EDUC90929 Chap.12 Datafication and AI in Education Policy
Datafication and AI in Education Policy
Datafication turns more educational activity, people and institutions into data that can be compared, predicted and governed. The Week 7 materials use five cases—student performance data, the Ofqual algorithm for score estimation, PISA, teacher value-added models (VAMs) including the Rubinstein case, and student experience—to show that every metric selects a purpose, definition of success and distribution of error. A proxy can provide useful information and remain too narrow or unstable for the decision attached to it.
O'Neil's audit prompts questions about data, integrity, success, accuracy and long-term effects. The subject also connects datafication to surveillance capitalism and builds generative-AI literacy around hallucination, bias, privacy, intellectual property, verification and declaration. Week 7 designates self-determination theory and constructivism, but the local page contains headings only; use the current Canvas theory resources before defining or applying their concepts.
The three required readings are Kohler (2024), McLean and Wheaton (2024), and the Ellucian and Nous (2025) report on scaling AI for student success. Required viewing is the September 2024 NBC News case about a Texas private school using AI, with the activity also linking Alpha School. Treat that short report and provider site as prompts about purpose, pedagogy, data, teacher roles, agency and evidence—not independent proof of effectiveness.
UNESCO's 2023 full guidance supplies governance depth: regulatory lag, human-centred agency, privacy and age safeguards, institutional validation, co-design, inclusion, equity, language, foundational and higher-order thinking, and assessment redesign. The Week 7 report page also names a TEQSA (2026) report, but the local file is a wrapper; detailed TEQSA recommendations are not reproduced here and must be checked in current Canvas or the report itself.
What this chapter covers
- 01Datafication, metrics, models and policy decisions
- 02Five education data-ethics cases
- 03O'Neil's algorithmic-audit questions and feedback loops
- 04Surveillance capitalism, ownership, consent and agency
- 05Generative-AI hallucination, bias, privacy, verification and declaration
Auditing an educational risk score
- 2Inspect the source, collection context, completeness and integrity of each data field.
- 2Interrogate how risk and success are defined and whose values the definition embeds.
- 2Disaggregate false positives and false negatives and identify who bears their costs.
- 2Trace long-term feedback effects on opportunity, behaviour and future data.
- 2Specify human review, contestability, privacy and evidence of actual educational benefit.
Key terms
- Datafication
- The translation of educational life and institutions into data that can be measured, compared, predicted or governed.
- Algorithmic audit
- A structured examination of data, integrity, definitions of success, errors and long-term effects in a modelled decision process.
- Feedback loop
- A process in which a model's output changes behaviour or opportunity and thereby shapes the future data used by the system.
- Surveillance capitalism
- A system in which human experience is turned into behavioural data, prediction products and interventions designed to shape behaviour.
- AI hallucination
- False or misleading model output presented with apparent confidence or plausibility.
- Human oversight
- Meaningful human responsibility for verifying, judging, contesting and acting on an automated system's output.
Datafication and AI in Education Policy FAQ
Is a highly accurate model automatically ethical?
No. The definition of success, distribution and cost of errors, data rights, feedback effects and available human review also matter.
Why can generative AI sound certain when it is wrong?
The subject explains that a language model predicts probable continuations from learned patterns rather than guaranteeing retrieval of verified truth.
May students upload subject materials to a public AI tool?
Do not upload copyrighted, confidential or sensitive material without permission. Follow current University and subject rules, protect personal information and declare or cite use as required.
What source limitation affects this chapter?
The subject materials available here contain no Week 7 lecture deck. The chapter relies on the data-ethics page, Week 7 readings, GenAI literacy materials and UNESCO's full guidance; the two designated theory entries and TEQSA report are treated as source-thin signposts.
What happened in the Ofqual case?
The UK regulator Ofqual developed an algorithm using four factors, including historic school grade distributions from 2017–2019. The activity states that students from historically lower-performing schools were disproportionately affected and that the share of private-school students receiving the highest marks increased. The government ultimately based results on teacher judgements. The case shows how historic institutional performance can be transferred onto current individuals through a model.
How should the Alpha School case be analysed?
Separate what the required NBC News video shows, what the school claims and what independent evidence establishes. Ask which learning purpose dominates, which teaching work is delegated, which data are produced, what choices teachers and learners retain, and how learning, wellbeing and equity are measured. A five-minute news item and provider site can frame the inquiry, but they cannot alone establish effectiveness, safety or scalability.
Assessment move
For each case, make an audit table with data source, missing data, success definition, group-specific errors, affected rights, feedback loops, remedy and human control. Keep the five cases separate: Ofqual concerns score estimation and historic school distributions; VAM concerns attribution of nonlinear, multi-teacher learning; Rubinstein provides a named VAM illustration; PISA concerns system comparison; student experience is a valuable but partial quality measure. Read Kohler, McLean and Wheaton, and Ellucian and Nous before using their arguments. Use UNESCO to test validation, privacy, age, co-design, agency, inclusion and assessment. Obtain the current Canvas sources before applying self-determination theory or constructivism and before attributing recommendations to TEQSA (2026). For AI-assisted study, attempt the task first, verify every factual claim in the original source, protect sensitive and copyrighted material, preserve prompts and outputs, and record use for an accurate declaration.
Working through Datafication and AI in Education Policy in EDUC90929? Sia is AskSia’s AI Education tutor — ask any EDUC90929 Datafication and AI in Education Policy question and get a clear, step-by-step explanation grounded in how EDUC90929 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
