Knowledge darwinism

Jump to navigation Jump to search

This article by Dr Piotr Wozniak is part of SuperMemo Guru series on memory, learning, creativity, and problem solving.

Definition of knowledge darwinism

Knowledge darwinism is the competition between memories for the best place in the jigsaw puzzle of knowledge.

To illustrate knowledge darwinism, consider three semantically identical questions, which store the exact same memory given sufficient knowledge of the context: (1) Which year did WW2 start?, (2) On September 1 of which year, did the Wehrmacht attack Poland?, and (3) Which year did Hitler unleash a world war?

Depending on the current knowledge of the student, one of those questions will have the best chance to survive in long-term memory. The question about World War 2 seems most universal. However, for a student from Poland, the question that involves her own country may stand a better chance. For someone with strong feelings about Hitler, the third question may have the greatest staying power.

Improved coherence

In incremental reading, asking the same question in different contexts is a norm. We call it knowledge redundancy. Paradoxically, adding more questions to SuperMemo may actually reduce the total workload. If one of the semantically equivalent questions takes root in long-term memory, it helps consolidate and retrieve the remaining questions. Moreover, gains in knowledge coherence may contribute to better recall of other memories in the context. For example, memories of Hitler, Poland, or war may all become refreshed, and more coherent.

Building stability

In incremental reading, new questions (cloze deletions) are generated dendritically on demand. When the student senses her recall of a given piece of knowledge may suffer due to the lapse of time, she may choose to generate a new question. In a statement "Hitler unleashed WW2 on Sep 1, 1939", few students will bother to cloze Hitler or WW2. This is a result of good recall of those concepts in that particular context. On the other hand, "Sep 1" or "1939" will be more volatile, and the student may opt to cloze. This can happen in one, or in many contexts depending on the quality of knowledge, and knowledge retrievability. This contributes to the overall stability of memory. While mutations of memory via generalization and forgetting are welcome, stability makes sure that winners of the Darwinian process establish a strong position in the structure of knowledge.

Building comprehension

Multiple cloze deletions may contribute to the maintenance of a specific semantic association (i.e. a specific atomic memory). Those multiple angles of the same thing favor generalization, i.e. the development of abstract knowledge with high applicability. Without student's participations, overlapping statements contribute to the emergence of new abstract quality, i.e. better knowledge and better comprehension. This type of emergence will continue over years both in waking and in sleep.

Convergent learning process

In a well-managed incremental reading process, knowledge redundancy is subject to saturation with respect to each atomic memory. In that sense, the understanding of the surrounding universe progresses incrementally, dendritically, and largely unidirectionally (i.e. towards better comprehension and better recall). In other words, while incremental reading is a form of life-long learning, individual memories become well entrenched and easy to maintain for life.

Darwinism in free learning

Knowledge darwinism underlies the mechanics of the learn drive, and has a powerful impact on signal learntropy. Understanding knowledge darwinism makes it possible to understand the power and importance of free learning. For the entire darwinian process to begin, memories need to be matched to the current status of knowledge. In simple terms, without comprehension, there is no learning. The dissonance between information on input and the current status of knowledge is the chief problem of schooling.

Knowledge evolution in a population

A similar Darwinian process occurs at the level of the population. Contradictory pieces of knowledge fight for supremacy in a viral process that leaves the fittest memes in the game. A paradigm shift may be necessary for a minority meme to break through on a larger scale. While population-level combat is based on verity, emotional impact, biases, prior knowledge, etc, brain-level competition is more dependent on stability, coherence and consistency. Those two levels work as a whole, and separate cultural transmission from neural memory in collective human intelligence.

For more, read about the science of memetics.


In summary, knowledge darwinism is the competition between semantically equivalent and contextually varied pieces of knowledge for survival in long-term memory. Knowledge darwinism maximizes knowledge coherence, which may be boosted by reconsolidation based on connectedness. Such reconsolidation might be taking place during memory optimization in sleep. Darwinism improves comprehension and the pleasure of learning. Exuberant knowledge darwinism may have a dramatic impact on reducing the cost of learning by building large sematic networks composed of best-fit pieces of knowledge.