How to Improve Academic Integrity: Evidence-Based Approaches

TL;DR

Improving academic integrity requires multi-factor intervention at multiple levels:

• Country level: legislation, regulator capacity, cross-border coordination
• Institutional level: detection deployment, disclosure infrastructure, penalty frameworks, honor codes
• Course level: assessment redesign, instructor training, integrity-aware syllabus design

Single-factor approaches are weak; multi-factor combinations work.

Country-level interventions

Specific contract cheating legislation

The most effective single national policy intervention. Templates:

• Ireland 2019 — Qualifications and Quality Assurance Act, QQI enforcement
• Australia 2020 — TEQSA Amendment, public provider list, AUD 100k fines
• UK 2022 — Skills and Post-16 Education Act section 80, OfS enforcement

All three jurisdictions score 100 on the AMI R-Score Legislation sub-component (the maximum). The bans target essay mill providers; individual students are handled by institutional frameworks.

Regulator capacity

Specific bans without enforcement capacity produce limited effect. Effective regulators:

• Investigation and prosecution authority
• Cross-border cooperation arrangements
• Industry intelligence gathering
• Coordination with detection vendors
• Public reporting of enforcement activity

TEQSA (Australia), QQI (Ireland), OfS (UK) provide the operational templates.

Public lists of known providers

TEQSA's distinctive feature: a public list of 2,300+ known contract cheating providers. Benefits:

• Enables institutional-level blocking
• Signals to students that authorities are monitoring
• Provides evidence for enforcement
• Demonstrates regulator accountability

Other regulators (QQI, OfS) maintain internal lists; the Australian public approach is unique and effective.

Cross-border coordination

Essay mill providers operate across jurisdictions. Effective enforcement requires:

• Regulator-to-regulator cooperation (TEQSA, QQI, OfS coordinate actively)
• Information sharing on identified providers
• Coordinated enforcement actions
• Detection vendor partnerships across markets

Institutional-level interventions

Detection deployment

Universal detection tool deployment across the university sector:

• Plagiarism detection — Turnitin, iThenticate, Compilatio, Antiplagiat, CopyKiller, JSA
• AI detection — Turnitin AI, GPTZero, Originality.ai, Copyleaks
• Image manipulation detection — Imagetwin, PaperWatcher for research

Detection alone does not deter (Scarfe 94% miss rate), but provides:

• Evidence for institutional misconduct findings
• Visible institutional commitment
• Triage for instructor review
• Baseline misconduct data for analysis

Mandatory institutional disclosure

Requiring institutions to publish misconduct statistics:

• Creates accountability
• Enables cross-institutional benchmarking
• Shifts student perception of peer cheating
• Provides evidence for policy effectiveness

Australia and the UK lead on Disclosure sub-scores (90 and 85 respectively). Other countries lag substantially.

Clear penalty frameworks

Consistent, applied penalty structures:

• Defined sanctions for specific misconduct types
• Consistent application across instructors
• Documented appeal procedures
• Public reporting on aggregate outcomes

Penalty severity matters less than penalty consistency. McCabe research showed that perceived likelihood of being caught matters more than penalty severity.

Honor code culture

The strongest single institutional intervention in McCabe research. Active honor codes feature:

• Student-led integrity processes
• Visible integrity commitments (pledges, ceremonies)
• Strong peer-norm reinforcement
• Disclosure transparency

US institutions with strong honor codes (University of Virginia, Princeton, Washington & Lee) show consistently lower cheating rates than peer institutions.

Integrity offices

Resourced integrity infrastructure:

• Full-time staff
• Coordinated investigation processes
• Faculty training programs
• Student outreach
• Policy development capacity

Post-COVID, institutional integrity office expansion has been retained at most universities. The expansion has improved both detection and process consistency.

Course-level interventions

Assessment redesign

The most discussed intervention post-ChatGPT. Effective redesign:

#### Demonstrable understanding formats

• Oral examinations
• Viva (oral defence)
• Live problem-solving
• OSCEs and practical demonstrations
• Project work with iterative review

These formats make AI use and contract cheating largely irrelevant — they assess what AI cannot perform.

#### Authentic assessment

• Real-world problem application
• Context-specific scenarios
• Locally relevant case studies
• Topic-specific data analysis

Authentic assessment ties the work to specific context AI tools cannot provide.

#### Varied question sets

• Per-student question variation
• Multiple equivalent problem sets
• Randomised question banks
• Personalised problem instances

Variation makes collusion harder and reduces benefit from question sharing.

#### Process-evidence assessment

• Required drafts at multiple stages
• Brainstorming notes submission
• Annotated bibliographies
• Step-by-step problem work

Process evidence demonstrates that the student engaged with the work — harder to fake with AI or contract cheating than the final product alone.

Instructor training

Faculty academic integrity training programmes:

• Identifying common misconduct patterns
• Using detection tools effectively
• Designing AI-resistant assessment
• Handling suspected cases appropriately
• Communicating integrity expectations clearly

Post-COVID expansion of these programmes has improved instructor capability across many universities.

Clear syllabus communication

Specific course-level expectations:

• AI use policy (permitted, prohibited, with disclosure)
• Collaboration boundaries (study groups OK; identical answers not)
• Specific examples of acceptable and unacceptable behaviour
• Process expectations (drafts, brainstorming)
• Consequences of misconduct

Clear communication reduces both honest mistakes and rationalization-based misconduct.

What works at the country level — what AMI data shows

Q1 countries combine multiple factors

The seven Q1 countries (Australia, UK, Ireland, Canada, NZ, Netherlands, US) all have:

• Some form of integrity-specific legislation (three with essay mill bans, others with research integrity laws)
• Strong detection deployment (R_det 65-90)
• Mature disclosure infrastructure (R_dis 40-90)
• Established penalty frameworks (R_pen 55-80)

The multi-factor combination produces the Q1 effect.

Q3 countries lack multiple factors

The twelve Q3 countries lack several factors simultaneously:

• Limited or no specific legislation
• Partial detection deployment
• Minimal disclosure
• Inconsistent penalty enforcement

The Q3 placement is not driven by any single missing factor but by the combination.

What does not work

Single-factor interventions

• Awareness campaigns alone — minimal effect
• Punishment severity increases without detection improvement — small effect
• Detection technology alone without enforcement — limited effect
• New policies without implementation infrastructure — symbolic effect

Symbolic policies without enforcement

Countries with strong written policies but weak enforcement (Russia's Antiplagiat + minimal sanctioning of Dissernet-identified plagiarists is the prototype) show poor R-Scores despite formal policy infrastructure.

Student-targeted criminalization

The supply-side legislative approach (targeting providers rather than students) has proven more effective than student-targeted approaches. No major jurisdiction criminalizes student use of essay mill services.

Practical pathways

For policymakers in Q3 countries

The most impactful single national reform: adopt essay mill legislation on the Irish/Australian/UK model. The legislation alone would lift Legislation sub-score from current 10-12 to 100. Combined with detection deployment and disclosure infrastructure, R-Score can reach Q1 thresholds within 5-10 years.

For policymakers in Q4 countries

The Q4 transition path: build disclosure infrastructure first. Q4 countries typically have moderate Detection and Legislation but very low Disclosure. Mandatory institutional misconduct reporting addresses the principal gap.

For institutional administrators

Multi-factor approach:

1. Detection deployment (where not already present)
2. Honor code culture investment
3. Assessment redesign training for faculty
4. Integrity office resourcing
5. Disclosure transparency

For instructors

Course-level approach:

1. Clear syllabus communication
2. Authentic assessment design
3. Process-evidence assessment
4. Variation in question sets
5. Engagement with institutional detection infrastructure

Sources

• AMI v1.5 dataset showing Q1 vs Q3 R-Score components
• McCabe (multiple) honor code and intervention research
• Newton (2018) systematic review
• Scarfe (2024) detection finding
• National legislative documentation (Ireland, Australia, UK)
• Institutional integrity office case studies

Full methodology | Download dataset