Professional Ethics
Domain 8: Professional Ethics
Professional Ethics tests adherence to the INFORMS Code of Ethics and related professional standards in analytics. Questions probe integrity, objectivity, confidentiality, conflicts of interest, data stewardship, fairness, transparency, and accountability.
Core rules and concepts
- Act with honesty and integrity: report methods, results, and limitations accurately; do not fabricate, falsify, or selectively report.
- Maintain objectivity: disclose conflicts of interest, separate advocacy from analysis, and avoid pressure to produce predetermined findings.
- Protect confidentiality and privacy: collect and use only necessary data, secure it, respect consent and legal limits, and avoid re-identification.
- Promote fairness and avoid harm: assess disparate impact, bias in data and models, and unintended consequences; mitigate and document.
- Be transparent and accountable: explain assumptions, uncertainty, and model behavior to affected stakeholders; identify an accountable owner.
- Follow applicable laws, regulations, and organizational policies; escalate concerns when asked to act unethically.
- Represent qualifications and work truthfully; give credit; do not mislead with visualizations or cherry-picked metrics.
Worked example
A hiring model improves selection speed but shows a lower pass rate for one protected group. The analyst must not ignore the result or hide it. They validate the disparity, check whether it stems from biased labels, features, or historical decisions, and quantify impact. They document the finding, consult legal and HR, and propose mitigation such as removing proxy features, reweighting, or changing decision rules. They disclose the limitation to the business owner and recommend a pilot with monitoring before full deployment. If leadership pressures them to suppress the finding, they escalate through the ethics or compliance channel. The ethical issue is not only the model's accuracy but the decision to deploy it with known disparate impact and no mitigation.
Traps that cost marks
- Cherry-picking favorable results or hiding known model limitations.
- Using sensitive data without a lawful basis, consent, or minimization.
- Ignoring disparate impact because the overall accuracy is high.
- Failing to disclose conflicts of interest or outside pressures.
- Presenting misleading visualizations or overstated certainty.
- Treating ethics as a one-time review rather than ongoing governance.