About AI Recommendations

What is X-ELRA?

X-ELRA is a learning recommendation system designed to help you discover relevant study materials. It combines multiple algorithmic signals to suggest resources that may support your learning goals.

Recommendations are generated by algorithms, not human tutors. They may not always be accurate or suitable for your specific situation. You should always use your own judgement when choosing what to study.

How recommendations are generated

The system uses a combination of signals to rank learning materials:

Content similarity — How closely a resource's topics overlap with what you have studied (Jaccard similarity on topic tags).
Collaborative filtering — Patterns from other learners who followed similar paths.
Popularity trends — Which materials are frequently accessed and completed.
Your progress — Items you have already completed are filtered out, and your mastery level is considered.
Sentiment feedback — Your self-reported feelings and ratings help adapt future suggestions (if enabled).

Different study groups may use different weightings of these signals. You can view which pathway you are assigned to on the Privacy & Transparency page.

Strengths

Personalised suggestions based on your actual progress and goals.
Transparent explanations for why each item is recommended (where enabled).
Adaptive: the system responds to your feedback and activity in real time.
Fairness monitoring: the platform tracks bias metrics to ensure equitable exposure across learner groups.

Limitations

The system cannot assess the pedagogical quality of a resource — it relies on metadata and behavioural signals.
Recommendations may be biased towards popular or recently added items.
The system does not have access to your full educational background or external learning.
Collaborative filtering can sometimes create "filter bubbles" where similar learners receive similar suggestions.
Algorithmic scores are approximations and should not be interpreted as absolute measures of suitability.

Your control

Set goals: Define mastery targets for specific skills to steer recommendations.
Provide feedback: Rate and comment on items to influence future suggestions.
Withdraw consent: You can stop data collection at any time via the consent mechanism.
Request deletion: Under GDPR, you can request full deletion of your data.

Frequently Asked Questions

Does the system use generative AI (e.g. ChatGPT)?

No. X-ELRA uses traditional recommendation algorithms (content-based similarity, collaborative filtering, and popularity scoring). It does not use large language models or generative AI to produce content or recommendations.

Can the system see my personal information?

The system identifies you by a pseudonymous learner ID derived from your account. It stores your email separately for authentication purposes. It does not collect additional personal information beyond what you provide during sign-in.

Why might a recommendation not be relevant to me?

The system works with limited signals (your clicks, completions, and feedback). It cannot assess your prior knowledge, learning style, or context outside the platform. If a recommendation seems off, rate it or skip it — this helps the system learn your preferences.

What happens if I withdraw consent?

Data collection stops immediately. Recent sentiment data is purged. You can still access the platform, but recommendations may be less personalised. You can re-consent at any time.

Who monitors the system for fairness?

The research team monitors fairness metrics including exposure distribution, CTR parity, and completion rate parity across learner cohorts. A bias audit dashboard is used to flag disparities exceeding defined thresholds.

How can I contact the team?

Email Simon.West@port.ac.uk with your learner ID. You can ask questions, report issues, or exercise your data rights.

What is X-ELRA?

How recommendations are generated

Strengths

Limitations

Your control

Frequently Asked Questions

Further information