The Surprising Origin of Conversational Cognition
There’s a long going conversation of whether human cognition is due to nature or nurture. I will argue here that Conversational Cognition is an innate capability of human brains and does not require mechanisms of cultural evolution (i.e. nurture).
The human brain likely has innate priors that enable an infant to learn incredibly quickly. It is key to recognize that infants learn skills in roughly the same order as other infants. This suggests that there exists a meta-learning algorithm that is innate in humans.
Chimpanzees also learn in the same order as human infants. Up to the age of 9 months, a chimpanzee has the same cognitive skills as a human infant. Yet after this 9-month threshold, a chimp hits a cognitive wall. This is despite the fact that chimpanzees have larger short term memories and faster reflexes than humans. Humans have slower reflexes and smaller short term memories than chimps, but we have larger brains. Nevertheless, we cannot ignore the reality that chimpanzees also have an innate meta-learning algorithm that is identical to humans.
This innate meta-learning algorithm likely exists to all mammals and this leads to the sophistication found in primates, dogs, dolphins, whales, primates, elephants, and seals. We need to discover this more fundamental algorithm. The human meta-learning algorithm likely bootstraps from this more fundamental mammalian algorithm.
We don’t know if the mammalian algorithm is different from the birds and octopuses. Discovering if in indeed it is the same then this suggests evidence for the independence of this algorithm on neural architecture. We do have a commonality between octopuses, birds and intelligent mammals. We all know that they have REM states in sleep. Interestingly enough animals with more innate priors (i.e. Ibex) appear not to have additional brain plasticity as evidence by having less or non-existent REM states in sleep. REM sleep clearly is highly suggestive of a general meta-learning algorithm across many species.
The argument of nature vs nurture can be framed in terms of the computer science idea of eager vs lazy evaluation. A specific case of this is known as premature optimization. Brain plasticity allows deferred evaluation which is a case of loose coupling in time. Another phrase that is more commonly used is “late binding”. The entire history of computer science can be tracked by later and later binding. (See: Recombinant Programming). Late binding is a kind of Loose Coupling.
Ockham’s Razor which is a popular principle in formulating theories via abductive reasoning can be recognized as a restricted case of the Loose Coupling Principle. Loose Coupling is an information dependency principle and the lower the dependency the more flexible the architecture. Loose coupling enables modularity and modularity enables recombination and thus emergent phenomena.
The nature versus nurture argument boils down to a question of allocation of fixed circuitry versus just-in-time circuitry. It is analogous to computer design where fixed software known as firmware exists on all computers and the boot process (derived from bootstrap) installs new software (i.e. the operating system). However, unlike computers an animal must bootstrap its software solely through a learning algorithm. There exists no knowledge upload mechanism for biological brains. At best, an animal can leverage social (or cultural) learning to further accelerate learning after birth.
The branching point for humans and modern primates was 5 million years ago. Hominid evolution preserved weaker but more dextrous hands and evolved more complex vocal cords. These two intrinsic human body features drove the capability of tool making and the teaching skills required for tool making. Octopuses have bodies and tentacles that are extremely dexterous (despite having 56 times smaller brain than Chimpanzees) are capable of astonishing feats of mental gymnastics. Cognitive development moves forward in progress with the adjacent possibilities afforded by features of the body (i.e. hands and voices). This is because minds learn through interaction with their environments and the richer an agent's ability to interact the richer its cognitive development (see: Embodied Learning). A passive observer is unable to gather sufficient signal from an environment.
The brains of humans have 3 times more neurons as chimpanzees. Evolutionary social pressures pushed the expansion of the human brain. Our brains evolved to become bigger and more complex. All intelligent animals are social by nature. Human brains are just scaled versions of chimpanzee brains. So what innate cognitive tools do humans have that are absent in chimpanzees?
As a consequence of our expressive hands and vocal cords humans have developed the corresponding neural machinery that leads to more effective social coordination. This machinery builds on existing cognitive machinery that exists for all mammals. Mammals need to assess the needs of their offspring as they are being reared. That is, mammals need to learn to effectively communicate with their children and their mates to survive. Just as animals must converse with environments to survive, mammals need to converse and coordinate with their social groups. This is another level of complex coordination in that effective conversation requires a level of empathy. Therefore the capacity for conversational cognition is an innate capability in mammals. The sophistication in conversation that humans have is a consequence of richer communication capabilities and larger brains. However, the core architecture exists in mammals in the limbic system.
How does the limbic system enable conversational cognition? The two key functions of the limbic system reside in the Amygdala and the Hippocampus. The Amygdala is the source of emotion, attention, and trust. Clearly the Amygdala is necessary for empathy. Conversation requires empathy but also the ability to ability to move conversations forward to explore more complex ideas. The Hippocampus is necessary for navigation and spatial memory. All abstract thinking is based on a foundation of spatial cognition. The limbic system contains necessary components for conversation but that does having the necessary components does not imply that it is used in conversation.
I will, however, argue that many mammalian brains are in conversation all the time. Brains are split into hemispheres and depending on the architecture are in conversation with each other. Different cognitive functionality is located in either hemisphere and normal cognitive function requires the conversation between these two hemispheres. Regardless of what function goes to which hemisphere there exists a decoupling of cognitive function in the brain that requires coordination or conversation to reach a consensus.
Humans have two brains that provide contrasting perspectives is how space is navigated. One is an allocentric brain (System 1) and the other is an egocentric brain (System 2). The former processes episodic information and the latter handles procedural information. An enactive cognitive process proceeds by alternating inference of what/where/when (episodic) and how (procedural). There is a flow of cognition where we find an abstracted form of navigation that steps from one abstract episode to another abstract episode. The mind intuits the meaning of an existing state and uses intuition to device the transition to another desired state.
Human cognition continuously exercises this intuitive process and conversation between two minds. It is also this same mechanism that allows the interpretation of conversation. Human conversation is intrinsically ambiguous and thus require intuitive machinery to resolve the ambiguity. Human language lacks the strict precision of mathematics and computer languages. For human language to be of use, humans require the intuitive machine to parse ambiguous sentences. Language however still needs to be generated in a manner that increases the likelihood of the conveyance to be understood. When speaking, a generative process is engaged that builds a sentence using the same machinery as that used for explaining the steps of navigation.
All conversation is about navigation. Humans have conversational cognition because we natively have navigation capable cognitive machinery. We assimilate stories best because stories are presented in terms of navigation. Conversations like navigation our empirical in nature. In conversation, we build a hypothesis of the meaning of an invocation and to refine our understanding we provide feedback in conversation. There is a back and forth of hypothesis making and testing. This is a meta-principle that generalizes to our practice of science:
Conversation is the meta-learning algorithm of human cognition. In a future post, I will detail the specific generative process that leads to the unique analogical thinking found in humans.
Further Reading
