30 terms from engagement analytics, learning measurement and AI Act compliance. Each one includes the distinction it is most often confused with, because that is usually the part that matters.
Completion rate is the proportion of learners who reach the end of a course or module out of those who started it.
Content effectiveness is a measure of how well a specific piece of material achieves its purpose — holding attention and producing learning — assessed per section rather than per course..
Drop-off rate is the proportion of learners who abandon a course before completing it, ideally expressed against the content position where abandonment occurred rather than as a single figure..
An engagement heatmap is a visual representation of attention across the length of a piece of content, colour-coding each section from high engagement to disengagement hotspot..
An engagement index is a single continuous score, usually normalised to 0–1, that summarises how much attention a learner is giving to content at a point in time.
Learning analytics is the measurement, collection, analysis and reporting of data about learners and their contexts, for the purpose of understanding and optimising learning and the environments in which it occurs..
Time on task is the duration a learner spends on a piece of content.
An at-risk learner is one whose behavioural pattern predicts non-completion or poor outcome, identified early enough for intervention to be possible..
Cyberloafing is the use of internet access for non-work or non-learning purposes during time allocated to work or study — in a learning context, switching away from the training content to unrelated browsing..
Disengagement prediction estimates the likelihood that a learner will abandon a session before it ends, using behavioural pattern changes observed in real time..
Learner persistence is the tendency to continue with a programme through difficulty or interruption.
Affective computing is the field concerned with systems that recognise, interpret, simulate or respond to human emotion.
Baseline calibration captures an individual's neutral resting state at the start of a session, then measures all subsequent readings as deviation from that personal baseline rather than against a population average..
Emotion AI is the commercial label for systems that infer emotional states from observable signals — most commonly facial expression, sometimes voice, text or physiological data.
FACS is a taxonomy developed by Paul Ekman and Wallace Friesen that decomposes facial movement into discrete Action Units, each corresponding to a specific muscle or muscle group.
Valence and arousal are the two axes of the circumplex model of affect: valence runs from unpleasant to pleasant, arousal from calm to activated.
Biometric data is personal data resulting from specific technical processing relating to physical, physiological or behavioural characteristics.
A DPIA is a structured assessment required under GDPR Article 35 where processing is likely to result in high risk to individuals.
Emotion inference is the act of drawing a conclusion about a person's internal affective state from observable signals.
An LRS is a data store that receives, holds and serves xAPI statements.
SCORM (Sharable Content Object Reference Model) is a set of standards from 2001 for packaging e-learning content so it runs in any conformant LMS and reports completion, score and time back to it..
xAPI, also called Tin Can API, is a specification for recording learning experiences as actor–verb–object statements sent to a Learning Record Store.
An air-gapped deployment runs entirely within a network that has no connection to the public internet.
Client-side inference runs a machine learning model in the user's own browser or device rather than on a server, so raw input never leaves the device.
Federated analytics computes aggregate statistics across many devices without collecting the underlying individual data centrally.
Adaptive learning adjusts content, sequence or difficulty in response to individual learner data — typically assessment performance, and increasingly engagement signals..
Cognitive load is the total demand placed on working memory during learning.
Microlearning delivers content in short, focused units, typically three to seven minutes, each addressing a single objective.
The Kirkpatrick model evaluates training across four levels: Reaction (did they like it), Learning (did knowledge change), Behaviour (did they act differently) and Results (did a business outcome move).
The Phillips ROI Methodology extends Kirkpatrick with a fifth level, converting Level 4 business results into monetary value and comparing them against programme cost to produce a return-on-investment percentage..