Humans are Empathy Machines
“We Are Not Thinking Machines. We Are Feeling Machines That Think.” — Antonio Damasio
Anil Seth calls us a “Beast Machine”. Along the same lines, I’ve argued that we are “Intuition Machines”. The argument goes that biological brains gain an understanding of their world through the embodiment in its environment. Understanding can be understood as fitness from the cognitive perspective. Fitness is a word that is used in the context of an evolutionary process. An evolutionary process is a learning process. A learning process is a knowledge discovery process. Knowledge is information with a useful purpose. Information with a useful purpose are known as affordances. So to tie everything in this chain, understanding is the learning of affordances. Understanding can’t be divorced from learning and I’ve argued that humans learn using an intuition or amortized inference process.
That said, one might ask, what is information with a useful purpose (i.e affordance)? A useful purpose or affordance is relative to the agent that discovers the information. We then break down this agent using Anil Seth’s model of selves.
Different selves are informed by information with different purposes. I’ll use the more awkward phrase ‘useful purpose’ rather than affordance because it seems many have difficulty with the latter word. To illustrate this, let’s walk up the list of self-models. The bodily self requires information that sustains its biological needs. The volitional self requires information that enables or constrains its agency. The perspectival self requires information that confirms or surprises internal mental models. The narrative self requires information that abstracts information into manageable information chunks. The social self discovers information that motivates shared experiences.
At the top of the hierarchy of selves that defines us, there exists a kind of capability that we can best describe as empathy. One might come to the wrong conclusion that the word empathy would have very ancient origins. It turns out though that it was invented just a century ago. The English word “empathy” was suggested in 1908 by a pair of psychologists. They based its meaning from the German word Einfühlung and the Greek “em” for “in” and “pathos” for “feeling”. Although its original meaning appears to be that cognitive capability that humans have when they extend their bodily selves to the tools that they use. Sympathy, by contrast, originated in the 16th century, it’s common use today is to express a shared experience of pain rather than empathy which is a shared experience of cognition.
It was only by the mid-20th century that the more common use of ‘empathy’ was established. That instead of the original meaning of projecting a body’s senses to objects, one would instead have empathy by placing oneself in the shoes of another. The biological origins of this capability were only discovered in the 1990s when mirror neurons were discovered. Empirical evidence shows that humans are best at predicting other people’s behavior when that behavior most closely matches their own. That is why we are more comfortable with people who are cognitively like us. This explains “Birds of the same feather flock together.”
A common misinterpretation of empathy is that it involves only the employment of emotional cognition and therefore is distinct from rational thought. This is analogous with bias against intuitive thought, that it appears to be based on cognition other than rational thought. This rational delusion or Cartesian dualist thinking is the main culprit as to why the GOFAI approach towards general intelligence has failed. One cannot have general intelligence without the capability of learning. Deep learning has shown how to build machines that learn and it is the correct first step towards more capable AI.
So we must use the word ‘empathy’ in a more universal sense. That is, empathy involves all our self-models (i.e. bodily, volitional, perspectival, narrative and social). I should not need to invent a new word because some people wish to equate empathy only to a restrictive sense of the word. All of human cognition involves pattern matching of observations of the world with one’s own internal self-model. We understand by discovering the equivalence of concepts with our own internal concepts. Category Theory (a branch of abstract mathematics) provides us with a universal notion of equivalence. Two concepts are equivalent when you can discover transformations back and forth between the two.
Empathy, therefore, is the human capability of discovering the equivalence of one’s internal self-model with that of its observations of the world. Empathy is how we make sense of the world. Conventionally, we employ the word ‘understanding’ to this capability. The origin of the word ‘understanding’ comes from the Latin prehendere, to grasp. Understanding means to grasp an idea, to hold on to an idea. The problem with this metaphor is that it fails to express the notion that understanding of any new idea is possible only when it relates to one’s own internal mental models. One cannot grasp a new idea if one has not developed the cognitive roots to hang that idea on to that mental tree. That is why you cannot achieve ‘symbol grounding’ in a top-down manner.
So when we understand ideas, we don’t grab and take ideas and separate them from their original sources. It is more like we empathize with what is external. What this mean is that we don’t internalize everything that is external. We only internalize the bits and pieces that we can attach to our existing cognitive models. The remainder remains in external form. So when we look out into the world with our eyes, we don’t internal everything we see. We only internalize what is important to our self-models, the rest we discard. We discard the rest because there are limits to our cognition and that our cognition is very frugal. The key to understanding the uniqueness of human cognition is understanding how we recognize what is relevant and what is not.
David Deutsch in his book “The Beginning of Infinity” asks what are the evolutionary pressures that drove the need for human ingenuity (Deutsch uses the world creativity, I prefer ingenuity)? Deutsch argues that little innovation was achieved by homo sapiens for 10,000 generations. It is only in the last 200 generations that technology has accelerated exponentially. If human ingenuity were indeed an intrinsic characteristic then humans we should expect a more rapid rise in innovation. How does individual ingenuity lead to an increase in one’s likelihood of survival in societies with very little innovation? In static societies with marginal innovation, survival will depend on one’s status in society. But how does one achieve higher status by having ingenuity?
Deutsch explanation is that ingenuity is critical in understanding society’s value memes (v-memes in Spiral Dynamics). Humans employ their ingenuity to discovery the explanations for the complex value memes that exist in society. Those that survive best are those that best understand the point of value-memes. Those that survive societies are those that are best able to conform to and perhaps even game the rules. It took over 9,000 generations of homo sapiens and their corresponding societies to favor individuals with creativity and as a consequence promote selection for social empathy. Deutsch writes that ingenuity’s original function was that of preserving cultural memes faithfully and how it has been reassigned to the function of creating new knowledge. Ingenuity is our ability to innovate and as an effect, it enhances our ability for empathy.
Further Reading
