Competence to read, work with, analyze, and argue with data — itself context-dependent.
In newsroom practice, data literacy is source criticism extended to datasets: knowing who collected the numbers, what the unit of observation is, what the margin of error allows one to claim, and when a comparison over time is broken by a definition change. A data-literate journalist treats a spreadsheet like an interviewee with interests — checking provenance and denominators before publishing — and builds charts that do not overstate certainty. The competence is enforced editorially: claims that outrun the data are challenged in the edit, not corrected after publication.
In practice: Interrogate a dataset's provenance, definitions, and denominators before publication, and present numbers with only the certainty and comparisons the data actually supports.
OmniGloss seed synthesis, 2026 (machine-drafted, pending expert validation)
In model-risk practice, data literacy is the competence that makes 'effective challenge' real: validators, senior management, and business users must understand what the data behind a model can and cannot support — sample construction, exclusions, performance windows, and metric definitions — well enough to question developer claims rather than accept them. Supervisory guidance names competence as an explicit element of effective challenge, so banks operationalize literacy through role-based training, model-committee membership criteria, and documentation standards written to be interrogable by informed non-developers.
In practice: Question the sample construction, exclusions, and metric definitions behind a model's reported performance, and escalate claims the underlying data cannot support.
OmniGloss seed synthesis, 2026 (machine-drafted, pending expert validation)
In clinical practice, data literacy is the working numeracy that lets a clinician act safely on quantitative evidence: reading sensitivity, specificity, and predictive values against local base rates; distinguishing relative from absolute risk when discussing options with patients; and recognizing when a dashboard metric or risk score is being applied outside the population it was derived from. It is judged at the point of care — a literate clinician can say what a 12% readmission risk does and does not warrant for this patient — rather than by statistics coursework completed.
In practice: Interpret risk scores, screening statistics, and dashboard metrics against local base rates and patient context, and communicate absolute risks accurately in shared decision-making.
OmniGloss seed synthesis, 2026 (machine-drafted, pending expert validation)
For hospitals deploying clinical AI, data-and-AI literacy is a compliance object: the AI Act defines AI literacy as the skills, knowledge and understanding that allow providers, deployers and affected persons to make an informed deployment of AI systems and to gain awareness of AI's opportunities, risks, and the harm it can cause, and Article 4 obliges deployers to ensure a sufficient level of it among staff dealing with the operation and use of those systems. Compliance teams operationalize this as documented role-specific training, competence records for staff assigned to oversee AI outputs, and procurement checks that instructions for use are actually intelligible to the clinicians expected to follow them.
In practice: Define, deliver, and document role-specific AI literacy training for staff using clinical AI, and verify that assigned overseers can interpret system outputs, limitations, and failure modes.
Regulation (EU) 2024/1689 (EU AI Act)
In government statistical practice, data literacy runs in two directions: officials drafting policy must read official statistics competently — confidence intervals, revision policies, administrative-versus-survey sources — and statistical offices carry a corresponding duty to present figures with the metadata and impartial commentary that make competent reading possible. The European Statistics Code of Practice treats accessibility and clarity as producer obligations, so literacy is operationalized institutionally: release notes, quality reports, and user guidance form the infrastructure on which any individual official's competence depends.
In practice: Read official statistics together with their quality reports and revision policies, and draft policy advice that reflects what the figures can and cannot establish.
OmniGloss seed synthesis, 2026 (machine-drafted, pending expert validation)
In civic and community data work, data literacy is not an individual skill deficit to be trained away but a collective capacity to contest how public data regimes classify and count people: reading a benefits algorithm's inputs, demanding the categories that render a community visible or invisible, and producing counter-data when official figures omit lived harms. On this framing, programmes that only teach chart-reading depoliticize the problem; literacy is measured by whether affected groups can effectively question and change data practices, not merely comprehend their outputs.
In practice: Equip affected communities to question the categories, sources, and uses of public data about them, and to produce counter-evidence where official data misrepresents them.
OmniGloss seed synthesis, 2026 (machine-drafted, pending expert validation)
Communities disagree about what data literacy is for and where it resides. Clinical and organizational framings treat it as an individual or workforce competence: skills that let a person interpret and use data and AI outputs correctly, deliverable through training and verifiable through records. Critical civic framings hold that this deficit model misplaces the burden: literacy is a collective, political capacity to question and reshape the data regimes that classify people, and training individuals to read outputs leaves the regimes themselves unexamined.