or change, because the agent will learn from its experience, thus becoming ever more effective in its actions. As a result, at different times it may react differently to the same sequence of sensations. Now not only the mental state of the agent is affected by its sensations, but also the structure of the mental system with which it interprets the sensations. This also means that initially identical agents that undergo different sequences of sensations will start to react differently. Because of their individual experience they will develop their own personality, character, or world view. As a result the same phenomenon will be experienced in different, unique ways by different agents, having a different meaning for each of them.

Self-awareness, reflection

Learning agents are still dependent on the environment to create new associations between sensations. At the next level of "thinking", agents become capable of creating their associations themselves, thanks to their capacity to symbolize experiences, and combine symbols into novel combinations that have never been experienced as such. At this level, the agent becomes aware of its own experiences, so that it can examine, analyse, integrate and manipulate these experiences. The agent also becomes aware of itself as an agent, similar to, but different from, other agents. It becomes capable of reflection or introspection, observing its own cognitive processes as if they were external to it. This allows the agent to be creative, to imagine situations and ways of achieving them, without ever having experienced them directly. It also allows the agent to improve its own mental functioning, to become more "conscious" of itself and the world.

First-person experience

The above sequence of levels of consciousness, from sensation, to awareness, experience, and self-awareness, in our view captures all the essential properties of consciousness. In principle, cybernetic agents that exhibit all these levels could be designed and built by engineers, e.g. in the form of some complicated neural network with sensors and effectors, that can learn from experience, and that uses symbols to represent learned conceptual associations. Such a neural network could be used to steer a robot. If the robot's goals, sensations and actions are chosen to be similar to those of a person, that robot would behave in a way not essentially different from a human being.

Yet, many consciousness theorists would claim that such a robot would still not be conscious, because it would lack what they call "first person experience" or "qualia". This so-called "hard problem of consciousness" vanishes if it is considered from a cybernetic point of view, according to which the property of consciousness is determined by the robot's organization, not by some mysterious substance, fluid or force.

Copyright© 2000 Principia Cybernetica - Referencing this page

F. Heylighen,

Apr 12, 2000


Metasystem Transition Theory


Prev. Next

Is there a "hard problem" of consciousness?


Add comment...