8.5 Ethical Considerations in Learning Analytics

Ethical considerations in learning analytics are not a supplement to technical practice — they are constitutive of it. A learning analytics system that violates learner privacy, produces biased predictions, or deploys AI without accountability mechanisms is not merely ethically questionable; it is professionally and legally deficient, and it undermines the trust that educational institutions require to function. Section 8.5 addresses the three primary ethical domains: data privacy and security, fairness and bias, and responsible AI deployment.

Why Ethics Cannot Be Deferred

The temptation to treat ethics as a post-hoc review — build the system, then check for problems — reflects a misunderstanding of how ethical failures occur in learning analytics systems. The most consequential ethical problems are typically built into systems at the design stage: in the choice of which data to collect, which outcomes to predict, which features to include in models, and which populations will be the primary users.

The U.S. Department of Education's AI in education policy report documented this pattern concretely: facial recognition systems used in e-proctoring that were demonstrated to have significantly higher error rates for darker-skinned, transgender, and neurodivergent students were deployed at scale before their differential accuracy was measured.[LE-LS-GL-004] The ethical failure was not in the post-deployment discovery of bias — it was in the absence of pre-deployment equity audit as a standard deployment requirement.

The LAK Conference has made ethical deployment a central theme since its founding, publishing work on algorithmic fairness, consent frameworks, data governance, and the power dynamics of educational data collection.[LE-LS-CE-002] The Journal of Learning Analytics has published the field's primary theoretical frameworks for ethical LA, including influential work on the concept of "learning analytics as power" — the recognition that data systems in educational institutions are not neutral tools but instruments of institutional authority that affect learners unequally.[LE-LS-JO-004]

The Regulatory and Policy Landscape

Learning analytics practice in the United States operates under a legal framework that includes FERPA (Family Educational Rights and Privacy Act), COPPA (Children's Online Privacy Protection Act) for minors, and a growing patchwork of state-level student data protection laws. The U.S. Department of Education's policy report represents the first federal attempt to extend this framework explicitly to AI-enabled educational systems.[LE-LS-GL-004]

The High-Leverage Opportunities report identified "algorithmic equity" — ensuring that LA systems do not produce disparate impacts on historically underserved groups — as a strategic priority, noting that the field lacks both the methods and the governance structures to systematically audit deployed systems for equity.[LE-LS-GL-003]

Internationally, the GDPR (General Data Protection Regulation) in the EU and equivalent frameworks in other jurisdictions impose data minimization requirements and rights of explanation for automated decisions that directly affect learning analytics deployment in those contexts.

Organizational Maturity and Ethical Governance

The Generalizable Learning Engineering Adoption Maturity Model frames ethical governance as an organizational capability dimension: institutions at low maturity have no formal processes for ethical review of LA systems, while institutions at high maturity have integrated ethics review into all stages of the LE lifecycle — from data instrumentation design through model development, deployment, and ongoing monitoring.[LE-LS-AP-013]

The Society for Learning Analytics Research has published ethical guidelines for LA practice, and the Penn Center for Learning Analytics has produced policy frameworks that help institutions translate those guidelines into institutional governance procedures.[LE-LS-CO-003][LE-LS-CO-004]

The Learning Engineering Toolkit treats ethical constraints as first-order design requirements — not filters applied after the technical system is built — and includes specific guidance on consent, data minimization, and equity audit as components of the standard LE design process.[LE-LS-GL-007]

The subsections that follow treat each ethical domain in depth: 8.5.1 covers data privacy and security; 8.5.2 covers fairness and bias in analytics; 8.5.3 covers responsible AI in learning analytics.

¶ Subsections:

From the Learning Engineering Toolkit

The Toolkit places data ethics in the investigation stage, where analysis happens and where those doing the work must be careful not to let bias creep into how they read the data, to safeguard learner privacy, and to interpret findings with ethical judgment in mind. [LET-07] When a solution is put into practice, the team likewise has to consider information ethics — privacy, consent, and user autonomy above all — and to remember that what counts as private is shaped not just by law but by cultural norms. [LET-07] The book's position is that learner data ought to serve only the purpose for which consent was given, stay confidential and secure, with learners told how their data will be used and given a way to have it deleted. [LET-07]

The book treats ethical judgment as a skill in its own right: exercising it means being able to recognize the ethically salient aspects of a situation — its ethical affordances — because even uses no one intended can raise fresh ethical problems. [LET-07] It draws on the American Psychological Association's five principles — beneficence and nonmaleficence, fidelity and responsibility, integrity, justice, and respect for people's rights and dignity — and lays them against every phase of the learning engineering process in what it calls the Sense-making Ethical Evaluation Matrix (SEEM). [LET-07]

A later chapter turns this into SEEM-ED, an instrument for evaluating a design by posing targeted questions along those same five dimensions at each phase of the work. [LET-17] Importantly, SEEM-ED is meant to be more than a checkbox exercise: responses fall on a scale — at least a three-way Yes/Maybe/No rating — different stakeholders answer on their own and the scores are averaged, and the whole exercise can be run anonymously when peer pressure might otherwise distort it. [LET-17] One of its prompts asks directly whether the product or process safeguards users' rights to privacy and confidentiality. [LET-17]

These warnings echo a point made earlier in the book: numbers alone rarely capture the full picture, and missing the surrounding context can push a team toward the wrong conclusion. [LET-06] Since one set of results can be read in more than one way, the book casts interpretation as something a team does together, so that a narrow reading does not cause them to overlook an important — and sometimes ethically consequential — change. [LET-06]

Sources from the Learning Engineering Toolkit

[LET-07]Jordan Richard Schoenherr (2023). Chapter 7: Learning Engineering is Ethical. In Jim Goodell & Janet Kolodner, Learning Engineering Toolkit (pp. 201–228). Routledge / Taylor & Francis. doi:10.4324/9781003276579
[LET-17]Jordan Richard Schoenherr & Jodi Lis (2023). Chapter 17: Ethical Decision-Making Tools. In Jim Goodell & Janet Kolodner, Learning Engineering Toolkit (pp. 361–363). Routledge / Taylor & Francis. doi:10.4324/9781003276579
[LET-06]Michelle Barrett, Erin Czerwinski, Jim Goodell, Daniel Jacobs, Steve Ritter, Robert Sottilare & Khanh-Phuong Thai (2023). Chapter 6: Learning Engineering Uses Data (Part 2): Analytics. In Jim Goodell & Janet Kolodner, Learning Engineering Toolkit (pp. 175–199). Routledge / Taylor & Francis. doi:10.4324/9781003276579

Community Recommended Reading

Data-Informed Course Improvement: The Application of Learning Engineering in the Classroom — Anne Fensie (2023, ICLS 2023)

ID: lecommons-rl-data-informed-course-improvement-the-application-of-learning-engineeri · type: reading list item · confidence: medium · expert-validated: false

Chapter 3: LE Toolkit — Introduction — Jim Goodell (2022)

Source: goodell-ch3-introduction · confidence: medium · expert-validated: false

Context enrichment applied 2026-04-17. See wrgr/lecommons for source data.