Why a dictionary?
Words such as “support”, “reasoning” and “learning” are familiar, but they do not automatically describe the same process in a person and in a system. This Dictionary records the specific meanings used in the EAI work. It can support lesson review, tool design and careful interpretation of evidence. It does not replace professional judgement or establish the effectiveness of an intervention.
What does it help with?
Shared meaning. Teachers, researchers and developers can describe the same action when they talk about “replacement” or “independence”.
Better design questions. For a particular stage, you can ask which learning action the learner must perform and what AI actually contributes.
Bounded conclusions. A polished product, observed activity and independent performance are treated as different kinds of information.
A classroom example
A learner is asked to compare historical sources for reliability. AI summarises both sources and provides a judgement about their reliability. The learner submits a polished comparison.
Process → Stage → Core Learning Action → AI
Process: learn to identify, weigh and justify arguments about source reliability.
Stage: the learner is practising how to weigh those arguments independently.
Core Learning Action: the learner must judge which source is useful and why.
AI contribution: summaries may support access to the material, but a ready-made judgement can replace the very action being practised.
The comparison provides information about the output. Discussion during the task can provide process evidence. A new source without the same AI help may provide information about independent performance. None of those, on its own, proves everything.
Find a concept
30 concepts
What must the learner still do?5
Where does the work move?7
Why can an answer sound more certain than it is?6
What can we actually infer about learning?11
How does this affect the design?1
Act of LearningDe handeling van het lerenEducation-native
The human process in which learning actually takes place: searching, choosing, attempting, doubting, recognising error, recovering, judging and acting again. A final product may result from this process, but is not the learning itself.
AI DriftAI-driftShared
A gradual shift in which the focus, intent or distribution of action within a task changes while its outward form still appears correct. Cognitive, decisional or regulatory actions may progressively move to AI without any single step clearly appearing to be a takeover.
Answer Machine EraTijdperk van de antwoordmachineTechnology-led
An educational context in which answers are available almost instantly, fluently and persuasively, making the preceding process of search, friction, judgement and doubt less visible or less likely to occur.
Closure BiasAfsluitingsneiging (closure bias)Technology-led
The tendency to experience a fluent and well-formed answer as complete. Its form suggests that the thinking is finished, making missing questions, assumptions and reasoning steps less visible.
Distinction: Closure Bias concerns the closing effect of a finished answer. Illusion of Completeness concerns the impression that the content itself is complete.
Cognitive DisplacementCognitieve verplaatsingShared
The shifting of relevant cognitive or learning actions from the learner to the AI system, reducing the learner’s opportunity to practise and develop those actions independently.
Distinction: Cognitive offloading can be useful. Cognitive displacement becomes a concern when the very work that matters for learning disappears or moves elsewhere.
Controlled DetachmentGecontroleerde ontkoppelingEducation-led
The deliberate and temporary removal of relevant AI support to determine whether the learner can now perform the core learning action independently.
Core Learning ActionKernhandelingEducation-native
The action to which the learner must personally give substantive meaning at a particular stage in order for learning to occur.
Distinction: The core learning action is not necessarily the hardest, longest or most visible step.
Didactic ControllabilityDidactische controleerbaarheidEducation-led
The degree to which it remains visible and verifiable who made the decisions and carried out the cognitive actions that matter for learning, and whether the learner can reconstruct, explain and justify those actions.
Didactic DistortionDidactische vervormingEducation-led
A shift in which the visible form of a learning activity largely remains intact while its educational meaning changes because a relevant action, decision or judgement is no longer performed by the learner.
Education in the Loop (EITL)Education in the Loop (EITL)Education-led
The principle that educational intent remains an active part of an AI system’s decisions and boundaries. Learning goals, core learning actions, evidence requirements and human authority are built into the design rather than checked only afterwards.
Evidence-Function ShiftVerschuiving van de bewijsfunctieShared
The shift by which, under generative AI, a final product becomes a less reliable indicator of the thinking and learning process that preceded it.
Illusion of CompletenessIllusie van volledigheidTechnology-led
The impression that an answer is substantively complete because it is presented coherently and conclusively, even though relevant perspectives, uncertainties, conditions or questions may be missing.
Independent EvidenceBewijs van zelfstandigheidEducation-led
Evidence showing what a learner can independently perform, explain, judge or reconstruct without the relevant AI support.
Justification MappingVerantwoordingsmapping (keuzes verantwoorden)Education-led
A structured method for making visible what the learner asked AI, what AI proposed, what the learner accepted, modified or rejected, and why.
Learning EvidenceBewijs van lerenEducation-native
Observable information that supports a defensible claim about what the learner has learned. Its evidential value depends on its relation to the core learning action; product quality alone is insufficient.
Evidence family: Output Evidence · Process Evidence · Independent Evidence
Meaning-Making vs AdditionBetekenis geven versus toevoegenEducation-native
The distinction between visibly adding something to a product and personally giving substantive meaning to the action that matters for learning. Editing, choosing or rewriting is not automatically evidence that the learner performed the underlying judgement.
Output EvidenceOutputbewijsEducation-led
Information derived from the final product or performance. It shows what was produced, but not automatically who performed the relevant cognitive work or what the learner learned.
OversupportOverondersteuningEducation-led
A condition in which support becomes so extensive, early or persistent that it begins to take over relevant cognitive, decisional or regulatory actions from the learner.
Presence vs PerformanceCognitieve aanwezigheid versus prestatieEducation-native
The distinction between demonstrable human cognitive activity during the process and the quality of the resulting performance. Strong performance can exist even when the relevant cognitive action no longer lies with the learner.
Process EvidenceProcesbewijsEducation-native
Observable information from the learning process showing how the learner chose, reasoned, doubted, verified, recognised errors, recovered or revised a judgement.
Pseudo-ComplexitySchijncomplexiteitTechnology-led
The appearance of depth or complexity created by rich, mature or technically convincing output without evidence of corresponding understanding or cognitive work by the learner.
Distinction: Pseudo-complexity concerns the appearance of depth. Illusion of Completeness concerns the appearance of completeness.
Question EmergenceHet ontstaan van de vraagEducation-native
The process in which a learner personally notices a problem, tension, missing relation or uncertainty from which a question emerges. A good question supplied by AI is therefore not always educationally equivalent to a question that emerges from the learner.
Reflective Re-EntryReflectieve herintredingEducation-led
The deliberate re-entry into the thinking process after AI use by reconstructing what the system proposed, what the learner accepted or rejected, why, and what the learner would have done differently.
Reverse Scaffolding — EAI meaningReverse scaffolding (EAI-betekenis)Education-led
A pattern in which AI support does not fade as the learner progresses, but persists, expands or increasingly takes over the core learning action. Performance may improve while the learner receives fewer opportunities to develop the underlying capability independently.
Source note: The term reverse scaffolding was used in the educational literature before the EAI work. There it means that technology takes over only actions a learner has already constructed or mastered. EAI uses the same term differently: support keeps growing where it should fade, and so takes over more and more of the work that matters for learning. This dictionary does not claim first use of the term; it records the EAI meaning explicitly.
Support–Replacement BoundaryGrens tussen ondersteunen en overnemenEducation-led
The point at which AI no longer merely supports a task but performs the core learning action to which the learner should personally give substantive meaning.
Sycophantic CompletionMeegaande voltooiingTechnology-led
The phenomenon in which a generative system smoothly affirms and completes the user’s assumptions, preferences or line of thought, reducing necessary disagreement, doubt or critical resistance.
Source note: Sycophancy and the phrase sycophantic completion exist outside EAI. This dictionary records their specific educational meaning.
Task DensityTaakdichtheidEducation-led
The degree to which AI performs the cognitive core learning action that the learner is meant to learn, practise or demonstrate in the task.
Key question: Who carries out the action that carries the learning?
Distinction: Heavy use of AI can be harmless for learning when AI does not take over a core learning action. A small AI contribution can be decisive when it carries out exactly that action.
Trace-back PromptingTerugtraceerpromptingEducation-led
An intervention in which the learner traces backwards from a final product, choice or revision to earlier AI contributions, prompts, personal decisions and reasons.
Truth ApathyWaarheidsapathieTechnology-led
A decline in the willingness or tendency to keep actively investigating what is true, uncertain or insufficiently supported because answers are available quickly, abundantly and persuasively.
Distinction: With Closure Bias, a single answer closes too early. With Truth Apathy, searching itself becomes less natural.
Visible Cognitive WorkZichtbaar denkwerkEducation-native
Learning-relevant cognitive actions that are made observable, such as analysing, choosing, explaining, verifying, judging, correcting and justifying.
Editorial review version based on the supplied 30-term EAI Dictionary. These terms are not a validated measurement scale.