The primary ethical framework I apply to Amazon’s experimental recruiting tool is justice and fairness. This framework asks whether people receive equitable treatment and whether the benefits and burdens of decisions are distributed justly.
In hiring, applicants should be evaluated using relevant qualifications and job-related criteria rather than being disadvantaged by characteristics unrelated to their ability to perform the role. Amazon’s case raises a central concern: a system can apply the same algorithm (Dastin, 2018; Raghavan et al., 2020) to everyone while still producing unfair results. Equal application of a flawed process does not necessarily create equitable outcomes.
Transparency and explainability
Applicants may not know that an algorithm is screening their applications (Raghavan et al., 2020) or understand why they were rejected. When decision-making criteria are hidden, applicants may struggle to correct inaccurate information or challenge unfair treatment. Employers should be able to explain the system’s role, identify the criteria that materially influence decisions, and provide appropriate ways for people to raise concerns.
Accountability and human judgment
Organizations cannot ethically transfer responsibility to a software vendor or claim that a computer made the decision. Human oversight can help identify inaccurate data, questionable recommendations, or circumstances the model failed to consider. However, human review is only meaningful if reviewers are trained, have enough information to question the algorithm, and are empowered to disagree with it.
Applicants should not be reduced to unexplained scores that determine their opportunities without a meaningful review process. Employers should evaluate both the system’s design and its real-world effects.
My ethical judgment: AI hiring should be evaluated by demonstrated fairness, not the appearance of objectivity. Employers should use a system only when its assessments are job-relevant, its outcomes are monitored, and responsibility remains clear.
The National Institute of Standards and Technology’s AI Risk Management Framework identifies (Tabassi, 2023) fairness, accountability, transparency, explainability, and reliability as important characteristics of trustworthy AI.
Three questions for a fair hiring system
These questions operationalize justice and fairness alongside transparency and accountability (Tabassi, 2023; Raghavan et al., 2020).
ETHICAL LENS
Three tests for a fair hiring system
| Principle | Question to ask | Evidence to look for |
|---|---|---|
| Fairness / justice | Do qualified candidates face unjustified disadvantages? | Outcome comparisons and job-related validation |
| Transparency | Can the employer explain what influenced a recommendation? | Documented criteria, limitations, and explanations |
| Accountability | Who can challenge or override a harmful result? | Named decision-makers, review records, and an appeal route |
This is an ethical evaluation framework for the case, not a claim that any single metric can prove a system is fair.
Sources: NIST (2023); Raghavan et al. (2020). See the full references.