Monday, September 16, 2013

Raising Funny Kids 41: Transitioning Imagination to the Next Level

The Backstory

A young boy of 12-years old walked up to his Headmaster and upon noticing a button on the outside the Headmaster's office inquired whether or not there was a lift inside. 

The Headmaster in his compassion and wisdom said that he did not have a lift in his office but that his office did in fact operate in a way that lifted others when they came looking for advice. Not like a Tardis or lift, but in the same spirit.

This is a wonderful story that leaves one feeling relaxed and soothed, like a warm cup of tea. 

But, What if..?

What if that boy had already been guided toward a conceptual understanding of the basic foundations of quantum theory... how would a young mind, amply sprinkled with a little bit of chaos theory to keep that adroit imagination strong and alert, naturally travel from point A to insight? 

Traditional education typically teaches children to travel from A to B or A to Z, but it is not with deductive logic and temporal lobe strengthening alone that can deliver the mind face-to-face with an epiphany ~ it is imagination; something this young boy naturally exhibits or has somehow managed to harness in order to ask adults questions. 

Using his imagination and "status" as a child, he might be one of those very clever kids who use imagination as a tool to communicate to those adults still capable of listening that he is ready for new insights. 

Either way, by inquiring about the button, this boy is in fact giving an adult a precious opportunity to help him transition his "lift theory" (his imagination) toward a place where his ability to see something others fail to notice is rewarded in deeper insight and understanding. 

This is the perfect time to transition imagination to a higher realm of understanding. And who better to introduce that world than Dr. Quantum...

Dr. Quantum

Dr. Quantum (in my opinion) has produced 
the best introduction to Quantum theory for kids.

There are some points in Dr. Quantum's descriptions that can be fine tuned with respect to the subject matter for those wishing to delve deeper into the field of inquiry, but it is not necessary for spark of insight to occur. It is this Spark of Insight that we want to give a child in order to help them transition or project their healthy imagination toward a sound understanding that wonderment is essential if one wishes to explore the nature of the universe. 

It is my earnest understanding that we cannot truly know all that which we think we know; that some things might not be as they seem, but that "reality" ~ in and of itself ~ is a poignant reminder that what we think things are might not exactly be what they are at all... and it is in this understanding where a new story begins to take shape.

Did you know that...

Some people might not have ever noticed that button and connected it together with the idea of a lift being in my office. But because you noticed, I'll let you in on a little secret... very few people people notice the button, much less connect it with the idea that I might have a lift in my office. 

In fact, some people dismiss the button entirely. They might think your question is funny or ridiculous. They might even laugh or feel uncomfortable thinking that someone out doesn't know the truth. 

But this is a big world, in an even bigger universe, in an even bigger dimension ~ of which there are many, so many that there are more dimensions than there are grains of sand on all the world's beaches combined. 

This tells us that there is more that we do not know than that which we think we know or could ever learn in our entire lifetime

If you are the type of person who prefers to say, "No, there is no lift in Headmaster's office," then you would be like most people, and those people are right ~ they can prove it! They can make a fist and bang it against the wall and walk all around the room and not uncover any hidden doors like on Scooby Doo... thus proving to the world that there is no hidden door. They will be happy and rejoice in their shared understanding that some things are real and some things are imaginary. 

But there is another type of person, there is a person who sees the button and immediately connects it to a device. After all, we know that objects have three characteristics in common. They have a "form" ~ a "fit" ~ and a "function" ~ because of this, you demonstrated sound reasoning, an understanding that with a button often times follows a mechanism that would unlock a device, such as a lift. 

And what better place to have a lift but in the most important office in all the school ~ The Headmaster's Office. Even though we do not know everything the Headmaster does, we know that he is in charge of sailing the ship. 

It could be that Headmaster's button operates a hidden lift that was actually built during war time ~ a secret entrance/exit of sorts built by Scottish officers who might have used the school as a secret command center. After all, our school's history dates back to the 1800s.

Given this knowledge, the first thing we should do is examine the room for clues. If we find something interesting, we can explore it further. 

If we do not find a hidden lift, we still have the satisfaction of knowing that we looked for one. And "looking" is something we can do for our brains to train them to "discover" or "see" more than we normally might. 

This is how Einstein saw the world. His teachers did not see what he said he saw and they tried to convince him in good conscience that there was no "lift" in his school. And by "lift" I mean whatever it was he thought he saw that others did not see. 

His teachers were no doubt good people just like your teachers are, and I'm sure they cared about him the way your teachers care about you. So when they told him there was no "lift", they believed what they were saying was true. 

But what some people fail to do is pause... 

...and wonder whether or not there "might indeed" be a lift that they, in their rushing about to do all the things people rush about to do, might have mistook for a wall. Imagine!

How many people will look for the lift, just in case? 

Even if you don't find a lift, it doesn't mean that one couldn't be built! 

And this where life gets really fun! We have the ability to build and create the things we think might belong or should naturally be there... and by doing so everyone sees the lift... 

Because... you see... in all reality, the lift is already there. It just hasn't been built yet. It is only a matter of waiting for someone like you to come along and ask the question: Is there a lift?

Maybe there is...

Further Reading...

For a long time philosophers have emphasized the active nature of perception and the intimate relation between action and cognition [1], [2]. Cognitive behavior results from interaction of organisms with their environment, which "appears to be filled with regularities" resulting from past experiences [3]. 

Our environment is usually filled with a plethora of stimuli and to discover those "regularities" we constantly have to discriminate between relevant and irrelevant features. Most kids, had they even noticed the button, would have learned to dismiss the button as "irrelevant". 

Why would they dismiss the button?

The seven deep brain nuclei of the basal ganglia are involved in a variety of crucial brain functions [4] and are tightly linked to the dopaminergic neuromodulatory system, which plays a fundamental role in predicting future rewards and punishment [5], [6].

By age 12, most children have been trained by well-meaning adults to seek reward and avoid punishment. The result of this thinking delays a child's natural ability to notice the button as well as easily and quickly solve the candle test. 

Children are taught to interact with their environment, initially guided by trial and error, in a way that helps them find a mapping between states and actions that will yield the maximal future reward. 

However, how do we keep our minds open to unknown but equally relevant inputs? How does our cortex determine which additional features might be relevant for selecting actions and obtaining new rewards? 

From a technical point of view it would be straightforward to first learn the state space, i.e., extract features, and then find the mapping between the states and actions. This two-stage learning is a common approach whereby we train a simple neural network based on rewards on top of features, which before have been extracted with the attention-gated reinforcement learning model of Roelfsema et al. [7], which represents a link between supervised and reinforcement learning. 

However, learning is slower due to insufficient feedback when the child guesses incorrectly. In the case of the button-lift hypothesis, not finding a lift would be disappointing and hence the temporal credit assignment problem would not be satisfied with this approach.  

Distancing ourselves from 19th and 20th Century pedagogical models, we can look to what was referred to as "imagination" what one might call a world model [8]. The child has a map of the surrounding area which allows him to do planning. The Headmaster in the above scenario simply provides the pupil who inquires about the presence of a lift with this new information (in this case an introduction to the quantum world or the history of the school). 

The variant of navigation lies in wonderment, a place in which the child navigates towards a goal position, e.g. to perform some action like pushing the button, user interaction. A task like pushing a button usually does not require a map, but is constrained by the affordances of the goal position, at which the child often needs to arrive with a specific belief (in the possible existence of a lift or something previously undiscovered). 

Most people only rely on information that is directly available to their ocular organ. The child is supposed to learn the relevant visual features and develop sensorimotor laws based on his or her interactions with the environment. Initially, a child does not know if there is a lift in the Headmaster's office, i.e., where the target region is, and reward is only received after the final exploration or movement leads to a successful discovery. 

However, by harnessing a child's natural imagination, they can be taught to extract task-relevant visual features as well as assign adequate actions to those, in a single-step procedure and within one united architecture or sense of coherency. 

This enables a child to build upon that winner-take-all layer that considers goal-relevance from sensory input dimensions valuable, and take that to a new level whereby he or she does not automatically neglect irrelevant parts but rather utilizes a prediction error that not only modulates learning into value functions and action strategies but that also uses these to adapt the weights of the features that are responsible for learning action-relevant input associated with specific actions. 

In other words, a child might not know whether there is a lift in the Headmaster's Office, but he or she still checks just in case. 

While it might not always be time-prudent to check everything just in case, this approach is the basis upon which science is founded: 


Knowing how to explore is just as important as knowing when to explore. 
The rest is merely fine tuning. 

  1. J. Dewey, "The reflex arc concept in psychology," Psychological Review, vol. 3, pp. 357-370, 1896
  2. M. Merleau-Ponty, The Structure of behavior. Boston: Beacon Press, 1963.
  3. H.R. Maturana and F.J. Varela, The tree of knowledge: the biological roots of human understanding, rev. ed. Boston: Shambhala, 1992
  4. V.S. Chakravarthy, D. Joseph, and R.S. Bapi, "What do the basal ganglia do? A modeling perspective," Biol Cybern, vol. 103, no. 3, pp. 237-53, Sep 2010
  5. W. Schultz, P. Dayan, and P.R. Montague, "A neural substrate of prediction and reward," Science, vol. 275, no. 5306, pp. 1593-9, Mar 1997.
  6. W. Schultz, "Predictive reward signal of dopamine neurons," J Neurophysiol, vol. 80, no. 1, pp. 1-27, July 1998.
  7. P.R. Roelfsema and A. van Ooyen, "Attention-gated reinforcement learning of internal representations for classification," Neural Comput, vol. 6, no. 8, 2010
  8. S. Thurn, W. Burgard, and D. Fox, Probabilistic robotics. Cambridge, Mass.: MIT Press, 2005.
  9. J. Kleesiek, A. K. Engel, C. Weber, S. Wermter, "Reward-Driven Learning of Sensorimotor Laws and Visual Features," Development and Learning (ICDL), 2001 IEEE International Conference, vol. 2, 24-27 Aug. 2011  

No comments: