TL;DR. Your learning time is tiny and fixed; the supply of things to learn is huge and growing. In the 2025 American Time Use Survey, adults aged 35 to 44 averaged 0.09 hours a day on educational activities, about 5 minutes, or roughly 33 hours a year. Worldwide output of science and engineering articles rose from 2.0 million in 2010 to 3.3 million in 2022 (National Science Board), and a long-run study of publication databases puts the growth of science since 1952 at 5.08 percent a year, a doubling every 14 years. Psychologists Ralph Hertwig and Christoph Engel call the conscious choice not to seek or use information “deliberate ignorance” and show that it often has useful functions. The CEO move: treat every new topic as an investment with an opportunity cost, and write down what you will not learn this quarter. The student move: protect the few things you learn deeply, because unused knowledge fades fast, and learn just enough of the rest to judge the people and AI tools you delegate it to.
This piece looks at learning from the strategy side: not how to learn, but what to leave out. For the selection side, pair it with the 4-factor decision matrix for your next skill. For the information diet that feeds the decision, see the signal-to-noise crisis and the attention portfolio.
Why is “what not to learn” now a strategy question?
Most career advice about learning is additive: one more course, one more tool. The constraint is old. In a 1971 lecture, Herbert Simon, then a professor of computer science and psychology at Carnegie-Mellon University, described the problem in one line: what information consumes “is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention.”
Economics later formalised the idea. Christopher Sims’s work on rational inattention, first published in the Journal of Monetary Economics in 2003, models slow reactions to news as “an inability to attend to all the information available”, treated as a finite information-processing capacity. Ignoring most information is then not a failure of rationality but a necessity. The open question is whether the ignoring is chosen or accidental.
Hertwig and Engel make the trade explicit in their 2016 article in Perspectives on Psychological Science: “the choice to know one fact invariably implies not knowing other facts.” They add that “the ability to select a few valuable pieces of information and deliberately ignore others may become a core cultural competence to be taught in school like reading and writing.”
That is the CEO and student lens in one sentence. A chief executive decides what to know personally, what to have summarised and what to leave to others. A good student reads the canon closely and skims the rest. The same discipline now applies to anyone whose work is touched by AI.
How much is there to learn, and how much time do you actually have?
Start with the supply side. The figures below are as printed in their sources.
| Measure | Figure | Source |
|---|---|---|
| Worldwide science and engineering articles, 2010 | 2.0 million | National Science Board, Indicators (Scopus data) |
| Worldwide science and engineering articles, 2022 | 3.3 million | National Science Board, Indicators (Scopus data) |
| Growth of science across four databases, whole series | 4.10% a year, doubling every 17.3 years | Bornmann, Haunschild and Mutz (2021) |
| Growth of science since 1952 | 5.08% a year, doubling every 14.0 years | Bornmann, Haunschild and Mutz (2021) |
Sources: National Science Board and NCSES, Publications Output: U.S. Trends and International Comparisons (NSB-2023-33); Bornmann, Haunschild and Mutz, Humanities and Social Sciences Communications, 2021.
Scientific articles are only one stream; add software releases, AI model updates, regulation and industry reports, and the total is unmeasurable. The point is not that you should read papers. It is that even one well-indexed stream of new knowledge grows faster than anyone can follow.
Now the demand side: how much time adults actually spend learning. The US Bureau of Labor Statistics measures this directly in the American Time Use Survey. The figures are averages per day across everyone in each age group, including the many people who did no learning at all that day.
| Age group | Educational activities (hours per day) | Reading, as leisure (hours per day) |
|---|---|---|
| 25 to 34 | 0.42 | 0.17 |
| 35 to 44 | 0.09 | 0.22 |
| 45 to 54 | 0.03 | 0.21 |
| 55 to 64 | 0.04 | 0.21 |
Source: U.S. Bureau of Labor Statistics, American Time Use Survey, 2025 annual averages, Table 3 (educational activities) and Table 11A (reading, a leisure activity). Released 25 June 2026.
Across everyone aged 15 and over, only 8.5 percent did any educational activity on a given day, and those who did spent 5.17 hours on average: learning is concentrated in students. Full-time workers spent 3.99 hours a day on leisure, 0.17 of it reading.
In the EU, 13.7 percent of adults aged 25 to 64 took part in education or training in the four weeks before the 2025 survey, up from 10.8 percent in 2019, and 46.6 percent did so over twelve months in the 2022 Adult Education Survey (Eurostat). More people are learning, in small amounts.
The table below converts the time-use averages into annual hours and sets them against the article count, as an illustration of scale.
| Age group | Educational activities, hours per year | Plus reading, hours per year | In 40-hour work weeks (combined) | New S&E articles per hour of yearly educational time |
|---|---|---|---|---|
| 25 to 34 | 153.3 | 215.3 | 5.38 | about 21,500 |
| 35 to 44 | 32.9 | 113.2 | 2.83 | about 100,500 |
| 45 to 54 | 10.9 | 87.6 | 2.19 | about 301,400 |
| 55 to 64 | 14.6 | 91.2 | 2.28 | about 226,000 |
CEOtudent analysis of BLS American Time Use Survey 2025 (Tables 3 and 11A) and National Science Board publication counts. Annual hours = daily average x 365; articles per hour = 3.3 million / annual educational hours. Population averages, so individual learners will differ widely.
The National Science Board counts imply compound growth of about 4.26 percent a year between 2010 and 2022, close to the long-run estimates above, or about 377 new articles every hour.
The practical reading: a mid-career professional with an average schedule has two to three working weeks of learning and reading time a year. Add a demanding certification and something else has to come out. The question is whether you choose what comes out.
Is choosing not to know rational, or just avoidance?
Not wanting to know is common. Gerd Gigerenzer and Rocio Garcia-Retamero ran the first representative national studies of deliberate ignorance, with 1,016 adults in Germany and 1,002 in Spain, and published the results in Psychological Review in 2017. They asked about ten future events, five negative and five positive.
| Event (would you want to know in advance?) | Share who would not want to know |
|---|---|
| When your partner will die | 90.2% |
| From what cause your partner will die | 89.4% |
| When you will die | 86.7% |
| From what cause you will die | 85.8% |
| Whether your marriage will end in divorce | 86.2% |
| The result of a recorded football match | 73.6% |
| What you are getting for Christmas | 64.4% |
| Whether there is life after death | 50.4% |
| Whether a 2,000 euro sapphire is genuine (test costs 50 euros) | 51.9% |
| The sex of your child before birth | 37.4% |
| All ten events | 71.6% |
Source: Gigerenzer and Garcia-Retamero (2017), Psychological Review 124(2), Table 7, total row across Germany and Spain (N = 2,018).
The authors summarise the pattern as “Between 85% and 90% of people would not want to know about upcoming negative events, and 40% to 70% prefer to remain ignorant of positive events. Only 1% of participants consistently wanted to know.” Their explanation is anticipatory regret: people avoid information whose answer they expect to regret knowing. People who preferred not to know were also more risk averse and more likely to buy life and legal insurance.
This matters in two ways. A deliberate not-to-learn list is not exotic; almost nobody wants to know everything. But much of this avoidance is emotional, and the comfortable version of strategic ignorance is skipping the topic that threatens you: the AI tool that automates part of your job, the feedback you would rather not hear. That is avoidance dressed up as focus.
Hertwig and Engel’s taxonomy helps separate the two. They list six functions of deliberate ignorance: emotion regulation and regret avoidance; maximising suspense and surprise; enhancing performance; serving as a strategic device; protecting impartiality and fairness; and what they call cognitive sustainability and information management. Only some of these are good reasons for a learning decision.
| Function (Hertwig and Engel, 2016) | What it looks like in a career | Use it for learning choices? |
|---|---|---|
| Cognitive sustainability and information management | Not following every AI model release; reading a monthly digest instead of daily feeds | Yes. This is the core of strategic ignorance |
| Performance enhancing | Not checking metrics or rankings hourly while doing deep work | Yes, with a set review date |
| Strategic device (including self-discipline) | Not learning a tool so you cannot be pulled into maintaining it; removing an app you overuse | Sometimes, if the trade is explicit |
| Impartiality and fairness | Blind review of candidates or ideas; not looking up who wrote a proposal | Yes, in decisions about others |
| Suspense and surprise | Avoiding spoilers | Not relevant to work learning |
| Emotion regulation and regret avoidance | Not looking at the skill that threatens your role | Rarely. Test it against the limits below |
CEOtudent analysis: the six functions are from Hertwig and Engel (2016), Figure 1; the career examples and the right-hand column are a CEOtudent editorial framework.
Why does learning you do not use fade?
A second argument for choosing what not to learn is that most learning you do not use decays. In 2015, Jaap Murre and Joeri Dros replicated Hermann Ebbinghaus’s forgetting experiment from 1880. One subject spent 70 hours learning lists of nonsense syllables and relearning them after intervals from 20 minutes to 31 days. “Savings” measures how much faster a list is relearned the second time; higher means more was retained.
| Interval after learning | Savings, Ebbinghaus (1880s) | Savings, Dros (2015 replication) |
|---|---|---|
| 20 minutes | 0.582 | 0.472 |
| 1 hour | 0.442 | 0.373 |
| 9 hours | 0.358 | 0.276 |
| 1 day | 0.337 | 0.317 |
| 2 days | 0.278 | 0.230 |
| 6 days | 0.254 | 0.168 |
| 31 days | 0.211 | 0.041 |
Source: Murre and Dros (2015), PLOS ONE 10(7), Table 3. Single subjects learning meaningless syllables; treat as a demonstration of shape, not a forecast for meaningful skills.
By 31 days, Ebbinghaus’s savings had fallen to about 36 percent of the 20-minute value, and the replication’s to about 9 percent (CEOtudent calculation from Table 3). Meaningful, practised skills behave better than nonsense syllables, but the direction holds. A meta-analysis by Winfred Arthur and colleagues in Human Performance (1998), pooling 189 data points from 53 articles, found skill loss ranging from an effect size of minus 0.01 immediately after training to minus 1.4 after more than 365 days without use. Physical, natural and speed-based tasks decayed less than cognitive, artificial and accuracy-based ones.
The implication: learning something “just in case” for next year is often a poor investment, because much of it will be gone when the need arrives. Learn it when you will use it, or learn only where to find it. For how quickly different abilities fade and how to schedule maintenance, see skill decay and the maintenance schedule.
Why do we keep saying yes to new topics?
Because the cost of a new topic is invisible at the moment of choice. Shane Frederick and colleagues studied this in the Journal of Consumer Research in 2009 under the name opportunity cost neglect. In one study, 150 students were asked whether they would buy a DVD for 14.99 dollars. When the option of not buying was described as “Keep the $14.99 for other purchases” instead of “Not buy”, willingness to buy fell from 75 percent to 55 percent, even though the two descriptions mean the same thing. In another study, describing the price difference between two iPods as cash left over raised the share choosing the cheaper model from 37 percent to 73 percent.
Time works the same way, and the trap is easier to fall into because learning feels virtuous. “Should I learn this new framework?” almost always gets a yes. “Should I learn it instead of finishing the statistics course I started?” often does not. Frame every learning decision as “this instead of that.”
The four-bucket test: learn deeply, learn to evaluate, delegate, or ignore
Here is the core tool. It sorts any topic, tool or skill into one of four buckets, using five criteria that the research above points to: how often you will use it, how costly an error would be, whether it compounds into other skills, how fast it goes out of date, and whether a reliable person or tool can do it for you in a way you can check.
| Bucket | What it means | Typical signs | Time budget |
|---|---|---|---|
| 1. Learn deeply | Build real, retrievable competence you practise regularly | Used weekly or more; errors are costly; it compounds (other skills build on it); slow to go out of date | Most of your learning hours; spaced practice |
| 2. Learn to evaluate | Know enough to judge quality, ask good questions and spot errors, without doing the work yourself | Used monthly; you buy, manage or review it; errors are costly but someone else does the work | A short primer plus a checklist; refresh yearly |
| 3. Delegate to AI or experts | Know who or what does it and how to check the result | Rare or one-off; a capable tool or specialist exists; output is checkable | Minutes: a source list and a verification step |
| 4. Deliberately ignore | Consciously decide not to follow it this period, and write it down | Low use, low stakes, fast-changing, no compounding, or pure novelty | Zero, until a review date or trigger |
CEOtudent editorial framework, built on Hertwig and Engel (2016) on information management, Murre and Dros (2015) and Arthur et al. (1998) on decay without use, and Frederick et al. (2009) on opportunity cost neglect.
Bucket 2 is the one most people skip, and the one AI has made most valuable: if you rely on an AI tool or a contractor, you need to judge their output even if you never do the work. That is a skill in its own right, covered in the evaluation skill: judging AI output. Delegating without it is not strategic ignorance; it is just ignorance.
The six-question decision test
Answer these for any topic you are tempted to learn. Score each question 0, 1 or 2.
- Frequency. Will you use this at least monthly in the next 12 months? (0 = no, 1 = a few times, 2 = weekly or more)
- Stakes. If you get it wrong, is the cost high or hard to reverse, for you or for others? (0 = minor, 1 = moderate, 2 = serious or irreversible)
- Compounding. Do other things you want to learn depend on it? (0 = no, 1 = somewhat, 2 = it is a foundation)
- Durability. Will what you learn still be valid in three years? (0 = likely obsolete, 1 = partly, 2 = mostly)
- Delegability. Can a person or AI tool do it well, in a way you can check? (2 = no, 1 = partly, 0 = yes, easily)
- Opportunity cost. Is it a better use of the hours than the best thing already on your list? (0 = no, 1 = about equal, 2 = clearly better)
Reading the score. 9 to 12: learn deeply (bucket 1). 6 to 8: learn to evaluate (bucket 2). 3 to 5: delegate (bucket 3), and make sure you can check the result. 0 to 2: deliberately ignore (bucket 4) and put it on your list with a review date.
Override rule. If question 2 scores 2 because of safety, legal or health consequences for you or others, the topic cannot go below bucket 2, whatever the total. See the limits section below.
CEOtudent editorial framework. The thresholds are a starting point for judgement, not a validated scale.
A not-to-learn list: the template
Writing the list down turns accidental neglect into a decision you can review, and answers the opportunity-cost question in advance.
| Topic or tool | Bucket | Why (which criteria) | What I do instead | Trigger to revisit | Review date |
|---|---|---|---|---|---|
| Example: every new AI model release note | 4. Ignore | Fast-changing, low compounding, high volume | Read one monthly summary from a source I trust | A model my team adopts | End of quarter |
| Example: a second programming language | 4. Ignore for now | Low frequency in my role; competes with statistics course | Finish the statistics course first | A project that requires it | Next quarter |
| Example: contract law basics for freelance work | 2. Evaluate | High stakes, low frequency | Short primer plus a lawyer for real contracts | A new client contract | Yearly |
| Example: spreadsheet modelling | 1. Learn deeply | Weekly use, compounds into analysis and forecasting | Spaced practice on real work | None, ongoing | Monthly check |
CEOtudent editorial template. The rows are illustrations, not recommendations for any particular role.
Two rules keep the list honest. Every “ignore” entry needs a trigger, so the decision is reversible. Every “delegate” entry needs a named check, so you can tell when the delegate is wrong. Ten entries is plenty.
The quarterly checklist
Use this once a quarter, in about 30 minutes.
- [ ] List everything you started learning, subscribed to or bookmarked “to learn” in the last three months.
- [ ] Run each through the six-question test and assign a bucket.
- [ ] For bucket 1, check that each item has regular practice scheduled; if not, it is decaying.
- [ ] For bucket 2, write or update a one-page checklist of what “good” looks like.
- [ ] For bucket 3, name the tool or person and the verification step.
- [ ] For bucket 4, add each item to the not-to-learn list with a trigger and a review date.
- [ ] Phrase your next learning choice as “this instead of that” and write both down.
- [ ] Check the list for avoidance: is anything there because it is uncomfortable rather than unimportant?
What this does not mean
The research that supports strategic ignorance also shows where it fails.
When others bear the risk. Hertwig and Engel note that “When ignoring information exposes others to risk (or imminent harm), Mill’s harm principle may be invoked.” If your work affects other people’s safety, money, health or rights, the duty to know overrides your preference. Safety procedures, data protection rules and the parts of your field where errors hurt clients are never in bucket 4.
When the law expects you to know. Hertwig and Engel describe how criminal law often requires proof that a defendant knew the facts, and how US courts use the so-called ostrich instruction, which tells juries they may treat a defendant’s wilful ignorance of the relevant facts as knowledge. Choosing not to know is a weak shield in professional and legal matters. Regulatory obligations relevant to your role belong in bucket 2 at minimum.
When it is your own health or money. Gigerenzer and Garcia-Retamero’s data show how strong the pull to avoid unpleasant personal information is. Their paper also cites earlier medical studies where refusal to test for diseases and infections typically ran between 10 percent and 30 percent. Avoiding a test or a bank statement because the answer might hurt is the emotion-regulation function, not information management.
When it is a compounding fundamental. Some skills look low-value in the short term but underlie everything else: clear writing, basic statistics, reading closely, the core concepts of your field. Their score on question 3 should dominate. Ignoring a fundamental because it is not urgent is the most expensive mistake this framework can produce. For why sustained reading is one of these, see deep reading as a competitive advantage.
When you delegate what you cannot evaluate. Handing a topic to an AI tool you cannot check moves the risk, it does not remove it. The decision test pushes high-stakes topics into bucket 2 for this reason.
When it is avoidance of change. If the topic you most want to ignore is the one most likely to change your job, look at it.
Frequently asked questions
What is strategic ignorance?
It is the deliberate, documented choice not to learn or follow something for a set period, because the time is better spent elsewhere. Psychologists Hertwig and Engel call the broader phenomenon deliberate ignorance: the conscious choice not to seek or use information.
Is it irrational not to want to know?
Not necessarily. With limited attention, ignoring most information is unavoidable. It becomes irrational when the ignored information is cheap to get and would change an important decision.
How much time do adults actually spend learning?
In the 2025 American Time Use Survey, adults aged 35 to 44 averaged 0.09 hours a day on educational activities, roughly 33 hours a year, plus 0.22 hours a day of leisure reading. In the EU, 13.7 percent of adults aged 25 to 64 took part in education or training in the four weeks before the 2025 survey.
How do I decide between learning something and delegating it to AI?
Delegate when the task is rare, the output is checkable and a capable tool exists. Learn to evaluate when the stakes are high but you do not need to do the work yourself. Learn deeply when you use it weekly, it compounds into other skills, and errors are costly.
Won’t I fall behind if I ignore new AI tools?
On some, yes; following every release is impossible. Learn a few deeply, keep a monthly summary for the rest, and set triggers, such as your team adopting a tool, that move an item back onto your list.
When is ignorance never acceptable?
When others carry the risk, when the law expects you to know, when it concerns your own health or finances, and when the topic is a foundation that other skills depend on.
Sources
- Hertwig R, Engel C. Homo Ignorans: Deliberately Choosing Not to Know. Perspectives on Psychological Science 11(3): 359-372, 2016.
- Hertwig R, Engel C (eds). Deliberate Ignorance: Choosing Not to Know. Strüngmann Forum Reports, vol. 29. Cambridge, MA: MIT Press. Chapter 1 and bibliography.
- Gigerenzer G, Garcia-Retamero R. Cassandra’s Regret: The Psychology of Not Wanting to Know. Psychological Review 124(2): 179-196, 2017.
- Simon HA. Designing Organizations for an Information-Rich World. In: Greenberger M (ed). Computers, Communications, and the Public Interest. Baltimore: The Johns Hopkins Press, 1971.
- Sims CA. Rational Inattention: A Research Agenda. Deutsche Bundesbank Discussion Paper, Series 1, No 34/2005. Refers to Sims CA, Implications of Rational Inattention, Journal of Monetary Economics 50: 665-690, 2003.
- Bornmann L, Haunschild R, Mutz R. Growth rates of modern science: a latent piecewise growth curve approach to model publication numbers from established and new literature databases. Humanities and Social Sciences Communications 8: 224, 2021.
- National Science Board and National Center for Science and Engineering Statistics. Publications Output: U.S. Trends and International Comparisons. Science and Engineering Indicators, NSB-2023-33, December 2023.
- U.S. Bureau of Labor Statistics. American Time Use Survey: 2025 Results, news release USDL-26-1022, 25 June 2026, Tables 1, 3 and 11A.
- Eurostat. Participation rate in education and training (last 4 weeks), dataset trng_lfse_01, accessed October 2026.
- Eurostat. Adult Education Survey, participation rate in education and training, dataset trng_aes_100, and survey metadata, accessed October 2026.
- Murre JMJ, Dros J. Replication and Analysis of Ebbinghaus’ Forgetting Curve. PLOS ONE 10(7): e0120644, 2015.
- Arthur W Jr, Bennett W Jr, Stanush PL, McNelly TL. Factors That Influence Skill Decay and Retention: A Quantitative Review and Analysis. Human Performance 11(1): 57-101, 1998.
- Frederick S, Novemsky N, Wang J, Dhar R, Nowlis S. Opportunity Cost Neglect. Journal of Consumer Research 36, December 2009.
Data tables report figures as printed in their sources. The learning-time versus publication table, the annualised time-use hours, the implied 4.26 percent growth rate, the forgetting percentages relative to the 20-minute value, the career mapping of Hertwig and Engel’s six functions, the four-bucket framework, the six-question decision test and the not-to-learn template are CEOtudent analyses or editorial frameworks. Not used: Eurostat average instruction hours per learner, because two Eurostat tables gave different values for the same group; the popular claims that knowledge “doubles every few months” and that a skill’s half-life is a fixed number of years, which we could not trace to a primary measurement; and Sims’s 2003 journal article in its published form, which we could not open, so the rational inattention description relies on his 2005 Bundesbank paper.
This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.
