Incremental reading and related techniques for learning knowledge available from the Internet

Dr Piotr Wozniak

This article is an adaptation of "Devouring knowledge" for the book "Knowledge Based Information Retrieval and Filtering from Internet" Marvin Minsky: Our cultures don't encourage us to think much about learning. Instead we regard it as something that just happens to us. But learning must itself consist of sets of skills we grow ourselves; we start with only some of them and slowly grow the rest. Why don't more people keep on learning more and better learning skills? Because it's not rewarded right away, its payoff has a long delay

Introduction

This article summarizes my 20-year-long effort in developing techniques and technologies that assist human learning. It illustrates ways in which learning can become a rationally controlled and conscious process. Among others, I discuss two concepts that can have a substantial effect on learning:

  • spaced repetition: a technique that ensures high recall (above 95%) with minimum investment of time via computing optimum inter-repetition intervals 
  • incremental reading: a fast reading technique derived from spaced repetition that makes it possible to read a large number of articles simultaneously in small increments (thousands of articles can be processed in parallel)

Knowledge as the driving force of progress

Knowledge and history: Throughout the ages, knowledge was the cornerstone of human progress. From Stone Age to Information Age, in pain, we have built a tiny oasis of civilization in ruthless expanses of the evolving universe. The history of the mankind is made of billions of individual lives that keep on sparking and fading. Born of self-preservation instincts imprinted by evolution, history books paint a picture of a constant string of wars, conflict of interest, loss and gain of influences, lust for power and submission to weaknesses of human nature. On the other hand, the greatest achievement of the evolution, the rational mind, kept on contributing to new findings, discoveries, technologies and philosophies. Progress has always hinged on discovering new truths and preserving them for posterity in form of stories, solutions, tools,  books, and other carriers of information. Knowledge is the basis of human power, yet it constantly struggles with two forces that regularly diminish it: death and forgetting. We can preserve knowledge in books and other forms of information storage. However, this knowledge translates to value only then when it is used by the creative power of the human brain. The limitations of the human brain will remain a bottleneck of progress for many years to come. We will develop artificially intelligent knowledge processors not earlier than in a decade or two

Knowledge and death: Death poses an ageless challenge to educating new generations. Years of hard work needed to gain knowledge on professorial level are obliterated in a single act of death. Newborns need to go through years of education before they are able to access, read, and comprehend this text. They all have to struggle with basic literacy skills, lessons of safe sex and teen pregnancy, lessons on superiority of altruism over egoism, the difference between wise and not-so-wise choices, existential questions, etc. Although constant reeducation may contribute to gaining a fresh perspective in each generation, it is also painfully wasteful. As yet, there is no efficient remedy to the death of knowledge. All we can do is to attach more weight to healthy lifestyle and health research. Those two promote longevity of knowledge in a single generation

Knowledge and forgetting: Forgetting is a natural process that makes it possible to efficiently use the limited memory space of the brain. We forget to dispose of knowledge deemed less important in order to make space for knowledge of higher importance. Currently we have only a limited control over what we remember and what we forget. Today, the most important tool that we can use to prevent forgetting is practice. We can minimize time needed for practice by using spaced repetition (i.e. learning technique based on computing optimum intervals between repetitions). Spaced repetition is the key to maximizing knowledge within a single human lifetime (see below for more)

Immortal knowledge: Artificial intelligence is our best hope for approaching immortal knowledge. It can nearly eliminate the problem of death (except for the heat death of the universe). It can also eliminate the problem of forgetting (at least within the bounds of the available storage). Today, however, the best path towards immortal knowledge must still rely on the use of the human brain by maximizing its learning capacity

Knowledge acquisition

There are five main areas where the learning process can be enhanced. All these areas will be discussed in this article:

  1. access to knowledge - this article will make an assumption that the Internet is the main source of knowledge
  2. selecting knowledge - we will assume that the reader is solely responsible for selecting knowledge. The ability to select valuable information grows in proportion to the acquired knowledge
  3. reading - reading is the process in which knowledge for the first time makes an actual intimate contact with the brain. Traditionally, it is streamed into memory in a more or less linear manner (i.e. paragraph after paragraph). This article shows how to delinearize this process and optimize reading by enhancing knowledge selection and prioritization concurrent with reading. For example, one should be able to say This paragraph can be processed later or This paragraph requires utmost attention now or This paragraph can be skipped for good even if I decide to read the article again, or I want to read this paragraph again in three days and in more detail or I want to mark this paragraph with lower-priority and come to it only after all higher priority paragraphs have been processed, etc. 
  4. representing knowledge - knowledge representation affects comprehension and retention. Things that are simple are easier to understand. Things that are simple are also easier to remember. Many people do not realize the degree to which simplicity can affect learning. Many people doubt that even the most complex material can be presented in a very simple way. Einstein noticed that "it should be possible to explain the laws of physics to a barmaid"
  5. remembering knowledge - we can eliminate the problem of forgetting once and for all. It can be done with the use of the learning technique based on the spaced repetition and commercially known as SuperMemo. As SuperMemo is currently the only software in existence that supports some of the discussed techniques, we will use it to illustrate practical applications of spaced repetition and incremental reading
  6. life cycle of knowledge - this article presents an approach to learning in which knowledge in human memory keeps on evolving and maturing. This involves continual rewording, reprioritizing, and re-associating pieces of knowledge. One will often need to give up portions of knowledge that become outdated or lose their high-priority status. Rules of knowledge representation should be used to make knowledge easy to remember. Knowledge should become more associative in time. In other words, it should become a more suitable ground for making intelligent choices 
  7. using knowledge - knowledge translates to value when it is properly used. In the long-run, skills discussed in this article should indirectly help one become more creative and skillful in using knowledge. Not surprisingly, skills needed to efficiently use knowledge are also part of knowledge itself and tend to grow spontaneously as knowledge itself increases

Access to knowledge

The Internet is an excellent source of knowledge. Its quality and role will for long continue increasing exponentially as more and more people appreciate its potential and contribute to its growth. There are still many complementary sources of information that compete successfully with the Internet. However, it is only a question of time before one is safely able to rely on the Internet as your sole source of information. 

The three main factors that limit the value of the Internet as the source of information today:

  • verification - there are no formalized peer-review or other verification mechanisms that would make it easy for you to ascertain the reliability of information available and trustworthiness of authors. The burden of proof is on you. You either need to tap onto reputable sites or judge the reliability of individual authors on preponderance of evidence. Consequently, reputable scientific journals are still an unmatched source of raw research data (cf. Pierre Salinger syndrome)
  • bandwidth - popular broadband access to the Internet is still to come. This limits availability of quality video documentaries, video reporting, video education, interactive material, etc. Consequently, television and multimedia titles have still a substantial role to play as an educational tool
  • micropayments - authors still see the printed matter as their main source of income. In the future, your credit account may be charged to a microscopic amount each time you access a selected page on the web if the author decides to make it a source of his or her income. This may ultimately cause a massive exodus of authors from the traditional publishing industry towards the net with immense benefit to those mining for quality data on the net

Despite the Internet's limitations, one can safely commit oneself today to making the net the chief source of knowledge in the quest for new knowledge in nearly all areas of well-rounded self-instruction. The number of free encyclopedias, journals, dictionaries, databases, stand-alone articles and thematic websites is increasing daily. 

This article will discuss efficient ways of accessing, prioritizing and reviewing information. It will also introduce a software application that implements some of the discussed techniques: SuperMemo 2002 for Windows. The most tangible value of SuperMemo comes from ensuring high retention of knowledge through spaced repetition. SuperMemo also makes it possible to increase the quality of reading by means of a technique called: incremental reading. In addition to its main strengths, SuperMemo also serves with an assistance in the field of knowledge access. We always face the need to fill the gaps in our knowledge in many more areas than the time permits or memory makes feasible. SuperMemo can scrupulously store and prioritize all areas of knowledge that need an enhancement! It uses a concept called a tasklist. A tasklist is a list of tasks sorted by the Value/Time ratio. The table below shows an exemplary knowledge-harvesting tasklist as exported from SuperMemo:

Tasklist: Knowledge (956 tasks)
Generated with: SuperMemo 2002 Build: 11.0 of May 1, 2002
Date: 01-May-02 3:25:43 PM
Source path: c:\sm2000\systems\all\tasks\MyKnowledge
  1. 20.0/0.16=125.00 - Calculus
  2. 18.0/0.17=105.88 - B.F. Skinner
  3. 30.0/0.32=93.75 - String theory
  4. 10.0/0.12=83.33 - Chomsky
  5. 10.0/0.12=83.33 - Laplace
  6. 9.0/0.11=81.82 - Leibnitz
  7. 12.0/0.15=80.00 - Napoleonic wars
  8. 10.0/0.13=76.92 - Koran
  9. 9.0/0.12=75.00 - Carl Linneaus
  10. 9.0/0.12=75.00 - Yasir Arafat and Yitzhak Rabin
  11. 14.0/0.20=70.00 - Copernicus
  12. 9.0/0.13=69.23 - Presenilin genes
  13. 9.0/0.13=69.23 - Gutenberg
  14. 9.0/0.14=64.29 - Kashmir
  15. 6.0/0.10=60.00 - Polgar sisters

etc.

In the tasklist above, you can see a list of task which all have a form of a small well-defined topic to study. Each time you add a task to your list, you must define how valuable it is (e.g. assign it a dollar value by answering the question How much would I pay for this knowledge if I was to transfer it painlessly to my memory at this moment?). Each time you add a task to your list, you should also approximate the time necessary to collect relevant information. This will give preference to encyclopedic topics that can be found in seconds in on-line encyclopedias such as Encarta (e.g. Carl von Linn or Old Testament). On the other hand, this will require more time for more esoteric topics such as cross-comparison of US state laws on dimpled ballot interpretation in manual under-vote recounts. In the exemplary tasklist above, the pink column specifies the dollar value of individual topics, the blue column indicates the access time in hours and the red column shows the Value/Time ratio. For example, the topic called Calculus is worth $20 and is supposed to take 0.16 hours to investigate (here this is the time needed to import Calculus to SuperMemo). This results in $125/hour valuation and brings Calculus as the most valuable topic on the tasklist

Selecting knowledge

The extent of global knowledge resources can be measured in terabytes. One terabyte is a thousand gigabytes, while one gigabyte is a rough equivalent of the Encyclopaedia Britannica whose 44 million words would most likely all fit your today's hard disk space. The US Library of Congress is estimated to amount to 25 terabytes of knowledge. Human DNA code kept at Celera takes 80 terabytes. At the same time, the so-called deep web (year 2000) may encompass as much as 7,500 terabytes (deep web includes the static web extended by dynamic information retrievable from web databases). It will take only three years from today to produce more data than in the whole of human history (say researchers at the University of California).

Only a fraction of those resources can be mastered by an individual in a single human lifetime. Even a single copy of the Encyclopaedia Britannica goes in detail far beyond what a single human being can encompass in a lifetime! 

The microscopic capacity of the human brain has not prevented it from building the present civilization as we know it. The human power comes from: 

  1. collective effort - a billion heads is more than one
  2. specialization of labor - all collective tasks are subject to top-down functional decomposition and a single brain usually only needs to process a fraction of information at a time 
  3. knowledge selection skills - the associative power of the human brain combined with the selective nature of forgetting help us retain memories that are actually most useful in problem solving

This article will show you how to prevent forgetting. However, forgetting plays an important role in our lives. It runs a valuable garbage collection on knowledge we acquire daily. If the power of forgetting is taken away, your responsibilities in the area of selecting knowledge increase manifold. The technique of spaced repetition can help one eliminate forgetting! At the same time, it increases the responsibility for selecting knowledge that is truly important and applicable. If used without care and attention, spaced repetition may actually be a waste of time by helping one remember a great deal of trivia.

A piece of information that occupies just several bytes of your hard disk may carry a relative value that my translate to a net gain of millions of dollars as well as a net loss of millions of dollars. It may also carry no value whatsoever. For example, a sentence written in French "SuperMemo vous aide mmoriser et apprendre diverses informations comme une langue, des chiffres, etc." may be of nearly zero value for someone who does not know French. At the same time, an item related to a Heimlich maneuver can save the life of a family member. We know that the expected payoff equals the value of the payoff multiplied by its probability; therefore, the low probability of a family member choking and the probability of actual successful application of the maneuver make the value of "Heimlich item" a fraction of the value of the human life. At the same time, even minor errors in medical knowledge of a physician can actual cost somebody's life and carry substantial negative value! Frequently, we find more benefit in memorizing the three best things learnt on a given day than in memorizing a whole monothematic article to the last detail!  It is possible to store the mind with a million facts and still be entirely uneducated (Alec Bourne).

I have mentioned earlier that tasklists can help you prioritize knowledge areas and topics that you want to study. In a similar way, tasklists can help you prioritize articles you locate on the net and decide to study in detail. In prioritizing study subjects, your tasklist contains only general names, e.g. nanotechnology, GM-foods, neural networks, American Civil War, etc. In prioritizing articles, your tasklist will contain whole articles. We will use the term reading list to describe a tasklist containing articles. 

A reading list is a list of articles sorted by the ratio: Value/Time, where Value is the estimated dollar value of knowledge contributed by the article and Time is the expected time needed for reading the article. SuperMemo can automatically estimate the Time field for you on the basis of text length. You can then modify Time manually. The basic value of a reading list is: you can import as many articles as you want (and as many as you hope you are able to read) and then only read as many articles as you are actually able to read. Your time will be spent optimally as you will always start with articles of the highest priority. You will quickly discover that you usually overestimate your reading abilities by one or two orders of magnitude! It is not unusual to want to read hundred times more articles than you are able to read! Using reading lists one can be sure that the small fraction that one is able to read is indeed populated by articles of the highest value! 

Blue inserts show how the discussed techniques are implemented in SuperMemo 2002. These inserts are not needed to understand the text
SuperMemo 2002: prioritizing learning material with reading lists
  1. Locate an article of interest on the Internet
  2. In your web browser, copy the text to read and remember
  3. In SuperMemo, choose Ctrl+Alt+R. You will see a dialog box similar to the picture below
  4. Type in the dollar Value of the article (you can use your own national currency or any other uniform measure of value). Choose the value by responding to the question: How much will reading this particular article contribute to my knowledge? How much would I be ready to pay for this knowledge?

Your articles will automatically be sorted for Priority (which is equal to Value divided by Time). You can begin reading articles from the top of your list

Reading

Traditional linear reading is highly inefficient. This comes from the fact that various pieces of the text are of various importance. Some should be skipped. Others should be read in the first order of priority. Old-fashioned books are quickly being replaced with hypertext. Hypertext will help you quickly jump to information that is the most important at any given moment. Hypertext requires a different style of writing. After all, all texts will live by the assumption that there has been some introductory section read before. The texts become context-independent, and all difficult terms and concepts are explained solely with additional hyperlinks. As the world wide web helped delinearize the global resources of information, the technique of incremental reading can help you delinearize your reading of whatever linear material you decide to read. While reading incrementally, you will see a linear text as a sequence of sections subdivided into paragraphs and individual sentences. Incremental reading will help you provide a separate and independent processing for each section, paragraph or sentence. Here are some typical actions you will execute on individual sections, paragraphs or sentences. Most typically you will decide one of the following:

  • read and remember - you can read a paragraph and schedule it for a future review. The timing of the review will depend on the paragraph's priority
  • skip to read later - you can mark a paragraph as worth reading later. The moment in which the paragraph will come back for reading will depend on the paragraph's priority. 
  • skip for ever - you can mark a paragraph as (1) not worth reading, or as (2) read and not worth remembering 

In incremental reading, all material worth remembering will be scheduled for review in the future. This is necessary to ensure you do not forget what you have learned. At the same time, all material that has not been read but is classified as worth reading will also be scheduled for reading in the future. In incremental reading, you will constantly face a serial process of reading, review, and repetition that will make sure that your work with the program is challenging, interesting and leaves permanent traces in your memory.

Reading with SuperMemo 2002

Here are the typical actions you will perform on individual sections, paragraphs and sentences:

  1. Read and remember - you can read a paragraph, select it with the mouse and choose Remember extract. Remember extract will create a new mini-article in SuperMemo and store it for review in a couple of days. The review of the extract will be subject to the same procedures as reading the entire article. In other words, you can only quickly read the selected paragraph to get the general idea and read it in detail only upon the first review

  2. Skip and read later - if you see that a longer section is important and requires more detailed reading, you can execute Remember extract without reading the section at all

  3. Skip for ever - if you believe that a given part of text is not worth your time, you can choose Ignore. This will mark the text with a grayed font, and ensure you will not get into reading the same text again in the future (e.g. on the next review)

If you believe the whole article should be skipped, you can do the following:

  1. Skip and read later - you will delay reading an article if you believe its priority is less or if you think that you need to read some other introductory articles first. For example, you can stop reading and manually choose an interval after which you will return to reading

  2. Skip and retain in the archive - if you do not expect to return to a given article but you still want to retain it in SuperMemo, you can choose Dismiss

  3. Skip and delete partially - if you believe an article is not worth keeping in your SuperMemo collection, you can delete it without deleting review material generated from the article

  4. Skip and delete entirely - if you believe an article is not worth keeping in your SuperMemo collection, you can delete it along with all associated extracts, sections and paragraphs

Representing knowledge

In incremental reading, all articles evolve. This evolution ensures the maximum comprehension and retention of knowledge. Initially the articles are split into sections and paragraphs. Those sections and paragraphs are later subject to regular review and further evolution. Individual paragraphs get enriched with context clues, reference labels, and converted to individual sentences. Individual sentences convey ideas which you want to remember. These ideas come back for review at increasing intervals. Initially, the ideas come back every couple of days, later after months and years. However, passively processed ideas in the form of sentences rarely leave a durable trace in your memory even if they are reviewed regularly. Very often, as soon as after 2-3 months, you will notice that at review time, you actually do not seem able to recall that you have ever had a given sentence in your collection. You will quickly discover that you need active recall in order to remember. Active recall is a process in which you must answer questions. For example, you may be presented with a picture of Charles Darwin and be asked to recognize his face. In the long run, you need to replace passive review with active recall. Otherwise, your memory of the fact will not be permanently consolidated. 

The fastest way of converting simple sentences into active recall material is to use a cloze deletion technique. In the cloze deletion technique, you convert simple declarative sentences like:

WW1 was precipitated by the assassination of Archduke Francis Ferdinand of Austria-Hungary by a Serbian nationalist in 1914

into question-answer pairs that can be used in actively stimulating your memory for best recall:

Question: WW1 was precipitated by the assassination of Archduke Francis Ferdinand of [...](country/empire) by a Serbian nationalist in 1914
Answer:
Austria-Hungary

Question: WW1 was precipitated by the assassination of Archduke Francis Ferdinand of Austria-Hungary by a Serbian nationalist in [...](year)
Answer:
1914

Question: [...](war) was precipitated by the assassination of Archduke Francis Ferdinand of Austria-Hungary by a Serbian nationalist in 1914
Answer:
WW1

Active recall in SuperMemo 2002

To convert a sentence to a cloze deletion in SuperMemo 2002, select the appropriate keyword in the sentence (e.g. Austria-Hungary, 1914, WW1, etc.) and choose Remember cloze.

Tip! Try to avoid using cloze deletion tools on conglomerate paragraphs. Your cloze deletions should be as simple as possible. Consequently, simplifying the parent paragraph to a simple statement will produce simple clozes that will require little processing. If you use Remember cloze on a longer multi-sentence paragraph, you will have to put extra effort on simplifying the resulting items. All cloze deletions should be short enough to ensure you read them entirely at repetition time. Otherwise, your brain will tend to "deduce" the answer from non-semantic clues. This will defeat the purpose of learning!

The active recall issue in representing knowledge is just a tip of the iceberg. You will need to master quite a number of skills that will ensure your knowledge is:

  • easy to remember

  • quick to process

  • highly applicable

You will hone those skills gradually in time. Your own mistakes will provide the best material for improvement.

Remembering knowledge

Forgetting has been the number one problem in learning for ages. All capable students know that it is not really hard to cram hundreds of pages of material before an exam. But a great deal of the learned knowledge is gone just 3-4 days later (esp. after an exhaustive all-nighter). 50% of the quickly crammed knowledge evaporates within a week or two. After a year or so, nearly all material is forgotten unless reviewed. It is easy to notice that the repetition is the key to remembering in the long run. Repetitio mater memoriae may date back to Horace (65-8 BC).

Spaced repetition and the review of the learned material is the key to efficient learning with high retention. The review algorithm has started the development of SuperMemo in the early 1980s and the first software implementation of SuperMemo in 1987 was based solely on repetitions of simple questions and answers.

Spaced repetition makes it possible to remember, by default, 95% of the learned material. You can program this retention level to fall between 90-99%. You can also determine the retention individually for each of the pieces of information subject to review. You can also use tools that will make it possible to substantially increase the flow of information into the learning process at the cost of retention. 

Remembering with SuperMemo 2002

All material introduced to SuperMemo 2002 will be subject to review. This is all you need to remember:

  • on a daily basis choose Learn and go through the scheduled portion of the material (this will include whole articles, their fragments scheduled for review, cloze deletions and simple question-answer items)
  • if you proceed with your daily review until you get the message Nothing more to learn, you are guaranteed to remember, by default, 95% of the learned material

You will often introduce to SuperMemo more material than you are able to review. You will therefore also need to consider the following:

  • the proportion of the material you remember can be changed globally or for individual pieces of information
  • if your review takes too much time, you can slow down the process by which material stored in SuperMemo is actually introduced into the learning process by answering No to Do you want to learn new material? This will make excess material wait for learning until you master the material that has already entered the learning process
  • if you cannot keep up with review, you can choose rescheduling tools

Life cycle of knowledge

Knowledge subject to incremental reading will gradually be transformed and reformulated. This will also reflect changes to the corresponding knowledge in the reader's memory. Three main principles will underlie the evolution of knowledge in incremental reading:

  • decrease in complexity - articles will be converted into sets of paragraphs. Paragraphs will be dismantled into sets of independent sentences and statements. Sentences will be shortened to maximize the contents-vs-wording ratio, etc.

  • active recall - all pieces of information will ultimately be converted into active recall material such as question-answer pairs, cloze deletions, picture recognition tests, sound recognition tests, etc.

  • incrementalism - all changes will take place gradually in proportion to available time, with respect to material priority, and in line with the gradually increasing strength of memory traces

Success in learning is based on review and repetition. Changes to individual pieces of knowledge will take place in steps upon successive reviews. Here are exemplary steps that show a complete evolution of a single article into a finished question based on active recall:

  1. Imagine that you find an article on the net, e.g. The criticism of global capitalism, and you decide to read it and remember its key points for ever

  2. You import the article into the incremental reading process

  3. You read the article (e.g. once it tops your reading list or once its turn comes up in incremental reading)

  4. While reading, you extract most important paragraphs. One of these, let us say, refers to Kuznets hypothesis

  5. The extracted paragraphs will assume a separate life in incremental reading and will be scheduled for separate review, i.e. independent of the review of the parent article. The extracted paragraphs in the parent article will be marked as processed. Once all paragraphs in the parent article are processed, you will terminate the review of the parent article and keep on reviewing only its components (e.g. selected paragraphs)

  6. Upon the first review, usually after a few days, you read the extracted paragraph again and analyze it as to how it should be processed further. You may decide to postpone it, remove it from the learning process, shorten it or extract the most important sentences that you want to remember

  7. If you decide to extract a single statement in reference to Kuznets hypothesis it will again be marked as processed in the original extract and will assume a separate review cycle in incremental reading

  8. Upon the first review of the extracted sentence, you make further decisions as to its further life. Let us say, this is the wording of the Kuznets sentence: Acc to Kuznets hypothesis, growth (from the low income levels associated with predominantly agrarian societies) would first lead to an increase, and then to a decrease in income inequality

  9. In order to capture the essence, you would probably decide to shorten the above sentence to the following form: Acc to Kuznets hypothesis, growth would first lead to an increase, and then to a decrease in income inequality

  10. At the same time, other parts of the same parent article might establish a memory trace that would say that Kuznets hypothesis has been based on relatively weak empirical data. Moreover, recent research clearly indicates that the hypothesis is false (growth actually seems to equally benefit both the poor and the rich). You could then enhance the extract with words controversial or even recently falsified. For example, Recently falsified Kuznets hypothesis claimed that growth would first lead to an increase, and then to a decrease in income inequality

  11. Upon the next review of the same sentence, you may decide to convert it into a number of cloze deletions. This conversion will be incremental, i.e. you may decide to first create an cloze deletion asking about the name of the controversial hypothesis and only later ask about its actual meaning (the meaning is relatively easier to remember and shall survive longer in your memory without active recall). Your cloze deletion could then look like this:

    Question: Recently falsified [...](name) hypothesis claimed that growth would first lead to an increase, and then to a decrease in income inequality
    Answer: Kuznets 

    This cloze deletion would again assume a separate life from the original sentence in which the keyword Kuznets will again be marked as processed. This is the original Kuznets sentence with one keyword marked as processed: Recently falsified Kuznets hypothesis claimed that growth would first lead to an increase, and then to a decrease in income inequality 

  12. The same sentence will generate a few separate cloze deletions that will be processed independently. Upon the first review of the cloze deletion created in the previous point, you may decide to simplify it in accordance with the rules of formulating knowledge in learning: 

    Question: Recently falsified [...](name) hypothesis claimed that growth would first lead to an increase in income inequality
    Answer: Kuznets  

  13. Upon the next review, you can, but you do not have to, convert the cloze deletion into a standard question-answer item:

    Question: What is the name of the hypothesis that falsely claims that income inequality initially increases with growth?
    Answer: Kuznets hypothesis 

  14. The above question-answer pair is probably as simple as it can only be. Certainly, it is simple enough to be relatively easy to remember. This item will be repeated in intervals determined by spaced repetition. You can decide how well you want to remember it. In SuperMemo, by default, it would be remembered with 95% probability of recall and require 5-15 repetitions in lifetime. The establishment of durable memory traces in your memory, completes the life cycle of this particular piece of knowledge. The only thing that remains is the memory-sustaining review in intervals ranging from months to years (as determined by spaced repetition)

  15. Once you convert all important keywords from the Kuznets hypothesis into separate cloze deletions, you will remove the original declarative statement with the hypothesis from the review process. You will continue repeating individual clozes and that will ensure your perfect recall of the hypothesis for as long as you deem necessary

Incremental reading

If you stop reading an article and return to it after a month or so, you may find it hard to understand the text due to the fact that the part read earlier will have already been partly forgotten while it might have built a solid introduction needed for understanding the remaining sections of the article.

In incremental reading, reading articles in small parts does not pose a problem. The parts read earlier enter the regular review process that ensures continual retention of all pivotal points. Even more, incremental reading encourages reading in small parts. This makes it possible to read many articles at the same time, and portion the inflow of information into the learning process in proportion to priority and available time.

Without spaced repetition, incremental reading would not be possible. This comes from the fact that only spaced repetition can ensure that you do not forget what you have read before. In other words, if you return to the same article in a week or in a few months, you will quickly recover the context and keep on reading as if you have read preceding parts just minutes earlier.

The advantages of incremental reading are numerous. All these advantages are interdependent. Good retention enables incremental reading. This provides variety which eliminates boredom, enhances attention and affects creative associations. Good prioritization enhances the speed of reading without affecting comprehension, and that in turn, eliminates the stress of fast reading (worry of missing important pieces of information) and the stress of massive reading (worry of losing track of one or more articles). 

Here is the list of the most prominent advantages of incremental reading: 

  1. knowledge structure and comprehension: Building knowledge in your brain is like assembling a jigsaw puzzle. Some pieces cannot be placed in the puzzle before the others. Some pieces capitalize on others. There is no point in memorizing facts about Higgs boson before you learn what a standard model is and that, in turn, should follow the general understanding of particle physics which itself requires some ABC of physics. In incremental reading, if you encounter texts related to Higgs boson you can manually delay it until the time you hope your Physics ABC will provide the ground for understanding. In traditional reading, you would just waste your time on reviewing Higgs boson material. Traditionally, your decision to skip the material would provide no definite way of coming back to material in the future. With incremental reading, you waste no time on reading material you do not understand. At the same time, you can safely skip portions of material and return to them in the future. You become the master of the conscious knowledge building process

  2. attention: Human brain has an in-built limit on the attention span. We all get bored with things. This is often true even with interesting articles once they get too long. Millions of people do a daily channel zapping on TV. This absurd activity is driven precisely by the craving for dense action and information variety. A gripping movie goes "too slow" for a typical channel zapper. This is why he or she prefers to watch three movies at the same time (even though the plot of all will suffer). Incremental reading is a perfect remedy to the limited attention span. Once you sense any sign of boredom or distraction, you can jump to the next article with few negative side effects. Unlike in the case of channel zapping though, you won't miss a bit of information. Just the opposite, you will maximize attention per paragraph. Your attention to the same piece of information may depend on your mood, amount of prior reading, today's interest that may depend on the piece of news you heard on the morning radio, etc. With incremental reading, you can fit your best attention to each individual piece of reading

  3. speed: Incremental reading will help you dramatically increase the speed of your first reading of a selected text. You can quickly jump from paragraph to paragraph, get the overall picture, mark fragments for later reading, mark fragments for detailed study, etc. You will be relieved of the greatest bottleneck of speed-reading: fear of missing important pieces of information. As soon as you suspect a quickly-reviewed paragraph may carry more importance than meets they eye, you can simply introduce it back into the review process (as opposed to backtracking and reading the paragraph again). Once you process the entire article, you can slowly digest it again from the very beginning in the incremental reading process. Incremental reading provides for speed-reading without detriment to comprehension  

  4. prioritization: There are always many more articles at hand than you can hope to read. Evaluating articles and prioritizing them is difficult because you cannot do a good evaluation without actually reading a part of the article in question. In incremental reading, you can read the introduction and then decide when to read the rest. If an article is extremely valuable or interesting, you can process it entirely at once. Other articles can slowly scramble through the learning process. Yet others may ultimately be deleted. The prioritization will continue while you are reading the article. If the evaluation of quality or content changes while reading, so will the reading-review schedule. Prioritization tools will ensure that important pieces of information will receive better processing. This is one of the most important things about incremental reading: efficient fishing for pieces of golden knowledge!

  5. consolidation: Everything we learn must be reviewed from time to time in order to be remembered. If you read an article in intervals, you already begin the consolidation of memory which may save you lots of time. In traditional reading, you would need to read the whole article, and then to review the article later several times. With incremental reading, you can begin the consolidation-review cycle already during reading! By the time you convert parts of the material into clozes or question-answer items, you will already have it well consolidated. This pre-consolidation will often dramatically reduce the number of repetitions required before your material gets to be reviewed in long intervals of months and years

  6. creativity: spaced repetition will throw at you various articles, paragraphs, statements and questions in a most unexpected order. You will be surprised to discover how this affects your creativity and helps you generate unexpected associations of ideas. This will also provide your brain with an entertaining form of mental training that will be highly appreciated in all forms of professions based on intellectual performance. Critics of incremental reading, who have never tried the method, are not convinced this argument is really tenable. You may be skeptical too. There is only one sure way to convince you: give incremental reading a good try! Try incremental reading for a month at times when your brain is in its peak shape. If the processed knowledge belongs to the areas that are critically important to you, so much the better

  7. stresslessness: Once you know you can rely on spaced repetition in presenting review material for you, you can eliminate the stress and anxiety related to having too much to study or too much to read. You will probably not manage to read or learn all that you would hope for, but you will at least not lose sleep over planning and scheduling. You can devote all your energy to comprehension and analysis of the learned material

  8. consistency: If your material contains contradictory parts, your brain will alert you to this fact. In classical learning, you would often relearn new facts that would contradict earlier learned facts. Then you would relearn the older version again and this wasteful cycle might repeat more than once. In incremental reading, the same process can take place; however, there will be two mechanisms that will prevent it. High degree of retention guaranteed by spaced repetition will often make it quite effortless to immediately spot the contradiction: Wait a minute! I have already learned this fact and the answer was different! Unfortunately, even incremental reading isn't hermetic to contradiction (your retention actually never reaches 100%). The second mechanism will quickly produce the convergence of contradictory facts in time. If you, for example, learn two different answers to What is the size of human population?, say, 5.5 billion and 6 billion. You will naturally provide a wrong answer to one of these questions. Once you relearn it the new way, you will provide a wrong answer to the other question. Inter-repetition intervals for these two contradictory items will get shorter with each relearning cycle. The timing of repetitions of contradictory items will tend to converge. It is only a question of time when the red alert is raised by your brain. You will quickly resolve the difference and delete one of the items. Similar process will affect hazy or incompletely specified information. Your knowledge will grow in consistency with time

  9. massive learning: Massive learning is the consequence of all advantages listed earlier. Incremental reading will run a wide stream of knowledge through your memory. Even though human memory is painfully limited, you shall be able to read thousands of texts pertaining to different fields and quickly swell the size of your knowledge to your and others' benefit

  10. enjoyability: Those who can compare the classic learning with incremental reading will testify that incremental reading is by far more enjoyable. Monotonous repetitions are here interspersed with reading and analyzing new material

Incremental reading tools in SuperMemo 2002

To introduce an article to incremental reading - use Edit : Add a new article. This will introduce the article stored in the clipboard into the incremental reading process

To resume reading - once you return to an interrupted article, the cursor in the text will be set on the last-processed text. When leaving the article, you can also manually set the so-called read-point and the place where you interrupted reading

To schedule the next review - if you want to schedule a given article for review on a selected day, choose Reschedule

To add pictures - to add a picture to an article use Copy on the picture menu in your web browser and then Paste in SuperMemo

Using knowledge

I am not aware of the tools that could effectively help you use your knowledge for good causes. Luckily, as the knowledge grows, so does the ability to use it efficiently. Incremental reading can boost your knowledge but you must consciously avoid spending time on gaining knowledge for the knowledge sake. Applicability should remain the ultimate goal of learning

1.3.21