Learning Analytics: Using Student Data to Support Learning Responsibly

Start with a support question

Learning analytics uses information about learners and learning activities to inform educational decisions. Its value depends on what people do with the information. Before selecting a dashboard or model, define the support question: which learners may need help with a concept, a transition or an access problem, and what assistance can the institution realistically provide?

Jisc's code of practice sets out institutional responsibilities for responsible learning analytics.^[1] A local implementation should translate those responsibilities into a clear purpose, accountable owner and practical response. Collecting more data without a support pathway can create surveillance without educational benefit.

Understand what an indicator represents

A low number of platform visits may indicate difficulty, but it may also reflect offline study, downloaded resources or a learner's different working pattern. A missed submission can have academic, personal or technical explanations. Describe each indicator's source, timing, missing data and possible interpretations.

Avoid using a proxy as though it directly measured motivation or ability. Check the data pipeline for duplicates, incorrect enrolments and delayed updates. Where a model produces a risk score, examine its validation, false alerts and performance across relevant groups. Staff need enough information to question the score rather than defer to it automatically.

Design a proportionate response

Decide who receives an alert and what they should do next. A supportive response might begin with a neutral invitation to discuss learning needs. It should not label the learner as failing or make an adverse decision without a fair process. Train staff to interpret uncertainty and document the outcome of the contact.

Match the intervention to available services. If the institution identifies a possible access barrier, it should have a route to equipment, connectivity advice or an alternative learning arrangement. An early-warning system that produces more alerts than staff can address may create unmet expectations and distract from urgent cases.

Explain data use and protect it

Tell learners what information is used, for which purpose and by whom. Set appropriate access controls and retention arrangements, and obtain the applicable legal and institutional review. Requirements for consent or another lawful basis depend on the jurisdiction and the type of processing; do not copy another institution's statement without checking its suitability.

The University of Edinburgh describes its own learning analytics approach and governance context.^[2] Such institutional examples can inform questions, but local policy must reflect the actual data flows and support model. Provide a route for students to ask questions and correct relevant inaccuracies.

Illustrative example: a missed-assignment alert

Imagine a college trialling alerts when a learner misses two early assignments in a course. An adviser reviews the record before contacting the student. One alert results from a submission recorded under the wrong group; another reveals that a learner cannot access the required software. These cases require different responses despite having the same initial indicator.

The pilot records whether the contact was timely, whether the issue was understood and whether support was available. It does not assume that every alerted student would otherwise withdraw. The institution uses the findings to improve both the data and the response process.

Evaluate educational value and unintended effects

Measure whether the intervention reaches learners who benefit, how often alerts are inaccurate and how students experience the contact. Review staff workload and whether the process overlooks learners whose difficulties are not visible in platform data. Where comparisons are made, account for differences between groups and avoid causal claims unsupported by the design.

Check for unintended incentives. Students may change online behaviour to satisfy an indicator without learning more, while staff may prioritise easily measurable engagement over deeper educational needs. Revise indicators when they distort practice or create unnecessary anxiety.

Keep human accountability visible

Assign responsibility for data quality, model changes, interventions and periodic review. Record who can approve a new use and whether the original notice still covers it. A dashboard developed for academic support should not quietly become a disciplinary or marketing system.

Learning analytics is a decision aid. Its strongest use combines limited, relevant information with professional judgement and a respectful conversation. Continue a system only where the institution can explain its educational purpose, manage its risks and show that the resulting actions are useful to learners.