Course engagement screening for personal tutors
A weekly screen of each student's activity in the learning management system and other campus systems against their course cohort, which flags students whose engagement has fallen far below their peers to their personal tutor or a support team, using no demographic inputs.
Discovery, not a referral queueTechnique: Anomaly Detection
This solution compares each student’s weekly activity in the learning management system, and where used their attendance, library, and building access, with the rest of their course in the same week, and flags the students who have fallen far below it.
The flag goes to the student’s personal tutor, or to a support team where tutors lack the time, as a prompt to get in touch, not as a judgement.
Nottingham Trent University has run a dashboard of this kind since a 2013 pilot, banding each student’s engagement from high to very low, and its vendor’s case study reports that students with lower average engagement progressed from the first year less often.
The university chose to measure what students do and not who they are, leaving socio-demographic data out of its model entirely, and in 2020 it set up a call centre to contact students with low and very low engagement when teaching moved online.
That campaign found the students it called came disproportionately from disadvantaged backgrounds, and that tutors sent alerts did not always have time to act on them, so the flags need a team with time behind them and their reach measured by group.
The screen finds students who have drifted away, which is a different group from those who are attending but failing, so it complements the degree map and grade alerts rather than replacing them.
It is the wrong choice for programmes that teach mostly offline, where platform activity says little, and for any institution that would use the flags for discipline rather than support.
Strengths for this problem
- Sees disengagement within weeks, long before grades or next term's registration show it, which is when a tutor's contact can still change the term.
- Compares a student with their own course cohort, so a course with little online work does not make every student in it look disengaged.
- Measures what students do rather than who they are, leaving socio-demographic data out entirely, as Nottingham Trent University chose for its dashboard.
- Can be shown to students as well as staff, so a student can see their own engagement against their cohort and act on it.
Limitations for this problem
- Online activity is a weak proxy for learning, and students who study from downloaded material, in groups, or on paper look disengaged when they are not.
- Activity is uneven across the term and across modules, so thresholds need to follow the teaching calendar rather than a fixed weekly number.
- Finds that a student has disengaged but not why, so the tutor still has to find out whether it is work, money, health, or the course.
- Depends on every course using the learning management system in a comparable way, which rarely holds across faculties.
- Its flags still fall disproportionately on students from disadvantaged backgrounds, as Nottingham Trent found when it called students with low engagement, so leaving demographics out of the inputs does not make the outcomes even.
- Personal tutors do not always have time to act on or log the flags, which is why Nottingham Trent added a dedicated team to make calls.
Data needed
- Learning management system logins, content views, quiz attempts, and assignment submissions per student, course, and week.
- Where the institution uses them, attendance records, library loans, and building access, with a clear statement to students that they are used.
- Course cohort averages and the teaching calendar, so activity is compared with the same course in the same week.
- The personal tutor or support team assigned to each student, and a record of each contact made from a flag and its outcome.
How it is measured
- Share of flagged students contacted by their tutor within a week, and share who re-engaged in the following weeks.
- Continuation into the next year for students who were flagged and contacted against those flagged and not reached.
- Share of flags the tutor judged to be false, such as students working offline, by course.
Pitfalls
- Treating low activity as low commitment and acting on it as a disciplinary matter, rather than as a reason to ask the student how they are.
- Using one fixed threshold across courses whose teaching uses the platform very differently.
- Collecting building access or location data without telling students, which damages the trust the system depends on.
- Adding demographic data to the screen to make it more predictive, which turns a measure of what students do into a judgement on who they are.
Human oversight
The personal tutor, or a support team where tutors lack the time, decides whether a flag needs a conversation, what support to suggest, and whether to refer the student to wellbeing, financial, or academic services. No academic or disciplinary decision is taken on engagement data. The student can see their own engagement data, and the learning analytics lead reviews each year which signals and thresholds produce flags that tutors find useful.