When Information Becomes Abundant, Attention Becomes Precious

Verification begins with a question. Reading does, too.

Imagine the subscription-business owner has asked an AI to analyze a long report on customer cancellations. The answer arrives quickly: several pages of summaries, possible causes, suggested charts, and recommendations. Reading every sentence from top to bottom may feel responsible. It may also be a poor use of attention. Before opening the report—or scrolling through the answer—the owner should decide what the reading is meant to accomplish.

There are several different jobs reading can do. You might be trying to locate a claim: What does the report say caused cancellations to rise? You might be trying to learn a concept: What is a cohort, and why would it matter here? You might be trying to compare options: Which proposed analysis would best test the owner’s hunch? Or you might be trying to check a risk: Does the recommendation rely on data the business does not have?

Each purpose directs attention differently. To locate a claim, scan headings, summaries, and relevant passages, then read the surrounding text closely enough to see what the claim actually says. To learn a concept, slow down around definitions and examples. To compare options, look for shared criteria—cost, evidence required, time, and what each analysis could establish. To check a risk, look for assumptions, missing evidence, and consequences if the advice is wrong.

This is selective reading, not careless reading. The aim is not to skim everything at high speed. It is to spend effort where it can change your understanding or decision. A report may contain many pages irrelevant to today’s question and one sentence whose qualification matters enormously: cancellations rose among new customers, for instance, rather than across the whole business. That qualifier can overturn a broad recommendation.

A useful habit is to state your reading purpose in a few words before you begin: find the claim; understand the term; compare the choices; check the assumption. Then change gears when the task changes. Scan to find the relevant passage; read closely to interpret it; pause to ask whether the evidence supports the conclusion. If an AI-generated answer includes a citation or points to a particular figure, follow that trail and inspect the source rather than treating the summary as a substitute for it.

This also makes reading more productive when you do not yet know enough to ask a precise question. You can begin with a provisional purpose—“find out what this report measures”—and refine it as you learn. The purpose is a compass, not a contract. If the reading reveals a more important question, turn toward it.

The owner need not absorb every sentence the AI produced. But before acting, they should understand the recommendation, its basis, and the uncertainty around it. Selective attention saves effort by separating what can be safely passed over from what deserves a closer look. The skill is not reading less at any cost. It is knowing where careful reading matters.

Knowing where careful reading matters is only half the job. Once you have found the important passage, you still need to understand what it says—and whether its reasoning holds together.

This is where the promise of speed-reading can become a distraction. A review of reading research by Keith Rayner and colleagues found a trade-off between speed and comprehension: doubling or tripling reading speed while preserving the same understanding is unlikely. Skimming can serve a different purpose—finding the gist or a particular detail—but it sacrifices comprehension. The review points instead toward reading practice and stronger language skills, including vocabulary, as ways to become a more effective reader.^[2]

That gives the language advantage a practical direction. Learn enough of a field’s language that you spend less effort decoding each term, and choose your pace according to the task. Moving faster through words is useful only if you still grasp the distinctions that matter.

A better aim is not maximum speed but appropriate speed. Skim a familiar overview. Slow down for a recommendation that could affect a budget, a lesson, or a customer. Reread a sentence when its meaning depends on a qualification such as unless, only if, or on average. These small words can carry large consequences.

One practical test is to close the text and explain its point in plain language. If a report says, “The intervention was associated with improved retention,” you might restate it as: “People who received the intervention stayed more often.” That is a useful start. But it is not the same as saying the intervention caused them to stay. Your paraphrase should preserve the claim’s limits, not quietly upgrade it.

Consider a short, invented passage from a customer report:

Customers who attended an optional onboarding session renewed more often than customers who did not. Attendance was voluntary. The comparison did not account for customers' previous experience with the product.

A hurried reading might leave you with “Onboarding improves renewal.” A closer reading leaves you with a different note: “Attendees renewed more often, but they chose to attend; the groups may already have differed.” The second note preserves both the finding and its limit. The words optional, voluntary, and did not account for are carrying much of the meaning.

That reading changes the next AI exchange. “Write a plan to increase renewals through onboarding” assumes the conclusion. “Using this passage, explain what we know about onboarding and renewal, what remains uncertain, and what comparison could help us investigate” keeps the inquiry open. The advantage came from understanding three sentences, not from discovering a special instruction.

Try the same exercise with a passage relevant to your own work. Read it, put it aside, and write two sentences: what it establishes and what it leaves unresolved. Reopen it and compare. If your version turns may into will, or drops the group to which the finding applies, repair the note. Then use that note to formulate one question. You have made reading, writing, and inquiry part of the same task.

Then ask what supports the claim. Does the passage point to a comparison, a source, or a method? Is the evidence strong enough for the conclusion, or does the conclusion reach further than the evidence? Following the argument is more valuable than admiring its polish. Fluency is not evidence.

Pay special attention when a key term appears to shift meaning. Retention might mean customers who renew, students who remain enrolled, or employees who stay through a given period. Significant might mean important in ordinary conversation or statistically distinguishable under a particular test. If a term changes the result, pause and ask what it means here. A definition can be a hinge: everything in the argument turns on it.

You can make this habit concrete with three short checks: restate the main point without copying the wording; identify the evidence offered for it; and mark any term or qualification that could alter the conclusion. This does not require reading every line twice. It requires noticing where comprehension is fragile and spending attention there.

The owner of the shop need not become a research specialist to assess an AI’s recommendation. But they can ask whether “improved retention” means more customers renewed, whether the comparison was fair, and whether the text actually supports the proposed change. Those questions turn the recommendation into something inspectable.

Reading well is not simply taking in more words per minute. It is building a reliable model of what the words mean, what they support, and what remains uncertain. When the stakes rise, the pause is not a failure of speed. It is part of understanding.

Sometimes the question that matters most is hidden inside a word you have not met before.

Imagine the shop owner from the previous example reading an AI recommendation about customer retention. The answer refers to “cohort analysis.” The phrase sounds technical enough to matter, but not so technical that it can safely be ignored. What does it mean—and would understanding it change the decision?

A useful first move is to notice which terms keep appearing. In an unfamiliar report, article, or AI conversation, mark words that recur, organize the argument, or carry more weight than their everyday meaning suggests. You do not need to stop for every unknown word. Focus on terms that seem central to how the analysis is built.

Then ask for an explanation in plain language, followed by an example: “Explain cohort analysis as if I run a small shop. Give me a simple example, and tell me what it can and cannot show.” In general, cohort analysis groups people or other observations by a shared characteristic—often when they entered a system or experienced an event—and examines how the groups behave over time. The owner might compare customers first acquired during a holiday promotion with those who came in through referrals, then examine whether their return-purchase patterns differ.

That example can help the owner follow the analysis, but the important learning is in the questions it makes possible: Were the groups defined sensibly? Is the comparison period long enough? Could another factor explain the difference? Learning a recurring term may make an analysis easier to inspect; it does not by itself establish that the analysis is sound or that the owner has mastered the field. The term is a foothold for examining the evidence and assumptions, not proof that the conclusion is reliable.

To move a term from recognition into use, keep a short note with four parts: its meaning in your own words, an example from your work, a boundary or non-example, and a question it now lets you ask. For the owner, that might look like this:

• Meaning: A cohort is a group linked by a shared starting event or characteristic, whose behavior we can follow over time.
• My example: Compare customers who first subscribed in April with those who first subscribed in May, measuring how many remain after their first month.
• Boundary: Comparing total cancellations in April and May is a calendar-month comparison; it does not, by itself, track groups from a shared starting point.
• Better question: “Compare retention at the same time since joining for each signup cohort. What differences in acquisition channel or offer could complicate the comparison?”

Check the definition in a dependable source. Later, try writing the explanation and question again without looking at the note. If you cannot explain why the groups need the same follow-up period, return to that gap before collecting another term. AI can offer a fresh example for you to classify and explain; the useful part is doing the explaining yourself.

Return to the cohort note a few days later with a fresh case. Suppose customers who joined in April have been observed for two months, while those who joined in May have been observed for only one. Would comparing the proportion still subscribed today answer which group stays longer? Explain your answer before consulting the note. The unequal time available to cancel is the problem; matching the time since joining makes the comparison more informative.

Then carry the idea into a different setting. A teacher comparing students who started a course in different months might ask how each group was doing after its first four weeks. The subject has changed, but the need for a comparable follow-up period remains. If you can recognize that connection and explain its limits, the word is beginning to belong to you.

Build this vocabulary from sustained reading around questions you care about. An introductory chapter can show how several terms relate; a worked case can show what they change in practice; a contrasting account can reveal where an explanation is incomplete. Follow a useful term back into those settings instead of collecting isolated definitions. You might finish a reading session with just one new concept, one example, and one better question. That is enough to bring something new to your next conversation with AI.

Curiosity starts this process with “What does that mean?” Writing pushes it further: “What can I now ask that I could not ask before?” Keep the terms that improve your questions, and use them in the next real task. A vocabulary becomes valuable when it changes what you notice and do.