⚖️ Ethics in Computing

Research Ethics

Why Research Ethics Matters

"Ethics is not a constraint on research; it is the condition of its legitimacy." — Paul Ricœur

Research involving human participants, data, or publication carries moral and legal obligations. Ethical failures invalidate results, harm participants, and undermine public trust in science.

Core Ethical Principles (Belmont Report)

Principle	Meaning	Application
Respect for Persons	Autonomy, informed consent, protection of vulnerable	Voluntary participation, right to withdraw
Beneficence	Maximize benefits, minimize harms	Risk-benefit analysis, safety monitoring
Justice	Fair distribution of burdens/benefits	Equitable selection, access to benefits

1. Institutional Review Boards (IRB) / Research Ethics Committees (REC)

When Is Review Required?

Category	Review Level	Examples
Exempt	Minimal risk, specific categories	Anonymous surveys, educational tests
Expedited	Minimal risk, no vulnerable groups	Blood draws, voice recordings, minor changes
Full Board	>Minimal risk, vulnerable populations	Clinical trials, deception, children/prisoners

IRB Process

graph TD A[Research Protocol] --> B{Exempt?} B -->|Yes| C[Exempt Determination] B -->|No| D{Minimal Risk?} D -->|Yes| E[Expedited Review] D -->|No| F[Full Board Review] E --> G[Approval / Modifications] F --> G G --> H[Informed Consent] H --> I[Recruitment] I --> J[Data Collection] J --> K[Analysis & Reporting]

Common IRB Pitfalls

Issue	Consequence	Prevention
Incomplete application	Delays/rejection	Use IRB checklist, consult early
Inadequate consent form	Non-compliance	Template + plain language review
Missing risk mitigation	Rejection	Pre-identify all risks
No data management plan	Conditional approval	Plan storage, sharing, destruction

Required Elements (8 Basic + 6 Additional per Common Rule)

Basic Elements: 1. Purpose — Research aims, duration, procedures 2. Risks — Foreseeable physical, psychological, social, legal, economic 3. Benefits — Direct (to participant) and indirect (to society) 4. Alternatives — Other options (e.g., standard treatment) 5. Confidentiality — Data protection, who sees data 6. Compensation — Payment, injury coverage 6. Contacts — Researcher + IRB for questions/complaints 8. Voluntary — Right to refuse/withdraw without penalty

Additional Elements (when applicable): - Unforeseeable risks (e.g., genetic findings) - Circumstances for termination - Additional costs - Consequences of withdrawal - Significant new findings - Commercial use / profit sharing

Principle	Implementation
Readability	≤8th grade level, avoid jargon
Modularity	Separate consent for distinct activities
Ongoing	Re-consent for new findings/procedures
Cultural sensitivity	Translate, respect community norms
Capacity	Assess understanding, use surrogates if needed

Special Populations — Enhanced Protections

Population	Key Safeguards
Children	Parental permission + child assent (age 7+)
Prisoners	Minimal risk only, no coercion, independent advocate
Pregnant women	Fetal risk assessment, no direct benefit required
Cognitively impaired	Legally authorized representative, ongoing assent
Students/employees	No academic/professional penalty for non-participation

3. Data Handling & Privacy

Data Lifecycle

graph LR A[Collection] --> B[Processing] --> C[Storage] --> D[Analysis] --> E[Sharing] --> F[Archiving/Disposal] A --> A1[Consent + IRB] B --> B1[De-identify + Code + Access Log] C --> C1[Encrypt + Access Control] D --> D1[Access Control + Logs] E --> E1[License Terms + DSA] F --> F1[Retain/Destroy]

De-identification Standards (HIPAA Safe Harbor)

Direct Identifiers to Remove (18 categories): 1. Names 2. Geographic subdivisions < state 3. Dates (birth, admission, discharge, death) 4. Phone numbers 5. Fax numbers 6. Email addresses 7. Social Security numbers 8. Medical record numbers 8. Health plan beneficiary numbers 10. Account numbers 11. Certificate/license numbers 12. Vehicle identifiers 13. Device identifiers 14. Web URLs 15. IP addresses 16. Biometric identifiers 17. Full-face photos 18. Any unique identifying number/code

Approach	When	Pros	Cons
Controlled access	Sensitive data	Protects privacy	Limits reuse
Synthetic data	High risk	Zero re-identification	May lose utility
Differential privacy	Aggregate release	Mathematical guarantees	Noise reduces accuracy
Data enclaves	Very sensitive	Full analysis possible	Expensive, limited access

FAIR Principles

Principle	Action
Findable	Persistent ID (DOI), rich metadata
Accessible	Standard protocol, auth if needed
Interoperable	Standard vocabularies, formats
Reusable	Clear license, provenance, domain standards

4. Publication Ethics

Authorship Criteria (ICMJE)

All four required for authorship: 1. Substantial contributions to conception/design OR acquisition/analysis/interpretation 2. Drafting or critical revision for intellectual content 3. Final approval of version to be published 4. Agreement to be accountable for all aspects

Contributors who don't meet all four → Acknowledge in "Contributions" section

Common Authorship Issues

Issue	Description	Solution
Ghost authorship	Uncredited writer	Disclose professional writers
Guest authorship	Honorary addition	Apply ICMJE criteria strictly
Gift authorship	Favor/reciprocity	Same as guest
Author order disputes	Credit allocation	Predefine order, document contributions

Publication Misconduct

Type	Definition	Detection
Plagiarism	Unattributed text/ideas	iThenticate, CrossCheck
Self-plagiarism	Reuse without citation	Text recycling guidelines
Salami slicing	Split one study into many	Same dataset/cohort flag
Data fabrication	Invented data	Statistical anomalies, replication
Data falsification	Manipulated results	Raw data audit, image analysis
Citation manipulation	Excessive self-cites, coercion	Bibliometric analysis

Image Integrity Standards

Manipulation	Acceptable?	Notes
Brightness/contrast (whole)	✅	Apply uniformly
Cropping	✅	Don't remove relevant info
Combining gels	❌	Unless explicit with lines
Cloning/erasing	❌	Never
Selective enhancement	❌	Misrepresents data

5. Research Misconduct & Accountability

Definitions (US ORI / UK Concordat)

Category	Definition
Fabrication	Making up data/results
Falsification	Manipulating materials/equipment/processes
Plagiarism	Appropriation without attribution
Questionable practices	Selective reporting, p-hacking, HARKing

Responsible Conduct of Research (RCR) Training

Core Topics (NIH/NSF required): - Human subjects protection - Animal welfare - Data management - Publication ethics - Mentorship - Conflict of interest - Collaborative research - Peer review - Research misconduct - Safe laboratory practices

Whistleblowing & Reporting

Step	Action
1. Document	Factual, specific, dated
2. Consult	Trusted mentor, ombudsperson
3. Report internally	PI → Department chair → Research integrity officer
4. External (if needed)	ORI (US), UKRIO (UK), journal, funder
5. Protection	Whistleblower laws, institutional policies

6. Emerging Ethical Challenges

Big Data & AI in Research

Challenge	Ethical Response
Consent at scale	Broad consent, dynamic consent platforms
Algorithmic bias	Diverse training, fairness audits
Re-identification risk	Differential privacy, synthetic data
Black box models	Explainable AI, model cards

Open Science

Practice	Benefit	Challenge
Pre-registration	Reduces p-hacking	Inflexible for exploratory work
Open data	Reproducibility	Privacy, competitive disadvantage
Open code	Transparency	Maintenance burden
Open peer review	Accountability	Reviewer reluctance

Global Research Ethics

Issue	Framework
Standards export	CIOMS guidelines — adapt to local context
Benefit sharing	Nagoya Protocol, fair benefit agreements
Capacity building	Equitable partnerships, not helicopter research
Dual use	WHO guidance, institutional oversight

7. Practical Checklists

Pre-Study Checklist

[ ] Research question justified, feasible
[ ] Literature review complete (avoids duplication)
[ ] Protocol written, version-controlled
[ ] IRB/REC approval obtained
[ ] Informed consent forms approved
[ ] Recruitment materials approved
[ ] Data management plan (DMP) completed
[ ] Statistical analysis plan (SAP) finalized
[ ] Sample size justified (power analysis)
[ ] Risk mitigation documented
[ ] Compensation plan approved
[ ] COI disclosures collected
[ ] Training certificates current (CITI, GCP, etc.)

During Study Checklist

[ ] Consent documented for each participant
[ ] Deviations reported to IRB promptly
[ ] Safety monitoring per protocol
[ ] Data quality checks scheduled
[ ] Enrollment logs maintained
[ ] Interim analyses per SAP only
[ ] Participant withdrawal handled correctly

Post-Study Checklist

[ ] Data locked, backed up
[ ] De-identification verified
[ ] Statistical analysis per SAP
[ ] Results reported completely (CONSORT, STROBE, PRISMA)
[ ] Authorship finalized per ICMJE
[ ] COI statements updated
[ ] Data sharing plan executed (repository DOI)
[ ] IRB closure report submitted
[ ] Records retained per policy (typically 3-7 years post-publication)

Resources

Key Guidelines & Frameworks

Document	Scope	Link
Belmont Report (1979)	US human subjects	HHS
Declaration of Helsinki (2013)	Medical research	WMA
CIOMS Guidelines (2016)	International health research	CIOMS
Singapore Statement (2010)	Research integrity	WCRI
Hong Kong Principles (2019)	Research assessment	WCRI
ICMJE Recommendations	Publication ethics	ICMJE
CONSORT / STROBE / PRISMA	Reporting standards	EQUATOR

Institutional Resources

CITI Program — Online RCR training
NIH Office of Research Integrity — Guidance, case studies
UKRIO — UK Research Integrity Office
COPE — Committee on Publication Ethics
Retraction Watch — Database of retractions