Unit 5: Research Ethics
Learning Objectives
- Explain the moral foundations of research ethics and the competing responsibilities that researchers carry
- Distinguish research misconduct from questionable research practices and honest error, with reference to documented cases
- Evaluate ethical issues arising from research involving human participants, including informed consent, anonymity, and proportionality of risk
- Describe the institutional ethics approval process and identify the appropriate pathway for a given research project
- Assess how UK GDPR obligations and publication ethics apply to research data, authorship, and collaborative work
- Critically evaluate the responsible use of generative AI in research and apply the AI4People framework to AI systems
- Explain how intellectual property rights affect research outputs, software, data, and academic writing
- Apply ethical frameworks to realistic research dilemmas and justify decisions through structured reasoning
Core Input — The Moral Foundations of Research Ethics
Read through each tab. The ideas introduced here underpin every subsequent activity in this unit.
A common reaction to research ethics training is "I'm a good person — why do I need to worry about this?" The question is understandable, but it rests on a false premise: that research ethics is primarily about preventing bad people from doing bad things deliberately.
In reality, ethical problems in research arise far more often from:
- Socially accepted practice later deemed wrong — practices that were once standard (such as conducting research on patients without explicit consent) are now recognised as violations
- Accident and oversight — a researcher genuinely fails to anticipate a harm
- Misdemeanours — small wrongs that seem easy to justify in the moment ("everyone does this")
- Missing something — an ethical dimension of the research simply wasn't noticed
- Complicity — being involved in someone else's ethical violation without initiating it
- Genuine moral ambiguity — the right course of action is genuinely unclear
- Self-deception — persuading oneself that a convenient choice is also the ethical one
Research ethics is therefore not a rulebook to be followed mechanically. It is a set of principles that require active, ongoing reasoning — before, during, and after a research project.
Researchers carry obligations to multiple parties simultaneously. Recognising all of them — and acknowledging when they pull in different directions — is central to ethical practice.
To whom do researchers owe the truth?
- Research participants — out of basic respect for persons and their autonomy
- The public — who fund much research through taxation and who rely on its integrity
- Institutions and universities — who stake their reputations on researchers' conduct
- Funding bodies — who have distributed public or charitable resource on the expectation of honest reporting
- Colleagues and collaborators — whose own work builds on shared findings
- Future generations — who inherit both the knowledge and the consequences of how it was obtained
- Yourself — intellectual honesty is part of professional integrity
The moral foundations underlying these duties include:
- Advancing knowledge and understanding for the benefit of society
- Doing good for people, the planet, and future generations
- Respecting the dignity and autonomy of individuals
- Fulfilling obligations to animals used in research
- Meeting wider obligations to society and democratic institutions
These foundations can come into tension. The duty to advance knowledge may conflict with the duty to protect participants. The obligation to a funder may conflict with the obligation to report negative results honestly. Ethical reasoning requires navigating these tensions — not pretending they do not exist.
Three ethical theories appear repeatedly in discussions of research ethics. None of them resolves every dilemma on its own; each illuminates a different aspect of an ethical situation.
Consequentialism holds that the right action is the one that produces the best outcomes overall. In research, this means weighing the potential benefits of a study (new knowledge, clinical advances, social understanding) against the potential harms to participants, society, or the research record. Consequentialist thinking is useful for risk-benefit analysis — but it can be used to justify serious harms to individuals on the grounds that the majority will benefit.
Deontology holds that some duties are binding regardless of consequences. Kant's categorical imperative demands that we treat persons as ends in themselves, never merely as means. In research, this grounds the requirements for informed consent and honesty: lying to participants or fabricating data is wrong even if the results would benefit many people. Deontological thinking protects individuals from being sacrificed for aggregate gain — but can seem inflexible when outcomes genuinely matter.
Virtue ethics asks not "what rule applies?" or "what produces the best outcome?" but "what would a person of good character do?" A virtuous researcher is honest, careful, courageous enough to report unwelcome results, and humble enough to acknowledge error. Virtue ethics is useful when rules and calculations fail to capture what is really at stake — but gives less practical guidance on specific decisions.
In practice, ethical dilemmas in research often arise precisely because these frameworks give different verdicts. A full ethical analysis considers all three perspectives before reaching a judgement.
Research Integrity and Misconduct
Expand each section to read the content, then try the Misconduct Detective exercise below.
The most serious forms of research misconduct are collectively known as FFP:
- Fabrication — inventing data, results, or findings that were never observed. This includes creating fictional participants, experiments that never took place, or citations that do not exist.
- Falsification — manipulating or altering data, images, instruments, or processes in a way that misrepresents what was actually found. This includes selective cropping of microscopy images to remove inconvenient features, or adjusting data points to achieve statistical significance.
- Plagiarism — representing another person's work, words, ideas, or data as one's own without appropriate attribution. This includes text lifted without citation, ideas gleaned from peer review and used as one's own, and self-plagiarism (republishing one's own prior work without disclosure).
Retraction Watch (retractionwatch.com) is a blog that reports on retractions of scientific papers. It provides a public record of cases where journals have removed papers following identified misconduct or serious error. A notable example: a 2014 paper by Haruko Obokata et al. claiming the discovery of STAP stem cells was retracted after image manipulation and data irregularities were identified. In 2024, Springer Nature retracted 2,923 papers — out of over 482,000 published — highlighting that misconduct, while a small proportion of the literature, is not rare in absolute terms.
It is important to distinguish proven misconduct from honest error and from unresolved allegations. A retraction does not always indicate deliberate wrongdoing; some retractions correct genuine mistakes identified in good faith.
Between clear misconduct and fully responsible research lies a wide zone of questionable research practices (QRPs) — behaviours that fall short of fabrication or plagiarism but still distort the scientific record.
- P-hacking — running multiple statistical tests or sub-group analyses and reporting only those that yield p < 0.05, without disclosing the others. With enough tests, a false positive is almost inevitable by chance alone.
- HARKing (Hypothesising After Results are Known) — presenting a hypothesis as though it was formulated before data collection, when in fact it was generated by inspecting the results. This gives the false impression of a pre-planned confirmatory study.
- Selective reporting — publishing only the results that support the desired conclusion, omitting null or negative findings. This contributes to publication bias and gives a misleading picture of the evidence base.
- Image manipulation — adjusting brightness, contrast, or cropping in ways that misrepresent what the image actually shows, without disclosure.
QRPs are harmful because they pollute the scientific record with findings that cannot be replicated, wasting future researchers' time and potentially leading to flawed policy or clinical decisions. Pre-registration — publicly declaring hypotheses and analysis plans before data collection — is one response to QRPs that is gaining wide adoption.
Understanding why research misconduct and QRPs occur requires understanding the structural pressures researchers face. These pressures do not excuse unethical behaviour, but ignoring them makes it harder to resist them:
- Publish or perish — academic careers depend on publication volume and citation counts, creating incentives to cut corners
- Positive results bias — journals are more likely to publish findings that confirm hypotheses, discouraging honest reporting of null results
- Grant competition — funding bodies want evidence of impact, which may distort how results are framed
- Being first — competitive fields create pressure to publish quickly before rigorous checking
- Supervisory pressure — junior researchers may feel unable to challenge a senior colleague's instructions
Recognising these pressures is the first step to resisting them. Institutional cultures that reward transparency, reproducibility, and the honest reporting of negative results reduce the conditions under which misconduct flourishes.
Misconduct Detective
The passage below is a fictional methods excerpt. It contains five ethical problems. Click any highlighted phrase to reveal an explanation. Find all five to complete the exercise.
A researcher collects data, finds that one sub-group analysis is significant, then writes the paper as if that sub-group comparison was the primary hypothesis from the start. This is an example of:
Which of the following best describes the purpose of Retraction Watch?
Research Involving People and Ethics Approval
Read both tabs, then complete the Ethics Screener exercise.
The ethical framework for research involving human participants has its origins in responses to documented historical abuses. The Nuremberg Code (1947), produced in the aftermath of Nazi medical experiments, established ten principles governing human subject research. Its first and most fundamental principle is that voluntary consent is absolutely essential. The Declaration of Helsinki (1964, revised multiple times) extended these principles specifically to clinical and biomedical research.
Informed consent requires that participants:
- Receive clear information about the purpose, procedures, and foreseeable risks of the study
- Understand what participation involves, including any data collected and how it will be used
- Agree to participate voluntarily, without coercion or undue incentive
- Retain the right to withdraw at any point without penalty
Informed and voluntary are both demanding criteria. Consent is not genuinely voluntary if there is a significant power imbalance — for example, between a lecturer and their own students, or a clinician and their patients. In these situations, additional safeguards are required.
Anonymity means that no one — including the researcher — can link data to a specific individual. Confidentiality means the researcher knows who provided data but will not disclose it. These are frequently confused. A survey where participants submit responses without names is anonymous; an interview where the researcher knows the participant but will not name them in reports is confidential. Both carry distinct obligations.
Research must not expose participants to disproportionate risk. Risks may be physical, psychological (distress, anxiety, retraumatisation), social (reputational harm, strained relationships), or reputational. The potential benefits of the research must be proportionate to the risks imposed on participants — a judgment that is especially important when studying vulnerable populations such as children, people with cognitive impairments, or individuals in custodial settings.
When deception is necessary (for example, in some psychological experiments where prior knowledge of the hypothesis would invalidate responses), it must be scientifically justified, approved by an ethics committee, and followed by a full debriefing in which participants learn the true purpose of the study and can withdraw their data if they wish.
Institutional ethics review exists to ensure that research meets agreed ethical standards before it begins, not after harm has occurred. The review process is not a bureaucratic obstacle — it is a structured opportunity to identify risks that researchers, who are naturally invested in their own projects, may overlook.
The approval spectrum typically runs from lower to higher risk:
- Self-certification — the researcher and supervisor confirm that the research involves only secondary data, poses minimal risk, and meets established criteria for exemption from formal review
- Supervisor or departmental sign-off — required when there are some risk factors but they are manageable; the supervisor takes responsibility for oversight
- Faculty or university ethics committee — required for higher-risk research involving primary data collection from human participants, sensitive topics, NHS data, children, or other vulnerable populations
Primary data is data you collect yourself — through interviews, surveys, experiments, or observation. Secondary data is data that already exists — datasets provided by a supervisor, publicly available repositories, or published records. Secondary data usually requires lighter-touch approval, but this does not mean it is ethically uncomplicated: the original consent given for one purpose does not automatically extend to a new research use.
Your ethics form will typically ask you to declare:
- Who your participants are and how they will be recruited
- What data you will collect and how it will be stored and protected
- Whether any special category data (see Activity 4) will be processed
- How you will obtain and document informed consent
- What risks exist and how you will mitigate them
Plans change. If the scope of your research changes in ways that affect the ethical analysis, you should seek an amendment to your approval — not simply proceed. Your supervisor is your first point of contact for any ethical uncertainty.
Ethics Screener
Read the project description, then answer the five questions. The screener will recommend an ethics approval pathway. You can switch the scenario to test a different case.
Data Ethics, Collaboration, and Publication
Expand each section, then try the Authorship exercise.
If your research involves information about living, identifiable individuals, it is subject to data protection law. In the UK, this means the UK General Data Protection Regulation (UK GDPR) and the Data Protection Act 2018.
Personal data is any information that relates to an identified or identifiable natural person. This includes names, email addresses, student IDs, IP addresses, and — crucially — any combination of attributes that could identify someone even without a name. Special category data includes data about health, ethnicity, religion, political views, sexual orientation, genetic data, and biometric data. Processing special category data requires a higher legal justification and stricter safeguards.
Researchers typically rely on one of two lawful bases: public task (for university research in the public interest) or legitimate interests. Where participants are asked to consent to data processing, that consent must be freely given, specific, informed, and unambiguous — the same criteria that apply to research participation consent, but distinct from it.
Anonymisation removes all identifying information irreversibly, taking the data outside the scope of UK GDPR. Pseudonymisation replaces identifying attributes with codes — the data is still personal data because re-identification is possible with the key. Re-identification is a genuine risk: research has repeatedly shown that combinations of apparently innocuous attributes (age, postcode, occupation) can uniquely identify individuals in supposedly anonymous datasets.
UK GDPR's data minimisation principle means you should collect only the data you actually need for your research. Collecting additional data "in case it is useful later" is not compliant. Data should be stored securely, with access limited to those who need it, and disposed of when the retention period ends. Open science mandates from funders to share data must be balanced against participants' privacy: sharing a dataset openly is only permissible if the data is truly anonymised or if participants consented to open sharing.
Authorship on a research paper is not simply a reward for contribution — it is an ethical responsibility. An author is accountable for the work they have authored. The ICMJE criteria (International Committee of Medical Journal Editors, widely adopted beyond medicine) require that all four conditions be met for authorship:
- Substantial contribution to conception or design, or to data acquisition, analysis, or interpretation
- Drafting or critically revising the intellectual content of the work
- Final approval of the version to be published
- Agreement to be accountable for all aspects of the work
Gift authorship — naming someone as author when they have not met these criteria — is a form of misconduct. It is common in hierarchical settings where junior researchers feel unable to refuse a senior colleague's expectation of inclusion. Ghost authorship — omitting someone who has made substantial contributions — is the mirror-image problem. Honorary authorship (naming a department head or funder who played no intellectual role) is also a recognised QRP.
In student research, intellectual property questions arise frequently. Copyright in a dissertation typically belongs to the student. However, research conducted under a funded project, using institutional resources, or in collaboration with an industry partner may involve more complex ownership arrangements — always discuss this with your supervisor before the project begins.
Peer review is the mechanism by which the research community evaluates the quality and validity of new work before publication. It carries its own ethical obligations:
- Confidentiality — reviewer identities and manuscript contents are confidential. Using an idea encountered while reviewing a manuscript as material for your own work — without attribution — is plagiarism of unpublished material and a serious breach of the peer-review system.
- Conflicts of interest — reviewers should declare and recuse themselves from manuscripts where they have a personal or professional conflict (e.g., being a direct competitor, collaborator, or having a financial interest in the outcome).
- Fairness — reviewing a manuscript poorly to delay or prevent publication by a competitor is misconduct.
Duplicate publication — submitting the same paper to two journals simultaneously, or republishing work without adequate disclosure — wastes editorial resources and distorts the research record. Coercive citation — editors or reviewers demanding the addition of citations to their own work — is increasingly recognised as a form of publication misconduct.
When errors in published work are discovered, researchers have a responsibility to correct the record. An erratum corrects an honest error; a retraction withdraws a paper when its findings are unreliable; an expression of concern flags an ongoing investigation. Failure to correct a published error, once identified, is itself an ethical violation.
Who Deserves Authorship?
A research paper has five contributors. Based on the ICMJE criteria, decide whether each person qualifies as an Author or should receive an Acknowledgement only.
A senior professor asks to be listed as corresponding author on a paper. She provided the research group's lab space but had no role in designing, analysing, or writing the study. Under ICMJE criteria, she should be:
AI in Research
Read all three tabs, then complete the GenAI Classifier and AI Ethics Audit.
Generative AI tools — including large language models, code assistants, and image generators — are now widely available to researchers. Their use raises ethical questions that are separate from, though related to, questions about AI systems as research subjects.
It is useful to distinguish three distinct relationships a researcher might have with AI:
- Using AI as a research tool — to assist literature exploration, understand concepts, debug code, reformulate arguments, or improve clarity
- Studying AI as a research subject — investigating the properties, outputs, or impacts of AI systems
- Using AI-generated content as research data — for example, using LLM outputs as a corpus for linguistic analysis
These three roles carry different obligations and are often confused.
Responsible uses of generative AI as a research tool include:
- Using a chatbot to explain a concept you do not yet understand
- Using a code assistant to help debug or structure analysis scripts, which you then review and understand
- Using AI to identify candidate papers for a literature search, which you then read and evaluate yourself
- Using AI for grammar and clarity review of text you have written
Inappropriate uses include:
- Submitting AI-generated text as your own academic work without disclosure — this is malpractice under university regulations
- Citing references generated by an AI without verifying they exist — hallucinated citations are a well-documented failure mode of LLMs
- Uploading confidential participant data or proprietary datasets to an external AI service
- Using AI to generate synthetic research data presented as real observations
Permissibility depends on the applicable policy. Institutional regulations, assessment briefs, funder requirements, and publisher policies may differ. There is no single universal rule. You must check the policy that applies to your specific context. Where AI assistance was used, disclose it clearly and specifically.
The fundamental principle: AI does not bear responsibility for research. You do. The researcher is accountable for the accuracy, integrity, and ethical conduct of their work, regardless of what tools were used.
AI is deployed in consequential domains — credit scoring, recruitment, medical diagnosis, criminal justice, content moderation, and national security — at a scale and speed that exceeds human oversight in many cases. This makes it powerful enough to require explicit ethical scrutiny.
Computers are amoral entities: they optimise for the objective they are given, without any independent judgment about whether that objective is good. Any ethical constraints must be explicitly placed by developers, enforced by institutions, and demanded by regulators and users.
Unintended consequences can arise even from well-intentioned systems:
- Proxy variables — an algorithm may use a variable such as postcode or name that correlates strongly with a protected characteristic (ethnicity, gender, disability), effectively discriminating without directly using the protected characteristic
- Biased training data — a model trained on historical data inherits the biases present in that history. A hiring system trained on past appointments will learn to favour the demographic that was historically hired, and present this as an objective score. The bias is laundered rather than removed.
- Oversimplification — reducing a complex human situation to a set of measurable variables can produce predictions that are statistically accurate on average but seriously wrong — and unfair — in individual cases
Not all harmful AI applications are accidental. Surveillance systems, manipulation algorithms designed to maximise engagement regardless of harm, and dual-use technologies developed ostensibly for beneficial purposes but readily weaponised are examples of systems where harms may be known to developers but tolerated or deliberately chosen. Ethical research on AI must grapple with both unintended and not-so-unintended consequences.
The AI4People initiative, launched in 2018, developed a framework of five ethical principles for AI that has been widely adopted in policy and research discussions. Each principle addresses a distinct aspect of how AI systems can fail to respect human values.
AI systems should benefit humanity: improving well-being, advancing sustainable development, and contributing to the common good. This principle considers not just immediate users but future generations and the planet. It asks: for whom is this AI beneficial, and are those benefits distributed fairly?
AI must not infringe on privacy, undermine security, or enable new forms of discrimination or exploitation. Importantly, non-maleficence is not equivalent to beneficence: a system that causes no harm but also provides no genuine benefit fails this principle. Potential for misuse, overuse, and capability escalation all fall within its scope. Caution is required when deploying powerful AI in high-stakes domains.
AI systems should protect and enhance human autonomy — the capacity to make informed decisions and choose between alternatives. Systems that nudge, manipulate, or gradually replace human judgment undermine autonomy even when they produce individually correct decisions. The growth of AI must be restricted and reversible: autonomous power to harm, destroy, or deceive should never be vested in AI systems.
AI systems must promote prosperity and preserve solidarity. They must avoid unfair discrimination and must not create new injustices or amplify existing ones. Justice requires active attention to who benefits and who bears the risks of AI deployment — and whether vulnerable groups are systematically disadvantaged.
AI systems whose decisions cannot be understood, inspected, or challenged undermine all the other principles. If a system denies a loan, recommends a sentence, or flags content for removal, those affected must be able to understand why and seek redress. Explicability includes both intelligibility (the system can be described in terms non-experts can understand) and accountability (someone is responsible for its decisions and can be held to account). A handful of developers should not be transforming the everyday lives of billions without any mechanism for oversight.
GenAI Classifier
For each scenario, decide whether using AI in this way is Acceptable, Conditional (acceptable under certain conditions), or Not Acceptable. Click to reveal feedback.
AI Ethics Audit
Read the AI system description below, then rate it across the five AI4People dimensions using the sliders. The radar diagram updates in real time. Submit to see an analysis.
Intellectual Property
Read both tabs, then try the IP Puzzle.
Intellectual property (IP) refers to creations of the mind — inventions, literary and artistic works, designs, symbols, and names — that are recognised as property and protected by law. Intellectual property rights (IPR) give the owner legal control over use, reproduction, and commercialisation of that IP.
The main forms of IP protection relevant to researchers are:
- Copyright — arises automatically on creation of an original work: written text, software code, datasets, images, and audio-visual materials are all protected. Copyright does not protect ideas, only their expression. Duration is typically the author's life plus 70 years.
- Patents — protect novel, non-obvious inventions with industrial application. A patent requires formal application; full protection takes 3–5 years and lasts 20 years from the filing date. The complete description is published 18 months after application, regardless of whether a patent is granted. Patents cannot be obtained for laws of nature, mathematical formulae, abstract ideas, or — crucially for CS researchers — software on its own (though software embedded in a novel technical process may qualify).
- Trademarks — protect brand names, logos, and signs that distinguish one organisation's goods or services from another's. Apple's lower-case 'i' prefix is a trademark; so is the Dyson brand name.
- Design rights — protect the appearance of a product. The distinctive shape of the original iMac was design-protected, giving Apple time to build the brand before competitors could copy the aesthetic.
- Database rights — a sui generis right under UK and EU law that protects substantial investment in compiling a database, regardless of whether individual entries are creative.
Multiple layers of protection can apply simultaneously to a single research output. Dyson's cyclone vacuum cleaner was protected by patents (the mechanism), design rights (the shape), and trademark (the brand) — each providing distinct and overlapping coverage.
For most students and academic researchers, the most relevant IP questions involve copyright, open licensing, and the ownership of research outputs.
Who owns your dissertation? In most UK universities, copyright in a student's dissertation belongs to the student. However, research conducted as part of a funded project, using institutional equipment, or in collaboration with an industry partner may involve agreements that transfer or share ownership. Always check your university's IP policy and your project agreement before publishing or commercialising any output.
Open-source software licences govern how software can be used, modified, and redistributed. The key distinction is between:
- Permissive licences (MIT, BSD, Apache) — anyone can use, modify, and distribute the software, including in proprietary commercial products
- Copyleft licences (GPL, LGPL) — derivative works must be released under the same licence. Using GPL code in your project means your project must also be GPL.
Creative Commons licences cover content and data. Key conditions include: BY (attribution required), NC (no commercial use), ND (no derivatives), SA (share alike). A dataset licensed CC-BY-NC cannot be used in a project with a commercial partner, even if the research itself is non-commercial.
Open access mandates from UKRI and other major funders require that research outputs be made publicly available. This is an ethical position as well as a policy requirement: research funded by the public should be accessible to the public.
AI-generated outputs and IP is a rapidly evolving area. Under current UK law, copyright is not automatically vested in AI-generated works in the same way as human-authored works. The legal position is unsettled; researchers using AI to generate significant portions of their work should be aware of the implications for ownership, attribution, and publication rights.
IP Puzzle
A research project produces five outputs. For each, select the most relevant type of IP protection.
Which of the following cannot be protected by a patent?
A researcher wants to use a dataset licensed CC-BY-NC in a project commissioned by a commercial company. Is this permitted?
Ethical Dilemma Simulator
Work through all three scenarios. In each case, choose an action, see the consequences, and reflect on which ethical principles are at stake. You can reset a scenario and try another path. Complete all three to unlock the final knowledge check.
Final Knowledge Check
A researcher runs 20 statistical comparisons and reports only the 2 that are significant, without mentioning the other 18. Which ethical issue does this most directly represent?
Under the AI4People framework, which principle specifically addresses the need for AI decisions to be understandable and challengeable?
A student submits an ethics form for a project using interview data. After approval, the study design changes significantly — new vulnerable participants are recruited. What should the student do?
Review
Use the self-assessment below to gauge your confidence on each learning objective. Marking an objective as needing revisiting reveals a direct link back to the relevant activity.
- Ethics problems arise from ignorance, pressure, misdemeanours, and ambiguity — not only from deliberate misconduct
- Researchers owe obligations to participants, the public, institutions, funders, colleagues, and themselves
- Consequentialism, deontology, and virtue ethics each illuminate different aspects of ethical dilemmas
- FFP (fabrication, falsification, plagiarism) is the most serious category; QRPs (p-hacking, HARKing, selective reporting) distort the record without always constituting proven misconduct
- Professional pressures create conditions for misconduct; recognising them is the first step to resisting them
- The Nuremberg Code established that voluntary, informed consent is a non-negotiable requirement for human participant research
- Anonymity (no one can identify) differs from confidentiality (researcher knows but will not disclose)
- The ethics approval spectrum runs from self-certification → supervisor review → ethics committee, depending on risk
- UK GDPR applies to any identifiable personal data; special category data (health, ethnicity, etc.) requires higher justification
- Authorship requires meeting all four ICMJE criteria; gift, ghost, and honorary authorship are all ethical violations
- Generative AI is a tool, not an author — the researcher is always responsible for accuracy, integrity, and disclosure
- The AI4People framework: Beneficence, Non-maleficence, Autonomy, Justice, Explicability
- Unintended consequences include proxy variable bias and laundered historical bias; not all harmful applications are accidental
- Copyright arises automatically; patents require application and cannot cover software on its own; open-source licences vary substantially in what they permit
- Ethical reasoning requires identifying stakeholders, weighing competing obligations, and justifying decisions — not merely following rules
Proceed to Unit 6: Qualitative Methods when ready.