Educational Item Banking

Educational item banking is the organised storage of examination questions, together with information about what they assess and how they perform. A calibrated item bank places questions on a common measurement scale using evidence from candidates’ responses, allowing different papers to be assembled to comparable assessment specifications.

How questions are calibrated

Calibration turns a collection of questions into a measurement resource. Expert review establishes content relevance, while statistical analysis examines how questions function.

• Each question is tagged by syllabus area, learning objective, cognitive demand and format, enabling selection against a test blueprint.
• Questions are tried out on an appropriate candidate sample, and responses are analysed for difficulty, discrimination and possible defects.
• Item response theory models the probability of answering correctly in relation to candidate ability and item characteristics; the parameters estimated depend on the model used.
• Questions calibrated separately require linking, for example through common anchor items, before their parameters can be treated as belonging to a common scale.

How random generation produces comparable papers

The process is constrained random selection, not an unrestricted lottery. Software selects different combinations of questions while satisfying a shared set of content and measurement requirements.

• Every paper must meet the same blueprint requirements, such as coverage of topics, cognitive demands, question formats and total marks.
• Selection also targets a comparable difficulty profile and test information, which indicates measurement precision at different ability levels.
• Matching average difficulty alone is insufficient: papers can have the same average difficulty but measure some ability levels with different precision.
• Randomisation within these constraints allows multiple question combinations without deliberately making one paper easier or less representative of the syllabus.

Conditions for equivalence and secure use

Equivalence is an empirical claim, not a guarantee created by software. It depends on the quality of calibration, the assembly constraints and evidence that the resulting scores support comparable interpretations.

• Where necessary, score equating adjusts for differences in difficulty between forms; equal raw marks should not automatically be assumed to represent equal performance.
• Items require periodic review because exposure, curricular changes or altered response patterns can undermine earlier calibration.
• Exposure controls, restricted access and audit trails protect the bank; random generation alone cannot prevent leakage or compromise.
• A sufficiently large, balanced bank is necessary to generate varied papers without repeatedly overusing a small set of suitable questions.