Research ethics, do no harm and informed consent in M&E

Why M&E is not exempt

The ethical obligations of research do not attach to the word “research”; they attach to the act of collecting data from people — usually people with less power than the collector, often people whose assistance depends on the organisation asking the questions. A baseline survey, a beneficiary feedback interview and a routine monitoring visit all involve that act. UNEG’s ethical guidelines state the position plainly for evaluation: everyone engaged in evaluation is bound by ethical principles throughout the process, whatever the exercise is called [2]. The practical consequence is that “it’s just monitoring” is never, by itself, an answer to an ethical question — the answer depends on what is collected, from whom, at what risk, and for what use.

The shared vocabulary: Belmont and UNEG

The Belmont Report (1979) remains the field’s common reference because its three principles carry almost every case [1]:

• Respect for persons — people are autonomous agents; those with diminished autonomy (children, some adults) are entitled to additional protection. Consent is this principle made operational.
• Beneficence — maximise benefit, and do not harm; risks must be justified by the value of what is learned.
• Justice — the burdens and benefits of data collection must be fairly distributed. A programme that repeatedly surveys its poorest participants for the benefit of reports they will never see has a justice problem, not just a courtesy problem.

UNEG’s Ethical Guidelines for Evaluation (2020 revision) translate this into the evaluation profession’s own four principles — integrity, accountability, respect and beneficence — and attach a pledge of ethical conduct for evaluators, commissioners and everyone else involved [2]. The guidelines sit within UNEG’s broader norms, where ethics stands alongside independence and impartiality as a condition of credible evaluation [4] — the norms themselves are covered under UNEG norms and standards.

For health-adjacent work — studies touching health status, health services, or biological data — the CIOMS International Ethical Guidelines (2016, with WHO) are the reference standard, with specific guidance on research in low-resource settings, community engagement, and the participation of people whose vulnerability is situational rather than intrinsic [3].

Informed consent that is real

The Belmont Report analysed consent into three elements — information, comprehension, voluntariness [1] — and each fails in a characteristic way in field practice:

• Information without comprehension. A consent script read at speed, in a second language, full of institutional vocabulary, produces signatures, not consent. The working remedies: scripts in the respondent’s language at everyday reading level; a comprehension check (“what will happen to what you tell me?”) before proceeding; literacy-appropriate formats, including oral consent where written forms exclude or intimidate — with the oral consent itself documented (script version, date, witness or recording as the protocol specifies) [2][3].
• Voluntariness under power asymmetry. When the interviewer’s organisation also delivers the assistance, “no” is expensive to say. Consent scripts must state explicitly that refusal and withdrawal carry no consequence for services — and programme behaviour must make that true. Perceived conditionality is the field’s most common consent failure, and it is invisible in the paperwork [2].
• Consent for children and vulnerable adults. Children require guardian consent plus the child’s own assent, sought in age-appropriate terms; adults with diminished capacity require a legally or culturally appropriate proxy plus the person’s own agreement so far as they can give it. Both require a prior question: does the study need this group at all? Inclusion of those who cannot fully consent demands justification, not convenience [1][3].
• Consent is purpose-bound. Agreement to a monitoring interview is not agreement to a photograph, a case story in a fundraising appeal, or a new analysis linking the data to another dataset. New uses need new consent — the data-protection counterpart of this rule is covered under data protection.

In group settings, consent has the additional limit that facilitators can promise their own confidentiality but not other participants’ — covered under qualitative data collection.

Do no harm in fieldwork

Do-no-harm planning is the operational half of beneficence, and it protects two populations: respondents and the field team itself [2][5].

• Distress and disclosure protocols. Interviews about loss, violence or hardship can cause distress, and can surface disclosures — of abuse, of danger — that the interviewer must be prepared to receive. The protocol specifies, before fieldwork: how to pause or end an interview compassionately, what the interviewer says on disclosure, and where the referral card points.
• Referral readiness. A referral list — health, psychosocial, protection, legal services actually reachable from the study sites — is prepared and verified before collection begins. Asking about violence without a live referral pathway is itself a harm [5].
• Sensitive-topic safety. For gendered and stigmatised topics: interviewer sex-matching where it matters, absolute privacy for the interview, and question technique that avoids re-traumatising — instrument-level practice covered under questionnaire design. Guidance developed for gender-responsive evaluation treats respondent and interviewer safety as one planning problem [5].
• Interviewer safety and welfare. Travel and communication protocols, working in pairs where risk warrants, and space to debrief — vicarious distress in teams collecting hard testimony is predictable and manageable.
• Compensation ethics. Reimbursing costs and time is respectful; payment large enough to induce participation that would otherwise be refused undermines voluntariness. The defensible line is set locally and stated in the protocol [3].

When formal ethical review is required

Three distinct machineries can apply to a study, and the threshold question — which of them does — is settled before fieldwork, not after [2][3]:

• Institutional ethics review (ERB/IRB). Studies designed to produce generalisable knowledge — and evaluations whose risk profile approaches that of research (vulnerable participants, sensitive data, experimental designs) — go to an ethics review board. Many international NGOs and UN agencies operate or contract standing boards for exactly this purpose; the ethics of randomised designs specifically are discussed under randomised controlled trials.
• National research clearance. Most countries license research conducted on their territory. In Kenya, the National Commission for Science, Technology and Innovation (NACOSTI) is the national research-licensing body, and studies that constitute research require its licence; the data-law obligations that run alongside are covered under data protection. This page describes the obligation’s existence; determining whether a specific exercise requires a licence is a judgement for the commissioning organisation to make with the regulator’s own guidance.
• The grey zone. Between clearly-research and clearly-routine sits most of M&E: baselines, outcome surveys, evaluation fieldwork. The honest resolution is a documented threshold judgement based on risk and intended use — will results be generalised or published? are participants vulnerable? is the data sensitive? — rather than on the activity’s label. Where the judgement is genuinely close, seeking review is the safer error [2][3]. What no branch of the flowchart permits is the inference that escaping the “research” label escapes the principles.

Ethics in the evaluation contract

Ethical practice is contractible, and UNEG’s instruments show what that looks like: evaluators and commissioners commit to the ethical guidelines and pledge; terms of reference state the consent, confidentiality and data-handling standards the evaluation will follow; and evaluators disclose conflicts of interest before appointment [2][4]. Independence has an ethical face too — an evaluator financially or personally entangled with the programme cannot deal fairly with participants whose testimony might cut against it [4]. Finally, ethical reporting closes the circle: findings presented so that no participant can be identified, dissent and negative findings reported rather than smoothed, and results returned to the communities that supplied the data — the accountability obligation covered under accountability to affected populations.

Checklist for an ethical data-collection protocol

• The protocol names the principles it applies (UNEG; Belmont; CIOMS where health-adjacent) and the specific procedures that implement them.
• Consent scripts exist in every fielded language, at everyday reading level, with a comprehension check and a no-consequences clause; oral-consent documentation is specified.
• Child assent and guardian-consent procedures exist wherever minors are in scope, with justification for their inclusion.
• Distress, disclosure and referral protocols are written and the referral list verified before fieldwork.
• The research-or-routine threshold judgement is documented, and any required ethics approval and national licence are obtained before collection.
• Interviewer training covers the ethics protocol as thoroughly as the instrument.
• Reporting plans protect identity, and results return to participants’ communities.