It pays attention to what's important to it

The memory patterns representing the past unhappy "hurling ball tears skin" encounter is not stored in a neutral way. What is stored is not just the internal representations of the ball, of visual memory of the ball hurling at the machine, and of the skin sensors' reports, but also the fact that the tearing reported by the sensors was Pain, and that this was Bad, plus the machine's reactions of withdrawal at that time. All these representations are usually bound together. The Pain and Bad representations give the memory patterns a high emotional significance value.

This high emotional significance, when the above memory patterns are evoked by the pattern of the yellow ball, leads to the machine's brain giving high attention to them. What this means is that these patterns stay around for longer, the voltages on the signal arrays' signal lines stay high, while other signal arrays (such as the "purple ball rolling up" imagination) die out.

Due to its associative nature, just about every signal pattern in the brain will evoke every other, and the result is a huge collection of signal arrays being activated. If the machine does not select a subset to pay attention to, it will probably day-dream all day. Emotional significance is a factor modulating attention.

An aside about the "everything evokes everything else" above: If each possible representation (a particular knowledge, imagination, motion, etc.) were thought of as a point, and we draw lines connecting points that relate to one another, then this is a "social network graph". The human social network graph is said to have about 6 degrees of separation. This graph is likely to be much denser, with far fewer degrees of separation. Just about anything will quickly evoke everything else.

It introspects

So far, the inputs to the perception process are from sensors that sense external information (e.g. camera) or internal (e.g. joint). But what if the various signal arrays floating around in the brain (as they are output by perception processes, or evoked from memory, or made up by the imagination process) get connected to the input side of perception circuits? Nothing stops this, and Haikonen's architecture mandates it.

The result is that the machine's perception neuron groups not only help it to perceive the outside world or the body's internal hardware, and help initiate a flow of mental images as discussed above, but these circuits also let it perceive this very flow. This is introspection, this is the machine examining and perceiving its own mental flow. Unlike a cow or a traditional computer, the machine perceives what goes on in its brain.

Is the Haikonen machine conscious?

The Haikonen machine's claim to consciousness starts with its ability to perceive the outside world and its body. Each of these percepts is grounded to the world in its meaning.

More importantly, it has a flow of imagery in its electronic brain. Perhaps most importantly, it perceives that flow by looping images in the flow to the circuits involved in perception.

What if the machine can speak, and tell us of that mental flow and its perception of that?

Haikonen does have a theory for the machine to learn to understand natural language (the learning of the word "ball", mentioned above, is an example). This theory also enables the machine to form elementary sentences such as "ball round" or "ball not purple". We do not know if this is a good enough foundation for the machine to grow up - cognitively speaking - then one day say, on its own, "There is a flow of images inside, and I can perceive it". But then neither can preschoolers say such a sophisticated sentence.

Outwardly, the Haikonen machine is able to behave in some of the ways which humans interpret as emotional or motivated (e.g. withdrawing from sources of pain, cameras' gaze turn to and focus on unexpected events). If we look inside, we can see its motors tense up in response to fear, we can probe to see that its brain is introspecting, and we can recognise mental patterns as the emotions of fear, curiosity, surprise, etc. But if we can only watch its external behaviours, will we dismiss it as just a non-conscious computer carrying out its programmer's instructions? Perhaps we will be less dismissive if it rushes over, grabs our phone charger, and tries to charge its hungry battery.