Unexpected Uncertainty, Einstein And the Mystery Of Learning
Unexpected uncertainty is a scientific term. It was coined in 2003 by neuroscientists Angela Yu and Peter Dayan from the Gatsby Computational Neuroscience Unit in London. Authors had an intention to clearly delineate two types of uncertainty: expected and unexpected. It is an important distinction because our mind treats those two types differently. Since then many neuroscientists have identified different areas and processes in our brain which get activated when we deal with those two different types of uncertainty.
If we know from our past experience that the odds of an event to occur are less than 100% and more than zero — it is an uncertain event. It may occur with some degree of probability that we also can know from our past experience. Thus, when tossing a coin several times we rightfully expect that it will fall heads up in 50% of cases on average. However, we are uncertain about the outcome of each single try. Hereof, the term expected uncertainty.
We can toss a coin a million times occasionally winning or losing if we make bets. We can exploit our knowledge of the probability of winning or losing in gambling but we will not learn anything new about that probability as long as it remains unchanged.
If by an accident the coin will get bent, the probability of heads against tails may change to 30/70 split. If we keep betting on heads after the change of probabilities we can lose a lot more than expected before we notice that something went wrong.
We will wonder what’s happening and start to observe the tossing more closely because something unexpected has happened. Yet, we remain uncertain about the new probability split. Therefore, unexpected uncertainty. We can learn the new probability split by exploring the new evidence further.
Unexpected uncertainty arises from errors in exploiting expected uncertainty when the known odds suddenly change. Like expected uncertainty it contains risks but unlike expected uncertainty it also provides a unique opportunity for learning from mistakes. Scientists defined as prediction errors such mistakes, on which we can actually learn, to distinguish them from mistakes of losing bets in the gamble with already known odds — i.e. from expected uncertainty.
Unexpected uncertainty is also known as ambiguity to neuroscientists who are specializing in adolescent cognitive development. They keep discovering new evidence that ambiguity plays a crucial role in the integration of the network responsible for autonomous decision making in an adolescent brain as it matures.
Unexpected uncertainty makes the very process of learning perpetually self-rewarding.
Expectedly uncertain events can make you happy or disappointed after their positive or negative outcomes will be achieved. If you are gambling with known odds you can either win or lose but neither result will surprise you. You won’t be curious to discover the odds as you already know them.
In the vast but yet limited number of cases when expected events occur almost certainly people can use rewards and punishments to facilitate sharing their knowledge of such events with each other. That practice is most commonly known as learning or reinforcement learning, in which acquisition of the new knowledge is reinforced with rewards or punishments external to the content of learning itself.
Reinforcement learning proved to be extremely successful for the exploitation of the existing almost certain knowledge simultaneously by many people, which do not necessarily understand the knowledge that they exploit.
Unexpectedly uncertain events surprise you. They make you wonder why they are happening even before their outcomes surface. You will keep wondering about their mysteries, their hidden causes for as long as it will take from you to discover them. That process of exploration, of seeking for the new knowledge may perpetually lead you from one mystery to the other because all human knowledge is inherently incomplete.
None can know everything but all can learn perpetually. Seeking for hidden causes of unexpectedly uncertain events, hypothesising about them, testing hypotheses for their validity and occasionally discovering the cause that works — only such process of learning on unexpected uncertainty leads to the emergence of understanding. Understanding allows us to modify our knowledge, to apply it under unfamiliar conditions, to build even more knowledge upon the existing base. Learning with understanding of already existing knowledge equals rediscovering it by each learner.
The process of learning from unexpected uncertainty may take milliseconds or decades to arrive at the next intermediate level of understanding. People use it in everyday life almost unconsciously. Great scientists over their entire lives steadily and sometimes consciously mastered that way of learning because it underpins the very nature of a scientific discovery.
As Albert Einstein once wrote, “The most beautiful thing we can experience is the mysterious. It is the source of all true art and science. He to whom this emotion is a stranger, who can no longer pause to wonder and stand rapt in awe, is as good as dead: his eyes are closed.”
