From Monkey Neurons to the Meta-Brain

What can one neuron tell us about brain function?  It can tell if we are looking at a picture of Jennifer Aniston. Brain surgeon and researcher Itzak Fried, in 2005, was probing a certain brain region in patients with epilepsy to pinpoint the source of their seizures.  This is open brain surgery done while the patient is conscious (the brain doesn’t have pain receptors).  These patients agreed to additional probing in the interest of science.  Fried was showing patients pictures, some of famous people, and kept running into neurons that would fire to multiple representations of the same person or object, and to nothing else (within the limited but large set of images used).  “The first time we saw a neuron firing to seven different pictures of Jennifer Aniston–and nothing else–we literally jumped out of our chairs,” recalled R. Quian Quiroga, who did subsequent work on the phenomenon with Fried.

In a study by Quiroga, Fried and others, severe epilepsy patients each had 64 tiny probes implanted in different parts of the brain, to study how the seizures manifested. The patients also agreed to view sets of images while the probes were monitored. A number of invariant responses (the same neuron firing to multiple views of the same person/thing) were found.  “In some patients, Jennifer Aniston neurons would also fire to her fellow actresses in Friends, … But they would never fire to other similar-looking, but otherwise unconnected, actresses” (Nature Magazine).  Either way, a connection was made between a concept and a single neuron.  Finding connections between a specific neuron and one specific memory has been going on for seven decades, and single neuron stimulation has triggered laughter, remembered childhood scenes or hearing snippets of music, but this association, apparently with the concept of a certain person, instantly became and remains a major focus of brain research.

Monkey See, Monkey Do?

About 10 year earlier, Italian neuroscientists were studying macaque monkeys with their brains exposed to probing, trying to map body movements to particular motor neurons.  They found a neuron that fired whenever the monkey grasped something.  But then the same neuron fired at other times, remarkably, when the monkey saw one of the lab workers grasp something.  So the monkey watches someone grasp something, and has an inner experience of grasping it himself?  And could imagining a hand movement possibly cause motor neurons to fire as if one were actually performing it?  The short answer is yes.  We know this occurs in dreaming, when studies show our brains generating a lot of starting sequences for action, that are somehow inhibited, unless, that is, a person has a disorder connected with excessive thrashing in bed and even sleepwalking.  The other startling thing is for monkeys to be so in touch with another being who happens to have hands, as to spontaneously imagine doing what the other is doing, and in such a neurologically realistic way.  We might call this phenomenon motoric empathy.

This finding caused a certain amount of giddiness, and visions of a neurological basis for empathy in all its generality.  Maybe so, but it’s a very long track or trek, with many skeptics as well as cheerleaders along the way.  The finding was replicated in many variations, and analogues were found in human beings.

Neurons that fire both when you are doing something, and when the same action is observed, were dubbed “mirror neurons”. It sounded, and still sounds, in some accounts, like mirror neurons were a particular unique kind of neuron whose function is to cause or indicate empathy, but as with the “Jennifer Aniston neurons”, it’s more likely that just some neuron in or connected to a brain circuit for producing grasping movements was encountered.

Mirror neurons are also used to support the “Simulation Theory” of mindreading in Alvin Goldman’s Simulating Minds: The Philosophy, Psychology, and Neuroscience of Mindreading (2008) .  By “mindreading”, also called “theory of mind”, philosophers and cognitive scientists mean our ordinary ability, such as it is, to intuit what other people are thinking; simple things like if Joe and Sally see a ball put in Box A, and while Joe is out of the room, it gets moved to Box B, and Joe reenters the room, and Sally is asked where Joe thinks the ball is, she will say Box A.  Testing whether Sally will intuit Joe’s false belief is called a “False belief test”.  You might think that we just reason this out; however high functioning adults with autism are prone to get it wrong.  This has lead to training people on the autistic spectrum in “mindreading” skills that are automatic for most people.  One such man wrote a wise and funny book called The Journal of Best Practices.  One “best practice” was “Don’t turn off the radio when my wife is singing along to it.  Simulation Theorists say “figuring it out” is too clumsy a tool to explain the speed and accuracy of most people’s “getting” what others must think, or know, or not know.

At the height of the mirror neuron excitement, Goldman and Vittorio Gallese (One of the original discoverers of mirror neurons) wrote “Mirror neurons and the simulation theory of mind-reading” Trends in Cognitive Science, 1998 Dec 1;2(12):493-501. I had the opportunity to ask Goldman “Why monkeys” and he thought, well that’s just what they happened to be studying.  But 25 years after the first discovery, we have hundreds of follow-up studies of monkeys and some of humans, and no other branches of the animal kingdom.  Has even more than one species of monkey been studied extensively?  I’m not sure.  But recent experiments have shown mirror neuron activity in marmosets, whose common ancestry with Macaques goes back 30-40 million years.  This suggests a very old adaptation common to all monkeys.

I’ll go out on a limb and bet that monkeys are something of a special case, and predict that close studies of monkeys traveling in groups from tree to tree (see video) will show that if monkey A is following monkey B, A will generally copy B’s way of grabbing the next branch.  We might also find that the most agile monkeys take the lead in traveling.  Jumping from a branch across several feet and catching hold of another branch high in a tree is a tricky business. I don’t know anything about the speed with which macaques or other monkeys roam among trees, but to move quickly in a group might be highly dependent on this ability.  It could be a key to the evolutionary success of monkeys, by which they leave behind many predators.  Chimpanzees are a striking exception.  In most accounts of them hunting for food, they are hunting monkeys

Are monkeys very good at imitating in general?  According to Goldman, his coauthor Gallese, who studied them in labs, said “No”.  This suggests that monkeys’ motoric empathy might be little used except in treetop locomotion, so perhaps we should call it limited motoric empathy.  On the other hand, a Science Magazine article, “Capuchin Monkeys Display Affiliation Toward Humans Who Imitate Them” notes that “wild capuchin groups routinely synchronize their behavior; for example, for travel, feeding, and predator defense”.

Where am I going with this?

This article, based on various scientific and philosophical works and my own thinking, suggests a path from simple motoric empathy to empathy in the broadest sense, and beyond that to what Philippe Rochat, cited in Sarah Perry’s The Essence of Peopling, calls “Others in mind” – having, in our minds, a continuous presence of models of others, as illustrated by this image from The Essence of Peopling:

 

Action parsing in monkeys and humans

The skill of “reading” physical actions, not necessarily for imitating, is called “action parsing” (or action processing), a theme touched on repeatedly in The Shape of Thought: How Mental Adaptations Evolve (Evolution and Cognition) by H. Clark Barrett, and is a major topic for both neuroscientists and AI researchers.  If we really mastered it, we could just show a task to a robot, rather than programming it.  This has begun to happen, but it’s taken a long time, and many iterations of Moore’s Law.  Our closest relatives, chimpanzees seem to have a high level of action parsing because they are very good at picking up techniques from one another, including rudimentary tool use, and do so much more quickly and with fewer trials than monkeys. Domesticated chimps living with humans have been able to learn sign language to a significant degree, and they seem to do a lot of communicating through hand gestures in the wild.

High expressions of how humans use action-parsing might be watching and performing a choreographed dance, or those comic routines in which one person pretends to be the mirror image of another.  Recognition that one is being imitated is more widespread than being able to imitate competently, according to the article noted above: “Capuchin Monkeys Display Affiliation Toward Humans Who Imitate Them”.  This makes sense as to imitate requires “reading” another’s action while action parsing might go on passively it there is no motive to imitate.

Or consider the degree of action parsing evident in this description of a hunting party from Sarah Perry’s essay What Is Ritual?

One day, deep within the forest, Agaso, then about 13 years of age, found himself with  rare good shot at a cuscus in a nearby tree. But he only had inferior arrows. Without the slightest comment or solicitation, the straightest, sharpest arrow of the group moved so swiftly and so stealthily straight into his hand, I could not see from whence it came.

At that same moment, Karako, seeing that the shot would be improved by pulling on a twig to gently move an obstructing branch, was without a word already doing so, in perfect synchrony with Agaso’s drawing of the bow, i.e., just fast enough to fully clear Agaso’s aim by millimeters at the moment his bow was fully drawn, just slow enough not to spook the cuscus. Agaso, knowing this would be the case made no effort to lean to side for an unobstructed shot, or to even slightly shift his stance. Usumu similarly synchronized into the action stream, without even watching Agaso draw his bow, began moving up the tree a fraction of a second before the bowstring twanged.

Quoted from E. Richard Sorenson, Preconquest Consciousness

While there must be action parsing, much more seems necessary for what the article aptly calls “group proprioception”, or the several boys moving like one body.

Infants train selves to be social

Much of the explanation, I believe, is summarized in 8 pages in “The Ultra Social Animal” European Journal of Social Psychology (Apr 2014) by Michael Tomasello, the briefest and most accessible overview of a vast set of research and analysis.

“The Ultra Social Animal” (tUSA) among other things, illustrates and analyzes some non-obvious aspects of how children engage in cooperative tasks.  If the task produces a reward that can be shared, children insist on fair division, and sharing only among those involved in the production; quite unlike chimps who simply grab and what they get depends mostly on how close they are to the prize, and if some bystanders are close to the prize and get some, they are no more resented than are participants.  Commitments seem to be phenomenally real to older children. “When 3-year-olds need to break away from a joint commitment with a partner, they even ‘take leave’ through some form of implicit or explicit communication—as a way of acknowledging and asking to be excused for breaking the commitment” (tUSA, p189) This suggests that being a “we” in such a joint endeavor might involve a special mental state

Another aspect noted is that often children operating in different roles are conscious enough of the others that they can switch roles with little or no evident lag in competence. Tomasello suggests we have a “bird’s eye view” of social situations.  Could some conscious or unconscious part of our minds have such a view?  Could it have anything to do with out of body experiences, in which one looks down at oneself and one’s surroundings?  Sarah Perry, in Cartographic Compression cites cognitive scientist Lera Boroditsky’s time among speakers of the Australian aboriginal Kuuk Thaayorre language.  This language incorporates what roughly amount to compass points, and a person is expected to have a sense of one’s orientation to them at all times.  Boroditsky failed miserably at this at first, but then

After about a week of being there, I was walking along, and all of a sudden I noticed that in my head there was an extra little window, like in a video game. And in that console window was a bird’s-eye view of the landscape that I was walking on, and I was a little red dot that was traversing that landscape.

Boroditsky shared the cognitive change she experienced with a native speaker of the language, who commented, “well of course – how else would you do it?”  It is noted casually that a third of human languages have such a feature.  There are approximately 6,000 identified languages, most of them spoken by small wilderness-dwelling tribes as in Australia, Papua New Guinea, and the Amazon, so this may be predominantly a feature of such cultures.  Tomasello’s “bird’s eye view” doesn’t of course call for such a literal phenomenon as an inset screen, but if the latter is even possible, the former is made more plausible.

According to Tomasello’s research, and summarized in tUSA infants begin, as early as 9 months old, to initiate their own preparation for being good collaborators.  Well before the 2nd year, infants begin to seek out “protoconversations”.  Perhaps you’ve experienced this with a stranger’s baby in a supermarket checkout line.  The baby smiles tentatively; you smile back, and the baby smiles broadly.  You may initiate some surprising but not too alarming gesture, which you then repeat the baby comes to anticipate and it is done over and over to laughs like a sort of little game.

Somewhat later, infants want to point out interesting sights to familiar adults, with the clearly desired result of a shared emotive attitude, of surprise, laughter, or maybe sometimes worry (“Uh Oh”, being one of the first things children learn to say).

In his book Origins of Human Communication (2008), Tomasello provides these examples of parents’ diary observations of infants’ pointing in the context of everyday social interactions.

• At age 11 months, J points to the closed window when he wants it open.
• At age 11.5 months, J points to the door as Dad is making preparations to leave.
• At age 11.5 months, after Mom had poured water into J’s glass at the dinner table; a few minutes later J points to his glass to request that she pour him some more.
• At age 13.5 months, while Mom is looking for a missing refrigerator magnet, L points to a basket of fruit where it is (hidden under the fruit).

Sarah Perry’s Meaning and Pointing spotlights the primacy of pointing, but treats it, I think, in a more traditionally philosophical way, not as a mental adaptation (although the “whites of our eyes” which facilitate our following another’s gaze is mentioned as an adaptation – apes have round eyes without visible whites, or sclera).

As in the example of the refrigerator magnet, Tomasello’s laboratory work shows many instances of children at about this age displaying a seemingly sophisticated understanding of both the fact that someone is looking for something, and a reason why they don’t know where it is.  Recall that a false belief perception is when you know that someone else has the wrong picture of what is going on due to a missing fact or observation. Previously it was thought to be proven that the ability to perceive others’ false beliefs, and draw the consequences, emerged at around 4 years old. This is apparently true for the ability to do this using words and reason, but these new experiments suggest that without words, 14 month olds, having seen that another was outside the room when the ball was moved from Box A to Box B, can with immediacy know that she would come into the room thinking the ball was where it used to be.

By adulthood, we use gestures to convey more than is possible with a “look there” implied by pointing.  Earlier in the book (p67-68), are some examples of this which the author had observed:

• I am in a cheese shop in Italy, and I ask for “parmegiano.” (sic) The proprietor asks me  something I do not understand, but guessing — and not having the appropriate   word — I twiddle my fingers as if sprinkling grated cheese onto my pasta.
• I am at the front of the lecture hall, getting ready to give a lecture. A friend  in the audience fiddles with her shirt button, frowning at me, and sure enough  when I look down mine is unbuttoned.
• The airport security guard motions his hand in a circular motion to tell me to  turn around so he can scan my back.

It would seem that typical infants are able to parse the actions of others, and by imagining them in their own bodies, intuit intentions, and what has and hasn’t been seen.  In the case of the child seeing a ball placed in Box 1 as an adult watches, and while the adult is out of the room, the ball is moved to Box 2, it is as if the child had a mental simulation of the setting, with an avatar of the person who missed the box-switching.  There was a wall between her and the ball being moved, so in the avatar’s “mind”, which is within the child’s mind, the ball is still in Box 1. If this sounds fantastic, bear in mind that the child won’t be able to analyse such a “false belief” in verbally presented scenarios until age 4. Just this, and the motivation to point, and bring the others’ imagined world view in line with ones own could account for a huge leap in social intelligence

Adults have a greater repertoire of gestures, but are operating on the same principle.  Imagine how, with this faculty, even mute prehistoric hunter-gatherers might excel other animals.  Elsewhere, Tomasello argues that this could pave the way for sophisticated gestural which could in turn lead to speech.

“Others in Mind” when we Dream

So far, we are looking at very contingent and ephemeral social situations. A recent analysis of dreaming suggests something about ongoing relations with people we see repeatedly, and it speaks explicitly of avatars.

One of the most common theories about dreaming held by researchers is that it is necessary for turning the ephemeral memories of a day into permanent memories.  What are memories like?  A naive idea might be they are like stored film clips.  The beginning of this paper dealt with a kind of memory, of who Jennifer Aniston is and what she looks like, which is more like a concept.  Another sort of memory is what that valley, or the top of that hill, or the first home that you remember is like.  Turning the film clips of the day into idea-like memories, or assimilating them to existing memories would seem to require a lot of processing, perhaps with a good bit of parallelism, probably best done when the brain is “off line”, and the idea that dreams are a by-product of such a process makes intuitive sense, and might shed light on the chaotic nature of many dreams.

“The Avatars in the Machine: Dreaming as a Simulation of Social Reality” (2015) by Antii Renvonsuo, Jarno Tuominen, and Katja Valli, introduces a theory of dreaming, which I think can be built upon the previous one.  The paper is part of a virtual forum organized by Thomas Metzinger, and appeared alongside papers by Daniel Dennett, Ned Block, Paul Churchland, Vittorio Gallese (mentioned previously), Allan Hobson, and Jesse Prinz.  The primary author, Renvonsuo, has studied dreaming and consciousness for two decades.

The paper, which admirably calls for “risky predictions” as an antidote to “just so stories”, summarizes a great deal of research by saying in essence that dreaming provides opportunities for trying out encounters with people in one’s life, or ogres in one’s imagination.  Why do we do this?  Maybe the practice makes us better at dealing with real situations, and while we don’t remember much of our dreams, they may be shaping our ability to imagine possibilities in response to a real situation.

A social reality simulator would be quite an impressive thing to have evolved. It is doubtful that animals we evolved from had such a strong need of social simulation, yet unlikely that the short period of human evolution created such a thing out of whole cloth.

Would animal’s dreams have been simulating something simpler?  If animal brains could have a simulation engine, what might be its best use?  While some animals have innate fear of snakes, higher predators have to learn much about some unpredictable set of prey animals, like where to find them, how to run them down, and what they do to defend themselves.  Prey animals have a corresponding need to know about predators, how to get out of their reach; how to fool them, etc.  So, simulation of other species based on observation makes sense.

No animal, unless it dwells in a featureless desert, or has a tiny range, like cave fish, can inherit full knowledge of its environment.  Animals,  especially w.r.t. species that prey on them, need to gain as much competence as possible from as few “close calls” as possible.  Does this remind you of anything?  Like one of the first things simulated on computers: flight. What if they had a simulator in which to play “war games” with antagonistic species?  If dreams can bring to mind threat or prey species, with characteristics we know by observation, and superimpose avatars of these on familiar and realistic landscapes, new response sequences can be imagined and stored away.  A squirrel is surprised by a fox and he escapes by popping into a hole.  If dreams make random combinations of places and significant other species, a dream might have the squirrel attacked by the same fox beside tree, for mental practice in a different kind of escape. Two suggestive bits of evidence. Studies of mice and rats have shown that if the shape of their environment is changed, the nature of their dreaming becomes more intense for some period, and brain areas concerned with navigation are particularly active.  Now if only the experimenters had thrown in some novel simulated predators, or new ways of acquiring food.

For the second bit, returning to Renvonsuo’s paper, one finding is that children are much more prone to dream of animals than adults, except if the adults have something like a hunter-gatherer existence, the proportion of dreams of animals remains high.  This suggests an older function of dreaming, plausible for lower animals, and out of which social simulation might have evolved, but which usually diminishes as we age, and we have no life or death encounters with other species.

In recurring dreams, we often face the same people, or for children, the same monsters repeatedly.  If these represent avatars, then might an avatar of, say your mother, brother, boss, or spouse be a sort of permanent fixture, that hangs around from night to night, helping you refine attitudes and responses to these people?

Might it turn up in the daytime?  Have you ever had an imagined argument, often a rehash of a bad, mad conversation, that just keeps playing and you can’t stop it.  You try to improve on a dumb response, and they come back with something else, and then you say whatever and then they say whatever, all pretty much in character (or our likely flawed version of the other’s character).  I know I have.  Anyone else?  People on the autism spectrum, Yes or No?

Dan Sperber is a cognitive scientist who thinks and writes more like a philosopher. In 1986 he wrote, with Deirdre Wilson, Relevance: Communication and Cognition. In effect, it says when we communicate with a person (one whom we know), we don’t simply “send a signal”.  We know as we bring the listener to mind what is relevant  to them and what is not.  We may know that in the context, a single word, or glance in a certain direction may be all it takes; and we act accordingly, usually without any consciousness of doing so.

When we communicate with someone, does the communication, in a sense, go by way of an avatar for that person in our mind, modeling what they know or don’t know or, phenomenologically speaking, modeling “being that person”.  We do this imperfectly, but often very well, and with little if any conscious effort.  And is this avatar the same, or at least continuous with, what we get when we happen to dream of that person?

Peer to Peer: Implications and Further Directions

The way humans seem designed to interact, and create action and knowledge seems to me very much like a sort of peer-to-peer protocol — for combinations of people in roughly equal relationship; not a protocol for one to control many.  As Venkat wrote in  Your Passport to the Meta-brain (Breaking Smart Newsletter), “In the non-exclusive mutual containment metaphor, however, each of us exists as both a living conscious being, and as an evolving digital ghost presence that others can include in their second brains.”  We seem designed for free agent collaboration.  Our sociality and our individuality are equally necessary. Infants take their own initiative in learning the social skill of language and in teaching themselves to be social.

We are by nature curious and driven autodidacts, but this produces a restless energy that educators have hardly ever known how to channel.  I would argue, given more space, that the Khan Academy’s methodology is far more suited to model of humanity I have been developing here.

Most evidence indicates the hunter-gatherer societies in which our nature took shape lacked extreme dominance.  A settlement typically had something anthropologists call a “headman”, but he generally needed a lot of persuasion to get much done.  The potential for the “alpha male” typical of chimpanzee society didn’t disappear from our natures, but reemerged with a vengeance on the heels of the ultimate disruptive technology: agriculture.  Early agrarian empires, at least in Mesopotamia, South and central America, and China, all seemed to pass through a horrendously cruel phase.  James C Scott’s not yet released work: Against the Grain: A Deep History of the Earliest States partly dealing with barbarian cultures having long parasitic, and fairly happy (for the barbarians) relations with such civilizations might indicate why.  Early agrarian societies were tied to a place, rapidly growing in population, too big for headman/consensus governance, and piling up surplus that made them the inevitable target of  less sedentary tribes.  The experience of war with its necessity for acting en masse, preferably with the unpredictability that a single dominant secretive mind can achieve shaped society, and more or less did so until the time of emergent democracies.

Since the acceleration of technology reached a certain point, the superior creativity of more peer-to-peer-ish societies tended to prevail, but in recent decades, a great many non peer-to-peer societies have literate elites capable of gathering technological power that they could not have created.

While much writing on the “singularity” is foolish, we do seem headed towards a very big, strange, and accelerating unknown.  We are experiencing something like the Gutenberg revolution which helped unleash two centuries of religious wars, but we are experiencing it in more like two decades.  In a torrent of disruptive changes, including those that might put the ultimate destructive weapons in the hands of the least wise and balanced, it is foolish to think we know how it will turn out, as many seem to both optimists and pessimists.  Just to give one example, genetic engineering could reach a state where any unbalanced but highly intelligent fool could design an “ultimate” virus, capable with the right dispersal methods of wiping out a population in days.  The same technology should in theory provide the means of quickly analyzing and stopping such a virus, but it won’t matter if we have a failure of imagination and are oblivious to such possibilities (and if I thought I could think of all such risks I’d again be foolish).

Many threats are analogous to the virus example; we have races between dangers and remedies, both accelerated by Moore’s Law and knowledge explosion.  If we are to reach a happy rather than nightmarish version of living in the meta-brain (Venkat, Breaking Smart Newsletter), or whatever the future looks like, this model, if correct might be one necessary addition to our toolkit.

The more exact sciences are slowly becoming relevant to human social problems, and our apparent stuckness at “Just So Stories” is, IMO, only temporary. Within a decade, we could make real progress in understanding how the genome produces the developmental process out of which humans are formed, and such an intimate knowledge of the genome, and a refactored way of seeing it should reveal more about our mental and social development, and what parts are built on what other parts (Sean B. Carroll in The Making of the Fittest gives interesting examples of this) and allow us to define more operationalized versions of human social and other qualities.  We are somewhat at the stage where we call anything that swims in the sea a fish, whereas deeper knowledge tells us there are fish and cetaceans, totally different from one another.