Consciousness & Emotionality integration - under review- do not cite

Cognition means 'knowing'. So how do we know anything? Who or what is it that is doing the 'knowing'? Most conceptions of mental computation have failed to answer these questions satisfactorily, and for rather obvious reasons- they don't even try. Computational frameworks which make 'honest' attempts to answer this question must include consciousness and emotionality. As with consciousness, emotions have traditionally been analysed using two differing points of view [13]-

(i) external/communicative- as observed behaviours, in which communication among individuals is considered to be helped by cues provided by emotion-based attitudes

(ii) internal/cognitive-as embodied mental states, cognitive functions crucial to the understanding of decision making processes.

However, if the declarative ('what') computation paradigm is deliberately used instead of the automatic use of the procedural/imperative ('how' paradigm (eg Turing Machines and their ilk), then the task becomes achievable. GOLEM is such a model, since it is goal-oriented, a sub-type of declarative. Instead of (unconscious) algorithms at its heart, goal-oriented computation has high-level (volitional) targets generated by low-level drive-state reduction. These targets are semantic state representations of exemplars. This situation is depicted in figure 8.1 below.

Figure 9.1

GOLEM places consciousness and emotionality centrestage, since it is an example of the analytic approach to AI, where the structure and function of real animal and human brains [10] is the target of the emulation. This is fundamentally different from the conventional synthetic approach, which try to use mathematical tools and software techniques to simulate intelligent behaviour.

Most modern approaches [1] to consciousness such as GOLEM theory assume that self-centered higher ordered representations (HOR's) are essential for conscious experiences. This includes conscious experience of emotional inputs, as well as those of more conventional perceptual inputs. GOLEM theory claims that conscious experience of self (eg internal emotions and external proprioception) is constructed from hierarchical data representations in a similar manner as our conscious non-self /world experiences. The theory consists of external conscious representations  associated with internal emotional valuations of them. Those past experiences with the greatest emotional 'value' are the easiest to recall. Similarly, of all possible future situations, the ones you believe will happen are those with the greatest emotional 'weighting'. Belief and recall are therefore conscious experiences of unconscious emotional evaluations. 

Consciousness and emotionality (volition) are included axiomatically in the structural derivation of the basic animal GOLEM (see section 2), and so play an implicit part in all subsequently derived models, such as the human GOLEM. TDE/GOLEM Theory (TGT) also defines consciousness and emotion in an abstract form, as follows:-

(i) Consciousness is that integrative (ie multi-level) feedforward mental function that manages the interaction between our thoughts & behaviour and spatial structures, consciousness tells us what, how, where. 
(ii) emotionality is that integrative (ie multi-level) feedback mental function that manages the interaction between our thoughts & behaviour and temporal processes, emotions tells us why, which, when.
(iii) Consciousness originally evolved from the need for the brain to locate (on the body) the cause of pain signals, ie find injury (aversive origin) - analyse structure
(iv) Emotions evolved from the need for the brain to locate (in the world) the solution to hunger drives, ie find food (appetitive origin) - synthesize process

Consciousness is directly involved with each animal's sensory and perceptual organs since they deal with space, shape and distance. However, the organ that best expresses the emotional function is, from the definition above, the one which is the best at measuring the effects of time- ie memory. 

It is therefore to be expected that the limbic system be functionally connected to memory, since it is the the limbic system that manages emotions. In the GOLEM model, the limbic system is modelled by the L-lobe, the place where each GOLEM's goal-oriented functions are based.

Some of these complex relationships are depicted graphically below. The left-hand diagram depicts the 3-level human GOLEM, with the addition of the neuroendocrine (zeroth) level,  representing instinctive processes, ie those which don't involve any learning or adaptation and so has not thus far been included in the 3-level human GOLEM. The right-hand diagram is a neocybernetic schematic, clearly indicating the complementary roles of
(i) consciousness  as a multi-level integrative feedforward representation
(ii) emotionality as multi-level integrative feedback evaluation 

(a)                                                                 (b)

Figure 9.2

amygdala vs hippocampus

According to published [2] correlations between clinical experiences and medical imaging, "..while the amygdala is widely regarded as the seat of human emotion, the hippocampus is well recognised as being the storehouse of memories. It is conceivable that all human experience activates both structures, with a feedback loop between the two providing a reference point to the memory and emotions surrounding past experience".

According to another source [3]-“The amygdala combines many different sensory inputs. Like the hippocampus it combines external and internal stimuli. Every sensory modality has input. These are integrated with somatosensory and visceral inputs-this is where you get your "gut reaction". The link between prefrontal cortex, septal area, hypothalamus, and amygdala likely gives us our gut feelings, those subjective feelings, about what is good and what is bad“.

The following example of their comparative effects on learning is taken from an undergraduate medicine course [11]. Two groups of patients were studied. The first group 'H' had hippocampus lesions, while the second group 'A' had amygdala lesions. Both groups were subjected to the same classical conditioning (CC) stimuli, which consisted of a series of coloured slides. Some of the colours were followed by a very loud, sudden, startling sound.  Amygdala patients (group 'A') did not become conditioned, since they showed little or no EDA [12] change when they were later shown any of the colours which preceded the startling stimuli. However, the formed explicit memories of both the colour sequence and which of the colours were followed by sound. Hippocampus patients (group 'H') had EDA changes which demonstrated their successful conditioning (ie they formed implicit memories), but had little or no recall of the colours and their pairing with sounds.

The 4-level human GOLEM model is simultaneously complex and abstract. Is there any empirical support for it? Consider the role of emotions in conscious experience. They are a 'felt' (embodied), integrative (both high and low level) measure of the temporal value of, say, a situation (how do I feel about my job, is it right for me?), or a person (I always seem to drink too much around them, maybe they are not good for me). That is, a subject S's emotions about experience X are based on the comparison between the actual outcome (success or failure) of X and S's expected outcome. 

Clearly, the mental faculty most relevant to emotions is memory. Now lets assume that emotions about (ie S's assessment of) some experience X are 'computed' by the limbic system. Observations of Blindsight sufferers tell us that each conscious experience X (typically vision) of a situation has both 'what' and a 'where' information streams [4]. Therefore the limbic system must have one channel dedicated to 'what' information, and one channel specialised in managing 'where' information. GOLEM theory claims that the amygdala is the specialised portal between the limbic lobe and semantic ('what') Temporal lobe memory codes, while the hippocampus is the specialised portal between the limbic lobe and syntactic ('where') Frontal lobe memory codes. 

H.M.'s anterograde amnesia

As we saw in the previous section 6, the memory of (ie the semantic state, or perceptual property set associated with) an experience X is originally created by the movable 'where' memory matrix row, which actively forms a new short-term memory by its intersection with the constituents of the 'what' semantic state (memory matrix column), in the form of the 'spotlight of attention'. Following memory consolidation that occurs during the next sleep episode, these STM's become LTM's. Clearly, damage to the mechanism that moves the 'where' row will prevent the formation of new memories [5]. 

This prediction of GOLEM theory is supported by the example of patient H.M. whose left hippocampus and some of his Temporal [6] lobe was surgically removed to eliminate all possible sources of a potentially fatal (Temporal lobe epilepsy) seizure. After surgery, H.M. was unable to form any new long-term memories at all- ie he had no anterograde (forward-looking) memory capability.  

 remapping of 'place' cells in lab rats

Laboratory rats who learn to navigate a submerged maze (they use milk so the rats can't see where to walk) fail the task after their hippocampal 'place [7] cells' are ablated. Place fields in rat brains exhibit the phenomena of  partial and total 'remapping', the ability to suddenly change their firing pattern from one pattern to another, due to a significant change in the perceptual characteristics of the environment. This activity closely resembles the 'unlatching' of 'latched' STM neurons by the ARAS, as a GOLEM changes from the 'previous' to the 'next' situation, thereby supporting this part of the theory.

Post Traumatic Stress Delusions (sic)

When people (eg war veterans) have experiences that involve repeated periods of extreme emotions esp. fear, they often suffer bouts of PTSD, involving 'flashbacks' at a later time. It has been suggested that sometimes false STM traces can be formed from highly charged percepts in the 'what' channel alone, without involving the subject's attention, ie without the voluntary participation of the 'where' channel. Unfortunately, in these cases, sleep-dependent memory consolidation only serves to preserve these intrusive, frightening and unwanted delusions.

As crazy as Eye Movement Desensitization and Reprocessing (EMDR) sounds, it has achieved clinical results as good as, if not slightly better than, those of traditional CBT in the treatment of PTSD. Begrudgingly, most international scientific psychiatric institutions have now given it their official imprimatur. The question remains - how does it work? The 'new-age' explanations given by its inventor are not helpful [8]. Maybe GOLEM theory can come to the rescue.

Marcel A. van den Hout & Iris M. Engelhard also asked the question-How does EMDR work? They published their findings in the Journal of Experimental Psychopathology JEP Volume 3 (2012), Issue 5, 724-738. They found that both vividness and emotionality of traumatic recollection decreased when forced to share working memory (WM = STM) with motor repetition tasks, such as lateral eye movements. The nature of the task was not as important as its ability to fully occupy STM. Visual tasks typically contain more information and so are the best ones to use. Mindfulness training is equally effective, because although it involves latent, rather than manifest visualisation targets, they are thought to occupy WM/STM in a similarly effective manner. 

The equal effectiveness of both external visual activity and internal mental imagery represents direct support for GOLEM theory. GOLEM theory (GT) is rather unusual in that manifest visual percepts and latent visual images are stored at the same places in the same memory matrix. The famous experiments by Shepard & Metzler [9] clearly demonstrate that real and imagined versions of any given image exhibit identical properties, eg the mental version rotated at the same speed as the real one.

Manifest visual activity such as EMDR involves the repetitive application of the 'spotlight of attention' to those 'what' columns corresponding to active percepts in the laboratory situation. According to GT, the brain responds to the excitatory effects of attention by 'latching' the neurons concerned. Presumably, this would apply to both the meaningless repeated tasks (eye movements or mindfulness exercises alike), as well as recalling the traumatic event ie re-invoking the PTSD patient's troublesome mental images in WM/STM.

As stated in previous section 6, STM 'latching' of neurons is relatively expensive of brain resources, second in cost only to attention itself. The substitution of STM latching with inhibitory (presumably EP) placeholders each time a new situation is encountered is clear evidence of this high cost. By deliberately 'taxing' WM/STM, the limited nature of the resource may cause the latching process to fail, presumably causing a similar failure in the ongoing memory consolidation of traumatic images. 

1. Lambie, J.A. & Marcel, A.J. (2002 ) Consciousness and the Varieties of Emotion Experience: A Theoretical Framework.  Psychological Review, Vol. 109, No. 2, 219-259 

2. https://nba.uth.tmc.edu/neuroscience/m/s4/chapter06.html

3. Krishnamoorthy, E.S. (2007) A differential role for the hippocampus and amygdala in neuropsychiatric disorders. J Neurol Neurosurg Psychiatry 2007;78:1165.

4. This viewpoint has been subsequently confirmed in laboratory animals and normally sighted people

5. since there are many thousands of 'what' columns, and only one 'where' row at any moment, it is clear that we needn't consider the issue of damage to the 'what' apparatus, since that would involve total destruction of memory.

6. The lower-case 't' spelling of 'temporal' means 'pertaining to time', while using an upper-case 'T' means 'pertaining to the temples'. This particular example of confusing terminology must surely rank as the most unfortunate in all science.

7. perhaps they should be renamed 'pose' cells since they are activated when a rat was "situated in a particular part of the testing platform facing in a particular direction". Pose is a dyad of position and orientation invented by roboticists.

8. Shapiro, F. (1995). Eye Movement Desensitization and Reprocessing: Basic principles, protocols and procedures. New York: The Guilford Press

9. Shepard, R. N. & Metzler J. (1971) Mental rotation of three-dimensional objects. Science, 171 , p 701-703.

10. one of the advantages of the analytic technique is (like real brains) its ability to sidestep the 'frame problem'

11. nba.uth.tmc.edu - University of Texas at Houston, McGovern Medical School

12. EDA = Electrical Dermal Activity.  GSR = Galvanic Skin Response is an older, equivalent term. Our skin surface becomes moist when we become anxious, due to the response of the Sympathetic Nervous System (SNS).  Its electrical resistance changes in a repeatable manner with changes in autonomic state. This is the basis of many devices, such as polygraphs, which (controversially) claim to be a 'window into the subconscious'. EDA responds independently to the SNS, while Heart Rate (HR) responds independently to the complementary Parasympathetic Nervous System (PNS).

13. Ventura, R. Pinto-Ferreira, C. (1998) Artificial Emotions - Good Bye Mr. Spock!