VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

How Summaries Fail | Why Compression, Omission and Distortion Break Meaning

Three secondary students review open books and notes together in a classroom.

How summaries fail is the study of what disappears when information becomes shorter.

A summary is useful because people cannot always carry the full source into every decision. A chapter becomes a page. A meeting becomes a note. A report becomes an executive brief. A long article becomes a few paragraphs. A lesson becomes revision notes. A dataset becomes a chart and a sentence. An examination passage becomes a compressed answer.

Compression is necessary.

The danger is that shortening does not remove information evenly.

A weak summary can preserve the topic while losing the claim. It can preserve the claim while losing the qualification. It can preserve the conclusion while removing the evidence that made the conclusion reasonable. It can preserve the evidence while losing who said it, when it applied or what population it described.

That is why a summary should not be judged only by whether it is shorter, fluent or easy to read.

A reliable summary preserves the information needed for the next decision.

This master article owns the general failure architecture for information summaries across study, research, work and decision systems. It does not replace eduKateSG’s English examination summary owner, summarisation-as-study owner, reading-to-write synthesis owner or SETC’s English Summary Writing system.

The three companion explainers go deeper into coverage, compression and traceability.


The Core Idea: A Summary Is a Lossy Representation

A summary cannot preserve everything.

That is its purpose.

The correct question is therefore not whether information was removed. Information must be removed.

The question is whether the summary removed the right information for the job it is supposed to do.

A useful summary is a controlled lossy representation: smaller than the source, but still adequate for a defined purpose.

That final phrase matters.

The same source can support several different valid summaries because different readers need different information.


A Summary Has a Receiver

A summary is not merely a smaller text.

It is information prepared for someone who will do something next.

  • A student will revise from it.
  • A parent will decide what to ask the school.
  • A tutor will plan the next lesson.
  • A manager will approve or reject a proposal.
  • A researcher will decide whether to read the full paper.
  • A reader will understand the central claim of an article.
  • A team will carry decisions from one meeting into the next.

The receiver changes what must survive compression.

A revision summary needs retrieval cues. A decision brief needs uncertainty and consequences. A research summary needs source identity, method and limits. An exam summary must obey task scope and word constraints.


Failure Mode 1: Topic Survives, Proposition Disappears

A common weak summary tells us what a source is about without telling us what the source says.

‘This article discusses sleep and learning’ identifies a topic.

‘The article argues that regular sleep supports memory, attention and emotional regulation, which together affect learning’ preserves a proposition.

Topics are labels. Summaries need claims.


Failure Mode 2: The Summary Starts Before the Purpose Is Defined

People often begin summarising by deleting sentences.

That is too early.

First define the use.

If the summary will support a purchasing decision, cost and risk may be central. If it will support teaching, mechanism and misconceptions may matter more. If it will support revision, definitions, relationships and retrieval cues may dominate.

Purpose determines information value.

That selection problem is developed in How Summary Coverage Works.


Failure Mode 3: The Main Idea Is Chosen by Position

Writers sometimes assume the first sentence is the main idea.

Often it is.

Sometimes the real claim develops later, or the opening frames a problem whose resolution appears near the end.

Summary requires reconstructing the source’s informational hierarchy rather than mechanically copying its first line.


Failure Mode 4: Examples Survive While the Rule Disappears

Examples are memorable.

That makes them dangerous during compression.

A vivid case can remain in memory after the general principle has disappeared.

The resulting summary becomes anecdotal.

Ask what the example was doing. Was it evidence? Illustration? Counterexample? Boundary condition? If the example can be removed without losing the governing relationship, it may not deserve scarce summary space.


Failure Mode 5: Qualifications Are Removed

Shortening often strengthens claims accidentally.

Words such as may, often, under these conditions, in this sample, on average, short term and where feasible can look expendable.

They are not expendable when they define the boundary of the claim.

Removing a qualifier can turn cautious evidence into certainty.

SETC’s paraphrase guidance repeatedly stresses preserving scope and certainty because polished wording is still wrong if the claim becomes stronger than the source.


Failure Mode 6: Negation Disappears

A single not can reverse meaning.

Compression systems that focus on keywords can retain the content words and lose the logical direction.

‘The intervention did not improve long-term retention’ cannot become ‘the intervention improved retention’ merely because intervention, improve and retention were preserved.

Negation, condition and exception are load-bearing.


Failure Mode 7: Causation Becomes Association

Summaries often compress relationships more aggressively than nouns.

A source may report that two variables are associated.

The summary may say one causes the other.

That is not efficient compression. It is a different claim.

Relationship words deserve protection because they determine how facts connect.


Failure Mode 8: Sequence Becomes Causation

Event B happening after Event A does not prove A caused B.

A short summary can accidentally convert sequence into mechanism.

Preserve whether the source says before, after, because, therefore, despite, conditional on or merely alongside.


Failure Mode 9: Contrast Is Flattened

Many texts are organised around contrast.

The summary may preserve both sides but lose the fact that the author distinguishes them.

For example, paraphrase and summary both restate source material, but they do different jobs.

Purdue OWL distinguishes summary as a broad overview substantially shorter than the source, while paraphrase generally retains more local detail.

If a summary collapses that distinction, the reader inherits a false equivalence.


Failure Mode 10: Uncertainty Is Rounded Away

A source may contain an estimate, range, confidence interval, disagreement or unresolved question.

A summary can make the message cleaner by deleting uncertainty.

Cleaner is not always truer.

If uncertainty affects the next decision, it must survive.


Failure Mode 11: Population and Context Disappear

A finding about one age group, country, school system, experimental condition or time period may not generalise.

A short summary that drops the population can make a narrow result look universal.

Context should be preserved when it changes applicability.


Failure Mode 12: Time Disappears

Information can be correct and stale.

A summary that omits the date of a policy, price, syllabus, schedule, software version or current event can become misleading later.

Temporal context is part of meaning when conditions change.


Failure Mode 13: Attribution Disappears

A source can report a claim without endorsing it.

A summary can erase this distinction.

‘The report says that critics argue X’ is not the same as ‘the report concludes X’.

Attribution protects who owns which statement.


Failure Mode 14: Evidence Is Removed but Confidence Remains

A summary may preserve a conclusion while dropping the evidence base.

That can be acceptable when the receiver only needs a high-level orientation.

It becomes dangerous when the receiver must judge trustworthiness.

Decision summaries should preserve enough evidence context to distinguish observation, expert judgment, model output, anecdote and speculation.


Failure Mode 15: The Summary Becomes a Stack of Sentences

Shortening each paragraph independently does not necessarily produce a coherent summary.

The original may contain a structure: problem → mechanism → evidence → limitation → implication.

A reliable summary compresses the structure, not merely the sentences.

That structural problem is developed in How Summary Compression Works.


Failure Mode 16: Every Section Receives Equal Space

Equal compression is rarely intelligent.

A source can devote ten pages to background and one paragraph to the decisive result.

A summary that compresses every section proportionally may waste space on background and understate the conclusion.

Space should follow informational value for the intended use.


Failure Mode 17: The Summary Copies Too Closely

Copying fragments can preserve wording while avoiding the real intellectual task of rebuilding meaning.

Purdue OWL recommends understanding the source, setting it aside, restating the main ideas and checking the result back against the original.

SETC uses the same mechanism in classroom form: meaning first, then compression and paraphrase, then verification.


Failure Mode 18: Paraphrase Is Mistaken for Summary

Changing wording is not the same as reducing informational scope.

A paraphrase may remain close to the source’s detail.

A summary selects the main point or points from a broader source.

The difference matters because a writer can produce excellent paraphrases and still fail to identify what deserves inclusion.


Failure Mode 19: Summary Is Mistaken for Synthesis

A summary normally compresses one source or one bounded body of information.

Synthesis reorganises ideas across multiple sources around a new question or structure.

eduKateSG’s Reading-to-Write Synthesis owner protects this distinction.

Stacking three summaries is not yet synthesis.


Failure Mode 20: The Compression Ratio Is Fixed

A chapter-to-page summary, a page-to-paragraph summary and a paragraph-to-sentence summary cannot preserve the same detail.

The more aggressive the compression, the more deliberate the information hierarchy must become.

SETC’s Reading to Notes lesson trains this by moving from paragraph to bullets to sentence to label, then reconstructing upward.


Failure Mode 21: The Summary Cannot Be Expanded Back Into a Coherent Model

A powerful test is reconstruction.

If a learner studies from a summary and cannot explain how the ideas connect, the summary may be too compressed or badly structured.

Good notes are not designed to store every sentence. They should preserve enough structure for the learner to rebuild the larger model.


Failure Mode 22: Source Boundaries Are Lost

A summary can combine ideas from multiple sources so smoothly that readers cannot tell where claims came from.

This is dangerous in research, reporting and decision support.

Source boundaries matter whenever credibility, disagreement or accountability matters.

That problem is developed in How Summary Traceability Works.


Failure Mode 23: Quotations Are Detached From Their Original Function

A striking sentence can survive summary while the surrounding qualification disappears.

The quotation may then appear to support a claim the source did not make.

Exact words still need context.


Failure Mode 24: Numerical Meaning Is Compressed Incorrectly

Numbers are especially vulnerable.

  • Average becomes typical.
  • Percentage-point change becomes percent change.
  • Absolute number loses its denominator.
  • Range becomes midpoint.
  • Nominal value is compared with real value.
  • One-year change becomes long-term trend.

A concise summary should not simplify the measurement into a different statistic.


Failure Mode 25: Outliers and Exceptions Disappear

Sometimes the exception is exactly what the decision-maker needs.

If a process works in 95 percent of cases but fails catastrophically in the remaining 5 percent, a summary that says ‘the process works’ may be operationally useless.

Coverage should follow consequence, not only frequency.


Failure Mode 26: The Summary Hides Disagreement

When credible sources disagree, a summary should not force false consensus.

It may be concise to write ‘research shows X’, but dishonest if the evidence is mixed.

Summaries should preserve disagreement when disagreement changes confidence or action.


Failure Mode 27: The Summary Adds Interpretation

A summary can drift from representation into commentary.

The writer may insert motive, evaluation or implication not present in the source.

Analysis can be useful, but it should be labelled as analysis rather than silently blended into the summary.


Failure Mode 28: The Summary Is Fluent Enough to Escape Checking

Fluent language creates trust.

That is why polished summaries can be dangerous.

A smooth paragraph may hide a missing qualifier, inverted relationship or unsupported inference.

Verification must compare meaning, not style.


Failure Mode 29: The Summary Is Never Checked Against the Source

A summary should survive a reverse audit.

  • Does every major claim map to the source?
  • Did any claim become stronger?
  • Did any population or timeframe widen?
  • Did a condition disappear?
  • Did attribution change?
  • Did the summary omit a decision-critical exception?
  • Did the summary invent a causal relationship?

This is summary verification.


Failure Mode 30: The Summary Becomes the Permanent Source

A summary is often a convenience layer.

Over time, organisations and learners can start treating it as the original.

Each later summary may compress the previous summary rather than the source.

Meaning drifts through generations.

Important summaries should retain a route back to the underlying source.


The Three Pillars of Reliable Summarisation

1. Coverage

Keep the information that the receiver needs for the next decision. Read How Summary Coverage Works.

2. Compression

Reduce redundancy and detail without destroying relationships, hierarchy or scope. Read How Summary Compression Works.

3. Traceability

Keep important claims connected to their sources, dates and evidence boundaries. Read How Summary Traceability Works.


A Practical eduKateSG Summary Test

  • Purpose — what will the receiver do with this summary?
  • Scope — which source material is inside the boundary?
  • Main proposition — what is the source actually saying?
  • Coverage — which ideas must survive?
  • Relationships — which causes, conditions, contrasts or sequences are load-bearing?
  • Qualification — which uncertainty, population or timeframe limits the claim?
  • Compression — what can be removed safely?
  • Attribution — who owns each important claim?
  • Traceability — can the reader reach the source?
  • Verification — has the summary been checked back against the original?
  • Freshness — does time affect validity?
  • Reconstruction — can the receiver rebuild enough of the model to act correctly?

Example: Student Revision Notes

A student reads a chapter and writes three pages of notes.

If the notes are simply copied sentences, little compression has occurred.

If the notes become ten disconnected keywords, too much structure may have disappeared.

A better summary preserves the chapter’s organising relationships: definitions, cause-and-effect links, contrasts, procedures, exceptions and retrieval cues.

The goal is not smallest notes.

The goal is a representation from which useful knowledge can be rebuilt.


Example: Mathematics

A Mathematics summary can be misleading when it stores formulas without conditions.

A learner writes the quadratic formula but omits that it applies to equations that can be written in the form ax² + bx + c = 0 with a non-zero leading coefficient.

The symbol survived. The applicability boundary disappeared.

Bukit Timah Tutor’s checking approach provides the complementary discipline: do not trust a representation merely because it looks familiar; verify it against the conditions of the problem.


Example: English

SETC’s Summary Writing owner treats summary as controlled reduction: select, compress, paraphrase and preserve meaning.

This general eduKateSG article deliberately sits above that exam-specific system.

The same mechanism explains why an executive brief, research note or lesson summary fails when it preserves words but loses the informational architecture.


Example: Research

A research summary that says ‘the intervention improved outcomes’ may be too weak if the full study only found a small short-term effect in one subgroup.

The key question is not how many words were removed.

It is which constraints were removed.


Example: Meetings

A meeting summary that records discussion but omits decisions, owners and unresolved questions is shorter but not useful.

For the next meeting, those state changes are often the most valuable information.


Example: AI-Generated Summaries

AI can produce fluent compression quickly.

The reliability test remains humanly familiar: compare the summary with the source, preserve uncertainty, keep source boundaries, and do not let a concise answer outrun the evidence.

eduKateSG’s Super Intelligence reading guide already owns the workflow for AI-assisted reading. This cluster owns the general information-reliability mechanics beneath it.


Across the eduKate Ecosystem

eduKateSG owns the general information-summary mechanism. SETC owns the English summary-writing system and summarising and paraphrasing in the English learning library. SETC Reading to Notes owns the classroom path from source to usable notes. Bukit Timah Tutor contributes the verification principle that a check should be able to disagree with the original route. eduKate Sengkang supplies the independence layer: notes and summaries should eventually support learner-owned reconstruction rather than tutor-dependent recall. eduKate Yishun contributes evidence-led recovery: summary labels should not replace diagnosis of the first broken dependency.


Sources and Further Reading

Purdue Online Writing Lab — Quoting, Paraphrasing and Summarizing

Purdue Online Writing Lab — Summarizing

George Mason University Writing Center — When to Summarize, Paraphrase and Quote

Singapore English Tuition Centre — How Summary Writing Works


Continue the Series

How Summary Coverage Works | Keep Decision-Critical Information

How Summary Compression Works | Shorten Without Distorting Structure

How Summary Traceability Works | Keep Claims Connected to Sources

eduKateSG