Sunday, June 9, 2013

STRESS -- new science of what it does to you and to schools

There is new science here, far beyond what you can find in popular articles on stress. But I'll try to work my way to the science through examples and stories.

Imagine a two-year-old playing with a ball in the living room. The ball rolls behind a couch.
When the kid was young, she might have suddenly thought the ball does not exist. Out of sight, out of mind. After a moment of perplexity, she would have moved on to other things in the room. But by two years of age (or earlier), the kid has learned "object permanence." She can "see" in her mind that the ball is still there, behind the couch, and she will probably crawl around to get it, and keep playing.

BUT -- as she plans to crawl around -- what if she suddenly hears her mother screaming behind her,
and hears other worrisome loud noises. She could turn around to look... and then... when the alarm is
over.. the ball may no longer exist in her mind. She forgot all about it. She is especially likely to forget if the alarm was really scary to her.

So what is really going on here?

"Object permanence" is a kind of memory, but it's not like the long-term memories which we have to work to recall. It's more like "short-term memory," part of our image of the present state of our immediate environment. The long-term memories are stored in a kind of hard-wired form in the brain, embedded in stable chemicals. But the short-term memory is stored in electrical form, as a kind of reverberation of electrical currents in the brain.

This past Wednesday, I heard a talk by Professor A. Arnsted, a neuroscientist at Yale, on her current research into short-term memory and the things which disrupt it, which have far-reaching implications for human life and human development.

It seemed as if Prof. Arnsted had been hired to try to fill the shoes of Professor Patricia
Goldmann-Rakic of Yale one of the truly great systems neuroscientists, whose work was
really essential to those of us trying to understand how brains really work (and how to build them).
Goldmann Rakic was famous for work on the frontal lobes of the human brain, the part which
are often considered the highest part of the human brain. "Great frontal lobes," said Isaac Newton --
and many after him.  She studied how "working memory" (short-term memory) results from
reverberations in the neutral networks of the frontal lobes.  Many of us in the neural network engineering area have followed up on this idea; "time-lagged recurrent networks" (TLRN) bring the same kind of power to artificial neural networks. This time-lagged "memory" is crucial to
good performance in things like controlling car engines in a way which reduces pollution and increases mileage; it is widely used in key industries, even though many ivory towers have yet to
catch on to what is really going on.

Years ago, I disagreed with the idea that our working memory, our short-term memory or "reconstructed image of reality," lives in the revervebrations in the frontal lobes. I argued that
it exists in ALL the time-lagged reverberatory connections in ALL of the cerebral cortex.
(See some of the books edited by Karl Pribram.) Many agreed. Students of Goldmann-Rakic have come to NSF, explaining that it is the reverberation, not the frontal lobes, which contain the memory.

(What disrupts our memory, and our "seeing" of things we do not see?
How does that affect our schools? I'll get to that...)

But then a story.

In 2005, the International Neural Network Society decided to hold its annual meeting
in Montreal. It invited a top neuroscientist from Montreal to give a big plenary talk on HIS research
on the frontal lobes. After fiddling with his laptop for half an hour, he described a whole series of experiments he did. In essence, he said: "My experiments do not agree with Golmann-Rakic's explanation. Let me give you all the details..." Many people did not understand the details. It took some energy to make real sense of what they mean. So I voiced and published my own interpretation of what he was saying (in the journal Neural  Networks): "The most advanced leading edge parts of the human brain are two specific parts of the frontal lobes -- orbitofrontal and dorsolateral. These
advanced parts of our brain begin to help us cope with the two most fundamental challenges facing our human mental abilities. These challeneges are to answer the two great questions in human life:
(1) where did I leave my car this time in the parking lot; and (2) what was it that I was trying to do anyway?" Of course, (1) is just an example of a more basic principle.

At Yale last Wednesday, I was delighted to hear that Arnsted has done experiments exactly on humans' ability to remember where they left the car in the parking lot. I asked Luda this morning:
"Did she get permission to meet hapless shoppers in Costco, terrify them, and watch how they forget where they left their cars?" Of course not -- but she did get permission to do something in the lab of the same flavor. (And also, experiments on people forgetting what it was they were trying to do.)

As she spoke, she reminded me of a woman I once knew named Kathie,
who spent much of her life trying to understand how strong hormones affect human thought and human behavior.  (I never talked about she affected my own hormones, except to Luda, but that was
certainly part of my own working memory, which did get recorded in my permanent chemical and even spiritual memory.) She talked about more than a dozen key hormones which change how things
work in the frontal lobes. She said that she has focused on three MOST important hormones, which have the biggest effect: (1) adrenalin (more precisely, norepinephrine), which is basically a fear response; (2) dopamine, which is a kind of reward or hope response; (2) acetylcholine (ACTH?),
which increases general awareness.

Basically, in situations of stress or fear, she has traced how adrenalin blocks or shuts down
those reverberations.  So the scared little girl forgets about that  ball behind the couch.
She has published lots of papers describing the details -- but in this talk to the engineering community, she basically told us to look up her papers for details. (I will, but not today.. too many emergencies..).

Since stress shifts the brain to reliance on lower, less capable levels of intelligence,
it reduces the use and development of the frontal lobes (and of higher cortical capabilities in general).

One quick corollary: if schools or parents or society terrify the children, intentionally or not,
it may have huge implications for the development of the capabilities of the frontal lobes,
and of the cortex in general.  Levels of adrenalin are a crucial variable which needs to be accounted for when we think about education.

That reminds me of a talk I heard, in one of the annual review meetings of the NSF Science of Learning Centers. People asked: "why is US mathematics education so poor compared to
many other countries?" In one study, they observed typical elementary school teachers, shifting form one subject to another. Sheer mention of the word "mathematics" led to an immediate spike
in the levels of adrenalin in the average teacher's bloodstream. Children being responsive...
it is no wonder that learning has been blocked so much.   It reminds me of the sci fi novel Dune, where hey said "Fear is the mind killer." But what of those Indian kids who seem to do better
in math so often? Two weeks ago, at the International Space Development Conference, a friend
who works with Indian schools explained: "A large fraction of that is kids who grew up
in special schools, where they were always told how great they were, and they didn't get ENOUGH
'fear of god' and humility..." Whatever. They did learn things in those schools.

Dr. Arnsted stressed how high levels of adrenalin or stress do get translated into different TYPES
of lower level responses -- sometimes fight or flight, and sometimes just freezing. She described how
she herself fell into a natural adrenalin freezing behavior, which worked quite well when she found herself close to a bear one time in the wild... But can people LEARN to keep their frontal lobes working in the face of stress and fear? (Is that more a subject for future research? And for adults?)

And what IS stress, and how does it relate to fear?

The neural network folks in the audience were a bit puzzled by this. "We don't have global hormones like adrenalin in our models. How can we make sense of this. Is this something we build computers
to be able to AVOID?"

So I got up to explain, and I add a bit here:

"Some of us in the neural network field moved on many years ago, before the PDP books, to
consider engineering designs which are far closer to the brain than the simple feedforward
networks you guys keep saying we are stuck with. We aren't the ones who are stuck. The ones who are stuck are the ones who keep referring to popularized accounts which oversimplify the math which was developed well before the popularized accounts were written.

"The brain actually uses recurrent or reverberatory connections for at least three different things,
each of which needs to be understood on its own terms. There is the time-lagged memory effect.
There is also what I call 'simultaneous' recurrence, like how we focus and gradually grow an understanding of something tricky, like how to 'see' the chessboard we are staring at. And there is also a kind of lower-level memory function, closed to the hard-wired stuff.

"If someone screams at you and really scares you when you are staring at a chessboard, you
can really lose that inner strategic picture you were developing. After the shock, you may have to start all over again, after your adrenalin levels off enough. And nasty chess players do sometimes try to raise the adrenalin level of their opponents. That's the simultaneous recurrence.

"This connects to the real math design as follows.  When we actually design simultaneous recurrent neural networks (to be able to play a good game of chess for example), we face a perpetual dilemma:
to converge quickly to the wrong answer, or to take the time to get to the right answer? Do we "reward" our neural networks based on the quality of their first guess, or the quality of what they come up with after some time to think? (In fact, this issue of "stress" or "tension" is described exactly like this in chapter 3 of the Handbook of Intelligent Control, which is posted at
www.werbos.com/Mind.htm. ) There are lower level circuits in the brain which are more feedforward, to provide fast initial guesses about what to do, and others which give better answers but need time to find them. Professor Arnsted appeared delighted that we could provide some level
of functional engineering understanding here, which did fit with what she sees.

Notice, then, that "fear" and "stress" are not the same thing, even if they often occur together.

Of course, "fear" and "hope" are much easier to understand; they are part of
"adaptive critics 101", which I have given endless tutorials on, some posted on my web page.

=====

For those who get to the more detailed paper on "how to build a mouse brain," which
I mentioned in an earlier blog post.... it should be sobering that good mechanisms for
creating and managing "stress" are important even to "VECTOR INTELLIGENCE,"
a level of brain or intelligence far below that higher level we see in the brain of the mouse,
which in my view is much lower in turn that the highest spiritual capabilities of the human mind.
We are not ABOVE this issue of stress; rather, it is part of the path we must cope with and understand, before we can really rise to a full mastery of higher levels.

Best of luck...


1 comment:

  1. Very nice summary.

    Let me add a couple of thoughts that may be relevant.
    Yes, the stressed child, the fearfull child may not learn.
    And yes at the special school where kids are told how great they are; they may learn better. But why? Is it lack of fear or something else? For sure lack of fear (inappropriate fear) is a good thing. But what about something else? and what something else?

    May I suggest the most powerful drug on planet Earth. The drug that is the standard against which all other drugs are compared. I recently read an advertisement for a new cancer drug that went something like this. "With this new chemo drug, the average patient can expect to live 18 months. Versus with placebo the patient will only live 15 months."

    I'm not here to argue the chemo case. I am here to argue the placebo case. We barely understand how powerful a placebo is. That mind body connection is seldom tested, and even less seldom taught and developed. Yet I would suggest adding the placebo as a positive effect in enhancing memory.

    Another thought on your seeing chess insight. Yes you are correct it takes time to see chess or physics or storeyline problem solving in writing a novel . Your words, "how we focus and gradually grow an understanding of something tricky", capture the question.

    How do we focus and gradually grow an understanding of something tricky?

    Let me suggest a way of enhancing seeing that I use most every night. Dreaming. I have done dream experiments upon myself for over 40 years. Many nights I give myself a task, solve that physics problem, solve that storytelling plot character problem, focus on this family problem; or find what I might be missing and not seeing in my understanding of such and such. Most mornings my mind has provided me a solution. I may not realize it at first; but as I work I say, "look at that." And I see a connection that I didn't see before I went to bed. My mind so to speak has been looking at the chessboard all night.

    Just those two thoughts.
    These are really excellent posts Paul. Thanks.

    ReplyDelete