1
Sketch of a Neurocomputational Explanation of
Emotional Consciousness
Sketch of a Neurocomputational Explanation of
Emotional Consciousness
Paul Thagard
University of Waterloo
2
Mechanistic Explanation of Emotional Consciousness
1. Consciousness
2. Explanation
3. Brains
4. GAGE
5. Objections
6. Conclusions
3
Origins of Consciousness
• Creation: God’s gift.
• None: consciousness is mythical, like demons and caloric.
• By-product of evolution of cognitive complexity.
• Evolution by natural selection: increases ability to survive and reproduce.
4
What is the Function of Consciousness?
• Emergency interrupt?
• Improve perception, sensation, inference?
• Improve problem solving?
• Improve teaching of skills?
• All of these seem like minor improvements.
5
Humphrey’s Social Theory
• Humphrey: The function of consciousness is social, improving the ability to understand, predict, and manipulate the behavior of others.
• Implication: Emotional consciousness is central.
6
What are Emotions?
• Cognitive theory: Emotions are appraisals of situations.
• Physiological theory: Emotions are physiological reactions to situations.
• Integration: Emotions are mental (brain) states caused by interplay of physiological reactions and cognitive appraisals.
7
Explanation Targets
• What is it like to be happy? Highly misleading question. Alive.
• Why does someone become happy?
• Why does someone go from being happy to being sad?
• How does happiness affect behavior?
8
How Consciousness Helps
• There are unconscious emotions. • If you become conscious of your emotions, then
you make approximate generalizations:– If <situation> then <emotion>– If <emotion> then <behavior>
• Linguistic representation of emotional states requires conscious awareness of them.
• Explanation target: How do brains become aware of emotional states in order to reason about them?
9
Mechanistic Explanation
• How does a bicycle move?• Parts: frame, wheels, gears, chain, pedals, etc.• Relations: e.g. pedal connected to gear.• Behaviors: e.g. pedal moves when pressed.
10
Brains: Neurons
• Parts of brains: neurons, glia, neural populations, brain areas.
• 100 billion neurons, with thousands of connections.• Main behavior: spike as result of chemical inputs.
11
Brains: Neural Populations• Computational Functions of Neural
Populations (Eliasmith & Anderson, 2003).– Encode information, e.g. perceptual input
encoded by spiking patterns of a population.– Decode information, taking inputs from other
neural populations. – Transform information, changing the internal
representation of information. EXAMPLES?
12
Brains: Areas
• Areas are anatomically identifiable collections of neural populations that are highly interconnected with each other.
• For emotion, some important areas are: amygdala (fear), nucleus accumbens (reward), insula, ventromedial and dorsolateral prefrontal cortex.
13
Neural Mechanism
• GAGE model: Wagar & Thagard,
Psychological Review, 2004.
VMPFC
NAc
HC
Amg
VTA
Somatic state
To Action/Overt
14
15
Key Brain Areas
• Prefrontal cortex: responsible for reasoning.• Ventromedial PFC: connects input from
sensory cortices with amygdala etc.• Amygdala: processes emotional signals,
especially fear. Somatic input.• Nucleus accumbens: processes emotional
signals, especially reward.• Hippocampus: crucial for memory formation.
16
How GAGE Explains Phineas
• Damasio: Effective decision making depends on integration of cognitive information with somatic markers.
• Damage to VMPFC prevents this integration.
• GAGE shows a plausible mechanism for integration that is disrupted by VMPFC damage.
17
GAGE II - under development
1. Incorporate additional brain areas: insula, anterior cingulate cortex, dorsolateral prefrontal cortex.
2. Incorporate higher level representations of relational information, to describe situation-emotion-behavior connections.
18
Components of Emotional Consciousness
1. Spiking neurons are organized into neural populations.
2. Some neural populations encode perceptual and somatic inputs.
3. Some neural populations decode, encode, and transform inputs from (2) plus cognitive inputs.
4. Feedback loops are common.
19
Higher Order Representations
Hypothesis 1: There are neural populations, possibly distributed across brain areas, that encode emotions.
Hypothesis 2: There are neural populations that encode generalizations of the form
<situation> <emotion><emotion> <behavior>
20
Agenda
• Design system of brain areas that conducts neural transformations of transformations of sensory, somatosensory, and memory inputs.
• Apply this system to explaining emotional phenomena.
21
Explanation Targets
• Onset and end of positive and negative emotions.
• Increase and decrease in intensity.
Intense
Weak
+-
grief
sadness
elation
happiness
anger
anxiousamused
bored
joy
22
Mechanisms
• Onset of positive emotions results from perceptual or memory input that activates reward areas.
• Intensity is a function of degree of cognitive evaluation and physiological inputs.
23
Scientific Objections
• Need for more detail about how the encodings work.
• Need application to specific aspects of emotional consciousness.
24
Philosophical Objections
• Zombies: We can imagine creatures just like us but lacking emotional consciousness. Response: imagination is a poor guide to reality.
• What Mary knows: Mary (without emotional consciousness) could know everything about the neuroscience of happiness, but not know what happiness is. Response: Mary would never have made it through kindergarten.
25
Philosophical Objections
• The mechanistic theory of emotional consciousness doesn’t tell us what it is like to be emotional.
• Response: it also doesn’t tell us how many hours there are in a kilogram.
• Better response: it should be able to explain why we feel positive/negative, weak/intense,
26
Conclusion
Mechanistic explanations of emotional consciousness are feasible.
They will require further understanding of the functions of different brain areas.
27
Web sites
• http://cogsci.uwaterloo.ca/• http://faculty.washington.edu/chudler/neurok.html• http://www.thebrain.mcgill.ca/flash/index_i.html