Cheating and Academic Integrity — What It Is, How Common It Is, Why It Happens, and the Two Ways to Read It

Education-Atlas foundations brief 03 — a neutral, globally-sourced survey of academic dishonesty and its meanings

This brief sits in the **foundations** layer of the atlas. Like its siblings, it deliberately does not argue a position. Its job is to **map the terrain fairly** — to define the contested category of "cheating," report what the evidence actually says about how common it is and why it happens, trace its long history, and then lay out the **two serious framings** of the phenomenon — integrity as a genuine good, and "cheating" as a predictable artifact of the system — each in the voice of its strongest defenders, without adjudicating between them. A document that takes a side here would be useless as a foundation; the rest of the repository can take sides, but only honestly if it stands on an even-handed account first.

Reading stance. Two failure modes haunt writing on this topic. The first is the moralizing register that treats every instance of cheating as simple individual vice and never asks what the system is doing to produce it. The second is the debunking register that treats integrity as a naïve fiction and explains all cheating away as a rational response to a rigged game. Both contain real truth; neither is the whole truth. This brief refuses both. It takes integrity seriously as a real value and takes the structural critique seriously as real analysis, and it is candid throughout that the definitions are contested and the data are weak.

0. Why this is harder than it looks

Most people think they know what cheating is. Pressed for a definition, they will say something like "breaking the rules to get an unfair academic advantage." That definition is serviceable and also conceals three genuine difficulties.

First, the rules are not the same everywhere, and some of what one institution calls cheating another calls collaboration, or good scholarship, or respect for one's teacher. The boundary between legitimate help and illegitimate collusion is drawn differently across cultures, disciplines, and even individual instructors — and students are frequently penalized for crossing a line they did not know existed.

Second, the data are thin and self-reported. Almost everything we "know" about how much cheating happens comes from anonymous surveys asking people to admit to behavior they have every incentive to hide. The numbers are real signals, but they are floors, not measurements, and they are sensitive to how the question is worded.

Third, and most consequentially, the same behavior supports two completely different explanations. One reading sees a student who cheats as failing a moral test that the institution rightly sets. The other sees the same student as responding rationally to an assessment that measures the wrong thing under stakes that are too high. The behavior is identical; the meaning is contested. Sorting out which reading applies — and how much of each applies — is the central unresolved question of the field, and this brief will not pretend to resolve it.

The structure that follows: definitions and their cultural variability (§1); prevalence and trends with the real numbers and their limits (§2); the evidence on why students cheat (§3); the long history (§4); the two serious framings, each steelmanned (§5); detection and its problems, segueing to the AI rupture (§6).

1. What counts as cheating — and why the boundary moves

1.1 The standard taxonomy

There is rough consensus on a working vocabulary, even if the edges are disputed. The common categories of academic dishonesty (or academic misconduct) are:

• Plagiarism — presenting someone else's words, ideas, or work as your own without

attribution. Ranges from verbatim copying to "patchwriting" (light paraphrase that tracks a source too closely) to self-plagiarism (resubmitting your own prior work).

• Collusion — collaborating on work meant to be done individually, or sharing work so

another student can submit it. The mirror image of legitimate collaboration; the line between them is the single most contested boundary in the field (see §1.2).

• Contract cheating — outsourcing assessed work to a third party (an essay mill, a

freelancer, a friend, or now an AI system) who makes a contribution substantial enough to cast reasonable doubt on authorship. The term was coined by Clarke and Lancaster in 2006.

• Fabrication and falsification — inventing data, sources, or results, or altering them.

Central to research-integrity contexts; also covers faking citations.

• Exam cheating — using unauthorized materials, copying from others, impersonation,

obtaining questions in advance, or smuggling information into a test.

These are descriptive buckets, not a natural taxonomy. The same act can fall into several at once, and institutions draw their internal boundaries differently.

1.2 The genuinely contested boundary: collaboration vs. collusion

Every educator wants students to learn from each other and also wants assessed work to reflect what each individual can do. Those two goals collide constantly, and the line between "working together" (encouraged) and "collusion" (punishable) is drawn case by case, instructor by instructor, with little consistency. A study group that is virtuous for one assignment becomes misconduct for the next, and students are often not told which is which until after they have crossed the line. This is not a fringe edge case; it is the modal ambiguity that produces the most aggrieved misconduct cases, because the student genuinely believed they were doing what was wanted.

1.3 Cheating is culturally variable — descriptively, not as an excuse

This is the part most worth getting right, and the part most easily mishandled in either direction. The empirical literature on international and cross-cultural academic integrity makes a careful, two-sided claim that is easy to caricature.

The case that norms genuinely differ. The Western academic regime rests on an individual-authorship assumption: ideas are property, attribution is owed to named individuals, and reproducing a text without citation is theft. This assumption is neither universal nor ancient. In several East Asian intellectual traditions, knowledge that is "beneficial to and shared by the community" is treated as common rather than owned, and faithfully reproducing a respected teacher's or canonical text's words can be read as a virtue — a sign of mastery and respect — rather than as theft. For Confucius, study itself meant "finding a good teacher and imitating his words and deeds." Memorization and faithful reproduction of authoritative texts were the core of classical examination traditions for two millennia (see §4). A student formed in such a tradition can land in a Western university and be baffled that citing many sources at the end of a paper is required — and may even find the practice mildly disrespectful, elevating individual authors above the shared body of knowledge. Cross-cultural scholars (e.g., work collected on "textual borrowing practices" of international students) argue that some proportion of what gets flagged as plagiarism is a literacy and acculturation gap, not dishonesty: students who have never been taught the citation conventions of a foreign academic culture committing "unintentional plagiarism." On this reading, treating every such case as a moral offense is both unfair and pedagogically counterproductive.

The case against overstating it. The same literature contains a sharp corrective, and honesty requires giving it equal weight. First, the empirical results often contradict the stereotype: several studies find that individualists plagiarize more than collectivists, plausibly because individualism is associated with the pursuit of personal advantage; and a number of studies find no significant difference in actual plagiarism rates between Asian and Caucasian students once confounds are controlled. The intuitive "collectivist cultures excuse copying" story does not survive contact with the data as cleanly as it is often told. Second, scholars warn that the cultural-difference frame can become a condescending stereotype — assuming international students cannot grasp that buying an essay or copying in an exam is wrong, when in fact deliberate cheating is understood as wrong across virtually all cultures. Most students everywhere can tell the difference between a citation-convention they were never taught and paying a stranger to write their dissertation. Third, some argue the cultural-relativism move can be weaponized to excuse misconduct or to lower expectations for some students, which is itself a form of disrespect.

The defensible synthesis — which most of the field now holds — is narrow and important: the boundaries of textual and collaborative conventions are genuinely culturally variable and frequently untaught, so honest mistakes are real and should be handled as teaching, not prosecution; but the core of deliberate deception is recognized as wrong nearly everywhere, and cultural difference is not a license to cheat. Both halves of that sentence are load-bearing. Dropping either one produces a position the evidence does not support.

2. How common is it? — the numbers and why to distrust them

2.1 The McCabe corpus — the closest thing to a long baseline

The late Donald McCabe built the largest body of self-report survey data on academic dishonesty, surveying hundreds of thousands of students across North America from the 1990s onward; the International Center for Academic Integrity (ICAI) maintains and reports the aggregate figures. The headline numbers, as reported by ICAI and McCabe's published work:

• Across 70,000+ U.S. high-school students, 64% admitted cheating on a test, 58%

admitted plagiarism, and 95% admitted to some form of cheating (test, plagiarism, or copying homework).

• Among undergraduates (data pooled roughly 2002–2015), **65–75% admit to cheating at

least once, and about 19–20% admit to cheating five or more times; roughly 62%** admit to cheating on written assignments at least once.

For long-run context, the Educational Testing Service has noted that admitted cheating among U.S. college students rose from around 20% in the 1940s to majorities in recent decades — though, crucially, this partly reflects changes in willingness to admit and in survey methods, not only changes in behavior.

The robust takeaway from McCabe's decades of work is not a precise rate but a structural fact: some academic dishonesty is normal — a majority behavior over a student career — rather than the deviance of a corrupt few. That reframing (from "a few bad apples" to "a near-universal temptation unevenly acted on") is McCabe's most durable contribution, and it is what makes a purely individual-vice account hard to sustain on its own.

2.2 Contract cheating — Newton's meta-analysis and the rising-trend debate

The sharper, more policy-relevant number is contract cheating, because it represents deliberate outsourcing rather than opportunistic copying. Philip Newton's 2018 systematic review (Frontiers in Education) is the standard reference. Pooling 71 samples from 65 studies, 1978–2016, totaling 54,514 respondents, it found a historical average of 3.5% (3.52%) of students self-reporting commercial contract cheating.

But the trend is the story. Newton found a statistically significant positive correlation between year and prevalence (r ≈ 0.37, p = 0.0016), and that in studies since 2014, the self-reported rate rose to 15.7% — roughly one in seven recent students — which, extrapolated globally, he estimated at on the order of 31 million students worldwide. Other studies in the same range cross-validate the historical baseline: Curtis and Clare (2017) found ~3.5%; Bretag et al. (2018) found ~5.78% overall self-reported engagement across a large Australian sample.

Newton is careful, and his cautions matter. He notes the rise "may be due to an overall increase in self-reported cheating generally, rather than contract cheating specifically," and that varying definitions inflate cross-study comparison. So the honest statement is: contract cheating appears to be low single digits historically and materially higher in recent samples, with the exact magnitude uncertain.

2.3 Exam vs. assignment, and country variation

Self-report consistently shows exam cheating and assignment cheating are different populations with different rates and different methods; aggregating them hides this. And the cross-national spread is large. A comparative study across European countries reported exam-cheating rates ranging from 62% in Portugal to 94% in Romania; cross-national work finds the educational system and social context of a country predict adolescent cheating as much as individual traits. In Russia, where educational bribery is described as endemic, estimates suggest up to 50% of students encounter corruption in their academic careers, and Russian students show among the highest tolerance for cheating measured, contrasting with low tolerance among U.S. students. Analysts attribute part of the post-communist pattern to a durable social condemnation of informers — a legacy that makes peer reporting and enforcement culturally costly. A 2007 survey found 18–20% of students in Bulgaria, Croatia, and Serbia, and ~40% in Moldova, reported using some illegitimate method to gain university admission. The point is not to rank nations morally but to show that prevalence is embedded in institutions and history, not a fixed property of "students."

2.4 The limits of self-report (read this before quoting any number above)

Every figure in this section comes from people voluntarily admitting misconduct on surveys, and that fact should temper all of them:

• Under-reporting / social desirability. People hide stigmatized behavior even when

anonymous. Newton explicitly argues non-responders are more likely to have cheated, which biases estimates downward. Reported rates are best read as floors.

• Convenience samples. In Newton's corpus, 70% used convenience sampling, only ~55%

reported response rates (median ~17%), and over half did not guarantee anonymity — conditions that further depress admissions.

• Definition drift. "Cheating" means different things in different surveys; "have you

ever cheated?" and "have you paid for an assignment in the last year?" are not the same question, and lumping them produces incoherent aggregates.

• Self-report ≠ detection. What students admit, what gets caught, and what actually

happens are three different quantities, and we mostly have only the first.

None of this makes the data worthless — the convergence across decades and countries that substantial cheating is common is robust. It means the precise percentages should be held loosely and never cited as if they were measured rather than confessed.

3. Why students cheat — the evidence

No single cause explains cheating; the literature is a set of overlapping, well-supported mechanisms. The most honest summary is that cheating is over-determined — most real cases involve several of the following at once.

1. Pressure and stakes. Experimental and survey work consistently finds that high-pressure, high-stakes conditions increase both the likelihood and the amount of cheating. When a single assessment carries heavy weight, the expected payoff of cheating rises and the perceived cost of failing honestly rises with it. This is the most replicated situational finding.

2. Performance goals vs. mastery goals. Students oriented toward performance goals (looking competent, getting the grade, beating peers) cheat substantially more than students oriented toward mastery goals (actually understanding the material). The orientation is partly dispositional and partly created by the assessment environment — grading on a curve, class rank, and credential gatekeeping all push students toward performance framing.

3. Opportunity and low vigilance. Cheating rises where it is easy and unlikely to be caught — unmonitored online assessments, recycled exam banks, inattentive proctoring. Opportunity is a necessary enabler even when motivation is present.

4. Perceived unfairness and peer norms. Students who believe an assessment is unfair, irrelevant, or arbitrary, or that the instructor "doesn't care if they learn," cheat more. And the single strongest social predictor in the meta-analytic literature is the perceived peer-cheating effect: believing that one's peers cheat is a powerful driver, partly through the "cheat or be cheated" logic — if competitors are gaining an unfair edge, honesty feels like unilateral disarmament.

5. Neutralization / rationalization. Drawing on Sykes and Matza's criminological neutralization theory, research finds students protect their self-image while cheating by deploying justifications — "the material is irrelevant," "the professor is unfair," "the test is too hard," "everyone does it," "I had no choice." Neutralization techniques show a medium-to-large positive effect on cheating. The significance is that most cheaters do not see themselves as dishonest people; they first reframe the act as not-really-wrong. This is why exhortations to "have integrity" often miss — the student already believes they are acting reasonably under the circumstances.

6. Assessment–learning mismatch. When assessment tests something the student does not see as the point of the course — rote recall they consider pointless, busywork, a hoop — cheating becomes, in the student's framing, a shortcut around an obstacle that is not measuring real learning anyway. This mechanism is the empirical bridge to the structural critique in §5b: it locates part of the cause in the design of the assessment, not only in the character of the student.

These mechanisms are not rival theories so much as different levels of one picture: dispositions (goal orientation), situations (stakes, opportunity, peer norms), cognitions (neutralization), and structures (assessment design). A complete account needs all four, which is precisely why both framings in §5 can point to real evidence.

4. History — cheating and its policing are ancient

Cheating is not a modern decline-of-values story; it is as old as high-stakes assessment itself, and so is the arms race to stop it.

Imperial China — the paradigm case. The Chinese imperial civil-service examination (keju), which over roughly 1,300 years selected officials by merit, created enormous incentives to cheat and correspondingly draconian countermeasures. Candidates were strip- searched on entry; smuggling techniques became an art form — miniature crib books, characters written on undergarments and the body, hired substitutes ("body-doubles"), and bribery of examiners. Surviving "cheating garments" covered in thousands of micro-written characters sit in museums today. The state's response escalated to extremes: candidates caught cheating could be expelled, banned, stripped of prior degrees, exiled, or executed; and the harshest penalties fell on examiners and administrators who failed to prevent fraud. In the roughly 260-year Qing dynasty there were at least nine cases in which exam officials were sentenced to death for examination scandals — five in the single notorious year of 1657, when examiners in the Jiangnan scandal were executed. The keju also pioneered the anti-cheating toolkit still in use: anonymous marking by candidate number, and recopying every paper by a clerk so graders could not recognize handwriting. Two themes that recur throughout history are already complete here: where a credential controls life outcomes, people will cheat in proportion to the stakes; and detection is a permanent, escalating arms race.

Honor codes — the Western institutional answer. A different lineage runs through the honor code tradition, formalized at U.S. institutions such as the University of Virginia (1842) and the College of William & Mary, and at the military service academies. Honor codes locate integrity in a community pledge and student self-governance rather than external surveillance — students promise not to cheat and, often, to report those who do. McCabe's own research found honor-code institutions tend to show lower self-reported cheating, evidence that culture and norms move behavior, not only enforcement — a finding both framings in §5 claim.

The modern detection arms race. From proctors and blue books to plagiarism-detection software (§6) to webcam proctoring to AI-text detectors, each new assessment technology has generated a new way to cheat and a new way to catch it, with the catchers perpetually one step behind. The keju clerk recopying papers and the 2020s AI detector are the same move 1,300 years apart.

5. The two serious framings — both steelmanned, neither adjudicated

Here is the crux. There are two intellectually serious ways to read the entire phenomenon. They are not symmetrical opposites — one is primarily a moral/normative position and the other primarily a structural/diagnostic one, and a person can hold parts of both. But they generate different prescriptions, and most public argument collapses one into a straw version of the other. This brief presents each in the voice of its strongest advocates and then deliberately stops, because choosing between them is a value judgment the foundations layer declines to make.

5a. Framing one — integrity as a genuine moral and epistemic good

This is the position of the academic-integrity field — the International Center for Academic Integrity, scholars such as Tricia Bertram Gallant and David Rettinger, and the broad consensus of educators who treat integrity as central to the enterprise. Its strongest form:

Integrity is not bureaucratic compliance; it is constitutive of what education is for. ICAI defines academic integrity as a commitment to six values — honesty, trust, fairness, respect, responsibility, and courage. The argument runs:

• Epistemic. Knowledge is a collective enterprise built on trust in attribution and

honest reporting. Plagiarism and fabrication corrupt the record itself; if we cannot trust that a claim is the author's and that data are real, scholarship stops working. This is the same value that makes research fraud a catastrophe, scaled down to the classroom.

• Fairness. Cheating is theft from honest students — it distorts the grade

distribution, devalues the credential others earned honestly, and lets the dishonest free- ride on the system's trust. The "cheat or be cheated" dynamic is precisely why fairness must be defended: tolerated cheating punishes integrity.

• The credential means something. A degree certifies a capability to employers,

patients, clients, and the public. A cheated medical degree, engineering license, or pilot certification is a public-safety problem, not a private moral lapse. The credential's value to everyone depends on its honesty.

• Character and formation. Education forms people, not just skills. Practicing honesty

under temptation is part of becoming the kind of person and professional society needs. Integrity is a virtue to be cultivated, not merely a rule to be enforced.

Critically, the contemporary version of this framing — Bertram Gallant's "teaching and learning approach" — is not primarily punitive. It explicitly shifts the question "how do we stop students cheating?" to "how do we ensure students are learning?" and locates integrity within the educational mission: build cultures of integrity through assessment design, clear communication, and trust, so that the temptation and the opportunity both shrink. Her 2025 work, The Opposite of Cheating: Teaching for Integrity in the Age of AI, frames the goal as cultivating integrity rather than merely detecting violations. So the strongest form of framing one already absorbs much of the structural critique — it agrees that bad assessment design drives cheating — while insisting that integrity remains a real and irreducible good, not merely an artifact to be engineered away. The value is genuine; the methods should be humane and pedagogical.

5b. Framing two — "cheating" as a predictable artifact of the system

This is the structural critique. It does not deny that dishonesty is real or that integrity matters; its claim is that much of what is labeled cheating is a foreseeable output of how we credential, motivate, and assess — a symptom of the system rather than (only) of student vice. Its strongest form draws several threads together:

• Goodhart's Law. "When a measure becomes a target, it ceases to be a good measure." The

grade and the degree began as proxies for learning. Once they became the targets that gate jobs, visas, status, and self-worth, optimizing the proxy — including by cheating — became rational, and the proxy stopped reliably measuring the thing it was meant to track. On this reading cheating is not an aberration in the credential system; it is the credential system working as incentivized. Campbell's Law makes the parallel point for high-stakes testing: the higher the stakes attached to a quantitative indicator, the more it will be gamed and the more it will distort what it measures. Teacher-side cheating (altering scores under accountability pressure) is the same mechanism on the other side of the desk.

• Extrinsic motivation crowds out learning. When the system rewards performance over

mastery (curves, rank, gatekeeping), it manufactures exactly the goal orientation that §3 identifies as the strongest driver of cheating. The structure trains students to care about the grade more than the knowledge — and then blames them for acting on what they were trained to value.

• The credentialing/signaling critique. Economists like Bryan Caplan (*The Case

Against Education, 2018) argue that much of education's payoff is signaling — certifying pre-existing intelligence, conscientiousness, and conformity to employers — rather than building human capital (his contested estimate: ~80% signaling). If the degree's value lies substantially in the signal rather than the learning, then cheating to obtain the signal while skipping the learning is, from the student's incentive structure, a coherent* move: they are buying the thing the market actually rewards. (This is a sharply disputed thesis — many economists hold education builds substantial real skills — but it is the strongest economic statement of the structural view.)

• Assessment tests the wrong things. When assessment rewards rote recall, formulaic

output, or volume — things that are easy to fake and that students correctly perceive as disconnected from real competence — cheating is a shortcut around a bad measurement. Fix the assessment (authentic, applied, oral, iterative, lower-stakes-but-more-frequent) and much cheating loses its point, because there is no longer an easy gap between the proxy and the real thing.

The structural conclusion is not "therefore cheating is fine." It is: if you build a high-stakes credential on extrinsic motivation and gameable measures, predictable rates of cheating are the system's own output, and treating it purely as individual moral failure is both inaccurate and ineffective — you are punishing students for a problem partly authored by the design.

5c. Why this brief does not pick a side

Each framing explains things the other struggles with. Framing one explains why honor-code cultures and high-trust environments genuinely reduce cheating, why research fraud is catastrophic regardless of incentives, and why most people across cultures recognize deliberate deception as wrong — integrity tracks a real value, not just an equilibrium. Framing two explains why cheating scales with stakes, tracks assessment design, rises under performance pressure, and recurs identically across 1,300 years and every culture the moment a credential controls life outcomes — patterns that a pure-vice account cannot accommodate.

The strongest contemporary practitioners (Bertram Gallant) and the strongest structural critics (the Goodhart/assessment-design camp) converge on much of the practice — better assessment, lower-stakes designs, cultures of trust — while diverging on the meaning: is integrity a genuine good we are failing to cultivate, or a label we apply to behavior the system manufactures? That is a normative question about value, not an empirical one that more data will settle, and per the convention of this foundations layer, it is left open. A reform thesis built downstream of this brief may take a side; it should do so knowing that it is taking a side, and that the other side has real evidence.

6. Detection and its problems — and how AI broke it

Detection deserves its own section because it is where the integrity field's tools meet their hardest limits, and where the whole topic collides with generative AI (handled in depth in the companion AI brief; this is the segue).

6.1 Plagiarism detection (Turnitin et al.) and the false-positive / equity problem

Text-matching software like Turnitin became near-universal by computing a similarity score against a corpus of prior submissions, web pages, and publications. It is genuinely useful for catching copy-paste plagiarism. But it has well-documented limits:

• A similarity score is not a plagiarism score. High similarity can come from correctly

quoted material, shared boilerplate, common technical phrasing, reference lists, or many students citing the same foundational texts. The number requires human judgment; treating it as a verdict produces false accusations.

• Equity / multilingual concerns. Critics argue the tools systematically disadvantage

non-native English writers, whose translated idioms and formulaic phrasing can inflate matches, and disciplines with highly formulaic conventions. (Turnitin disputes the magnitude; the contested 2023 figures around false positives for short papers are widely cited and widely challenged — note the dispute rather than treating any single number as settled.)

• It detects copying, not authorship. Text-matching never caught contract cheating

or any bespoke work — purchased essays and ghost-written assignments score low similarity because they are original text. The most serious form of deliberate cheating was always invisible to the dominant tool.

6.2 The proctoring backlash

Remote and algorithmic proctoring (Proctorio, ProctorU, Proctortrack, and peers) exploded during COVID-19 — Proctorio reported business up ~900%, proctoring ~2.5 million tests in April 2020 alone. The backlash was substantial and rests on documented concerns:

• Surveillance and privacy. Webcam monitoring, screen and keystroke recording, room

scans of students' homes, and gaze/behavior tracking struck many as disproportionate intrusion, eroding the trust that the integrity field itself says matters.

• Bias and accessibility. Facial-detection systems failed to recognize **darker-skinned

students**; behavior flags penalized students with disabilities, anxiety, or non-normative movement; the systems were argued to reproduce existing inequities.

• Effectiveness in doubt. Several studies found students performed **better on

un-proctored than on proctored remote exams** — consistent with the tools adding stress and surveillance harm without clearly delivering the integrity they promised.

• Institutional retreat. After student boycotts, multiple institutions **declined to

renew** proctoring contracts. The episode is a cautionary case: detection technology can damage the educational relationship faster than it protects the credential.

6.3 How AI breaks the whole detection paradigm

Generative AI is the rupture that the rest of this topic now bends around, and it breaks detection from both ends:

• AI-written work is original text. Like contract cheating, an essay generated by a

large language model is not copied from anything, so similarity detection sees nothing. The dominant tool of the last two decades is structurally blind to it.

• AI-detectors are unreliable in both directions. Tools claiming to detect AI-generated

text produce false positives and false negatives at high rates. Turnitin itself acknowledged elevated false positives in the low-AI-percentage range and the difficulty of short documents. Studies report AI detectors flagging human writing as machine-written — with particular concern about non-native English writers, whose more uniform, less idiomatic prose pattern-matches to "AI" (a bias Turnitin's own data partly disputes; the research is unsettled). Accusing a student of AI use on a detector's say-so risks punishing the innocent, and several institutions have disabled AI detection rather than rely on it.

• The boundary itself dissolves. AI does not just evade detection; it **collapses the

collaboration-vs-collusion line** from §1.2 into incoherence. Is using AI to brainstorm cheating? To outline? To edit grammar? To draft a paragraph you then rewrite? There is no consensus, conventions differ by instructor and institution, and students are once again penalized for crossing lines nobody clearly drew — the §1.2 problem at civilizational scale.

This is where the topic hands off. The arrival of AI does not introduce a new chapter so much as it forces the unresolved question of §5 into the open: if the cheapest tool in the world can produce the proxy (an essay, a problem set, a passable exam answer) without the learning, then either we recommit to integrity as a cultivated good and redesign assessment to make the learning the point again (framing one's prescription), or we concede that high-stakes credentialing on gameable measures was always going to end here and rebuild the measures (framing two's prescription) — or, most likely, some contested mixture of both. The companion brief on AI and education takes that question up directly.

7. What to hold onto

A short, honest summary of a contested field:

1. Definitions are real but their edges are cultural and untaught. Deliberate deception

is recognized as wrong nearly everywhere; the conventions of citation and the collaboration/collusion line are genuinely variable and frequently the source of unfair accusations. Both halves matter.

1. Cheating is common, the exact rate is unknown. Majorities admit some cheating over a

career (McCabe/ICAI); contract cheating runs low-single-digits historically and materially higher in recent samples (Newton). All figures are self-reported floors, not measurements, and country variation is large.

1. The causes are structural as much as individual — stakes, opportunity, peer norms,

performance-goal framing, neutralization, and assessment–learning mismatch — which is exactly why a pure-vice account is incomplete and why the two framings both find evidence.

1. It is ancient. The keju and the honor code already contain every modern dynamic: high

stakes breed cheating, and detection is a permanent arms race.

1. Two serious framings, left unadjudicated. Integrity as a genuine moral/epistemic good

(ICAI, Bertram Gallant) and "cheating" as a predictable artifact of high-stakes, extrinsically-motivated, badly-measured credentialing (Goodhart, the assessment-design and signaling critics). They converge on practice and diverge on meaning, and the meaning is a value question.

1. Detection is failing. Similarity detection never caught bespoke cheating, proctoring

provoked a justified backlash, and AI breaks both detection and the definition of cheating itself — pushing the field toward the assessment-redesign and culture-of-integrity moves that both framings, for different reasons, end up endorsing.

Sources

Prevalence and the integrity field:

• International Center for Academic Integrity, Statistics — https://www.academicintegrity.org/statistics/
• ICAI, Facts & Statistics — https://www.academicintegrity.org/aws/ICAI/pt/sp/facts
• Plagiarism.org, "Plagiarism: Facts & Stats" (McCabe high-school/undergraduate figures, ETS historical) — https://www.plagiarism.org/article/plagiarism-facts-and-stats
• McCabe, "Cheating among college and university students: A North American perspective" — https://www.researchgate.net/publication/228654731CheatingamongcollegeanduniversitystudentsANorthAmericanperspective
• Newton, P. (2018), "How Common Is Commercial Contract Cheating in Higher Education and Is It Increasing? A Systematic Review," Frontiers in Education — https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2018.00067/full
• Newton, "How Prevalent is Contract Cheating and to What Extent are Students Repeat Offenders?" — https://www.researchgate.net/publication/316275881HowPrevalentisContractCheatingandtoWhatExtentareStudentsRepeat_Offenders
• phys.org, "Latest study reveals sharp rise in essay cheating globally" — https://phys.org/news/2018-08-latest-reveals-sharp-essay-globally.html

Tricia Bertram Gallant / teaching-and-learning approach:

• ICAI author bio, Tricia Bertram Gallant — https://academicintegrity.org/integrity-matters-author-bio/4-tricia-bertram-gallant
• "Academic Integrity as a Teaching & Learning Issue: From Theory to Practice," Theory Into Practice 56(2) — https://www.tandfonline.com/doi/abs/10.1080/00405841.2017.1308173
• Bertram Gallant & Rettinger, The Opposite of Cheating: Teaching for Integrity in the Age of AI (2025) — https://www.amazon.com/Opposite-Cheating-Teaching-Integrity-Engaging/dp/0806194952
• UCSD, "How to Teach in the Age of AI" — https://today.ucsd.edu/story/how-to-teach-in-the-age-of-ai

Cross-cultural definitions:

• "Cross-Cultural Differences in Plagiarism: Fact or Fiction?" (Duquesne Law Review) — https://dsc.duq.edu/cgi/viewcontent.cgi?article=3836&context=dlr
• EBSCO Research Starters, "Cross-Cultural Perspectives on Source Referencing and Plagiarism" — https://www.ebsco.com/research-starters/ethnic-and-cultural-studies/cross-cultural-perspectives-source-referencing-and
• "Deconstructing Plagiarism: International Students and Textual Borrowing Practices" — https://www.researchgate.net/publication/233052529DeconstructingPlagiarismInternationalStudentsandTextualBorrowingPractices
• "Culture and Unethical Conduct: Understanding the Impact of Individualism and Collectivism on Actual Plagiarism" — https://www.researchgate.net/publication/228237737CultureandUnethicalConductUnderstandingtheImpactofIndividualismandCollectivismonActualPlagiarism
• Turnitin blog, "Understanding cultural differences in plagiarism" — https://www.turnitin.com/blog/cultural-differences-in-plagiarism

Why students cheat:

• Ives, B. (2020), "Your Students are Cheating More Than You Think They Are. Why?" — https://files.eric.ed.gov/fulltext/EJ1250265.pdf
• "Theoretical Frameworks for Academic Dishonesty" (Univ. of Michigan) — https://quod.lib.umich.edu/t/tia/17063888.0028.018/
• "Using Neutralization Theory to Understand Cheating" — https://distantreader.org/stacks/journals/cieatasu/cieatasu-190.pdf
• "Under Pressure to Achieve? The Impact of Type and Style of Task Instructions on Student Cheating," PMC — https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6702756/
• "Academic dishonesty and its relations to peer cheating and culture: A meta-analysis of the perceived peer cheating effect," ScienceDirect — https://www.sciencedirect.com/science/article/pii/S1747938X22000240
• "Drivers of academic dishonesty: situational pressures versus student motivation" — https://www.tandfonline.com/doi/full/10.1080/02673843.2025.2560649

History:

• World History Encyclopedia, "The Civil Service Examinations of Imperial China" — https://www.worldhistory.org/article/1335/the-civil-service-examinations-of-imperial-china/
• ChinaFile, "Heirs of Fairness?" — https://www.chinafile.com/reporting-opinion/out-of-school/heirs-of-fairness
• The World of Chinese, "Three Famous Chinese Imperial Examination Cheats" — https://www.theworldofchinese.com/2021/06/three-famous-chinese-imperial-examination-cheats/

Country variation:

• "Academic Cheating in Austria, Portugal, Romania and Spain: A Comparative Analysis" (SAGE) — https://journals.sagepub.com/doi/pdf/10.2304/rcie.2006.1.3.198
• Good Authority, "Here's the academic evidence on cheating in post-communist countries" — https://goodauthority.org/news/heres-the-academic-evidence-on-cheating-in-post-communist-countries/
• WENR/WES, "Academic Fraud, Corruption, and Implications for Credential Assessment" — https://wenr.wes.org/2017/12/academic-fraud-corruption-and-implications-for-credential-assessment
• "Cheating and Plagiarism in Higher Education" (Higher Education in Russia and Beyond) — https://www.researchgate.net/publication/329964165CheatingandPlagiarisminHigherEducationHigherEducationinRussiaandBeyond_HERB

Structural critique:

• Bryan Caplan, The Case Against Education (review/summary), Cato Institute — https://www.cato.org/cato-journal/spring/summer-2018/case-against-education-why-education-system-waste-time-money-bryan
• "Signals, not Learning: A review of Bryan Caplan's The Case Against Education" — https://safs.ca/newsletter/signals-not-learning-a-review-of-bryan-caplans-the-case-against-education/
• Edutopia (Denise Pope), "Academic Integrity: Cheat or Be Cheated?" — https://www.edutopia.org/blog/academic-integrity-cheat-or-be-cheated-denise-pope
• Wikipedia, "Academic dishonesty" (overview, teacher-side/Campbell's Law) — https://en.wikipedia.org/wiki/Academic_dishonesty

Detection, proctoring, and AI:

• Turnitin, academic integrity / AI writing solutions — https://www.turnitin.com/solutions/topics/ai-writing/
• K-12 Dive, "Turnitin admits there are some cases of higher false positives in AI writing detection tool" — https://www.k12dive.com/news/turnitin-false-positives-AI-detector/652221/
• Inside Higher Ed, "Turnitin faces new questions about efficacy of plagiarism detection software" — https://www.insidehighered.com/news/2015/07/14/turnitin-faces-new-questions-about-efficacy-plagiarism-detection-software
• MIT Technology Review, "Software that monitors students during tests perpetuates inequality and violates their privacy" — https://www.technologyreview.com/2020/08/07/1006132/software-algorithms-proctoring-online-tests-ai-ethics/
• Vice, "Schools Are Abandoning Invasive Proctoring Software After Student Backlash" — https://www.vice.com/en/article/schools-are-abandoning-invasive-proctoring-software-after-student-backlash/
• The Conversation, "Online exam monitoring can invade privacy and erode trust at universities" — https://theconversation.com/online-exam-monitoring-can-invade-privacy-and-erode-trust-at-universities-149335
• "Survey on Plagiarism Detection in Large Language Models: The Impact of ChatGPT and Gemini on Academic Integrity" (arXiv) — https://arxiv.org/pdf/2407.13105
• Paperpal, "Can Turnitin Detect ChatGPT? AI Detection in Academia" — https://paperpal.com/blog/news-updates/can-turnitin-detect-chatgpt

Foundations brief 03. Neutral by design: it reports the contested definitions, the weak self-report data, the multiple causal mechanisms, the deep history, and the two serious framings without adjudicating between them. The reform thesis lives elsewhere in the atlas.