A dataset is useful only when people can find it, understand it, and access it in a way that respects participants. Learners prepare a data package with an access statement, repository record, Licence, README, data dictionary or codebook, and collection history. They also practice writing a clear explanation for legitimate restrictions on sharing.
Distinguish open, controlled, and non-shareable data based on an access statement and privacy context.
Create a data package that includes a repository record, documentation, provenance, and a clear access or restriction statement.
Explain how metadata, licences, and persistent identifiers support responsible discovery and reuse.
Review a data-sharing statement and identify missing information about access, documentation, collection, processing, or permitted use.
A paper links to a dataset, but the file contains abbreviated variable names, no definitions, no indication of who was included, and no explanation of how access is granted. A future researcher cannot tell whether event_type = 4 means a quiz attempt, a page view, or an instructor action; they also cannot tell whether the file represents raw events or a processed subset.
Posting a file alone does not make data understandable or responsibly reusable. Researchers need enough context to find the record, interpret the variables, understand how the data were collected and changed, determine whether they may use it, and follow any access process. This matters especially for learner data, where detailed records can be sensitive and data management needs to address transparency, access, security, and accountability. Ethical and privacy principles for learning analytics
Open data can be accessed without an individual access decision. An open-data statement should say where the data can be found, identify the dataset record, state the licence or terms of use, and point readers to the accompanying documentation. A clear licence matters because reusable data should have accessible terms that explain permitted use. The FAIR Guiding Principles for scientific data management and stewardship
Controlled-access data are not directly downloadable by everyone. Instead, a repository record or data statement can describe the dataset and explain how an eligible researcher may request access, including the route for making a request and any conditions that apply. FAIR explicitly allows authentication and authorization where necessary, and rich public metadata plus clear access rules can support FAIRness even when sensitive data themselves are not published. The FAIR Guiding Principles for scientific data management and stewardship
Non-shareable data should not be provided to external users when responsible sharing is not possible in the study context. The data statement should still be informative: state that the data are not shared, give a brief rationale, and describe what can be available instead, such as metadata, a codebook, a processing description, aggregated outputs, or analysis code where appropriate. A justified restriction is not a failure of transparency; it tells future researchers what exists, why direct access is limited, and what can still be inspected.
For learner, behavioural, and administrative data, a clear restriction rationale helps a future researcher understand why a download is unavailable rather than mistaking missing data for poor reporting. Detailed access policies are recommended because sensitive data can have serious consequences when exposed, and an access policy should identify permitted operations and who may access which areas. Ethical and privacy principles for learning analytics
A useful data package helps a reader answer three questions: What is this? How was it produced? What may I do with it? Documentation should travel with the data or be clearly linked from the repository record.
Artifact | What it tells a future reader | Example question it answers |
|---|---|---|
README | The package contents, file formats, folder structure, and how the files relate to one another | “Which file contains the analysis-ready data?” |
Data dictionary or codebook | Variable names, definitions, value labels, units, missing-value codes, and derivation notes | “What does |
Metadata record | Searchable descriptive information about the dataset, including title, creators, topic, dates, and related materials | “Is this dataset relevant to learner activity in online courses?” |
Collection description | The setting, population, recruitment or inclusion approach, time period, instruments or systems, and event-generation process | “Which learners and learning environment produced these records?” |
Processing history (provenance) | How raw data were cleaned, linked, filtered, transformed, or aggregated into released files | “Which rows were removed before the final dataset was created?” |
Licence or terms of use | Conditions for using, adapting, sharing, and citing the data | “May I reuse these data in a new study?” |
Repository record and persistent identifier | A stable, citable landing point for the dataset, its metadata, documentation, versions, and access information | “What exact dataset record should I cite and return to later?” |
Metadata is information that describes a dataset: what it contains, who created it, when it was collected, how it can be accessed, and how it relates to other research objects. Rich metadata supports discovery, while a persistent identifier is a stable identifier for the dataset or its record; FAIR calls for globally unique and persistent identifiers, rich metadata, and registration or indexing in a searchable resource. The FAIR Guiding Principles for scientific data management and stewardship
Processing history is also called provenance: the account of where data came from and what happened to them before release. It should distinguish, where relevant, between source data and processed data, name major transformations, and link the resulting files to the methods and code that produced them. Detailed provenance is part of the FAIR reuse guidance. The FAIR Guiding Principles for scientific data management and stewardship
FAIR means Findable, Accessible, Interoperable, and Reusable. It is a set of guiding principles for making research objects easier for people and machines to discover, access, combine, interpret, and reuse. The principles apply not only to conventional datasets but also to algorithms, tools, and workflows that led to the data. The FAIR Guiding Principles for scientific data management and stewardship
In plain language:
Findable: the dataset and its description can be discovered through a searchable record and a persistent identifier.
Accessible: a reader can retrieve data or metadata through a stated route; that route may require authorization.
Interoperable: the data and metadata use understandable formats, shared terms, and meaningful links that help them work with other resources.
Reusable: the dataset has enough accurate description, provenance, and usage terms for someone else to assess and use it appropriately.
FAIR does not automatically mean open. A sensitive or personally identifiable dataset can be FAIR when its metadata are discoverable and its access process is explicit, even if the dataset itself is restricted. The FAIR Guiding Principles for scientific data management and stewardship
FAIR also does not guarantee ethical access, appropriate consent, or protection from re-identification. De-identification or anonymization can reduce direct identifiers, but it does not automatically eliminate re-identification risk: combinations of information can be used to derive identity, and the risk depends on the type of data. Ethical and privacy principles for learning analytics
Treat FAIR as a guide for responsible stewardship, not as a label that settles every privacy or ethics question. In learning analytics, transparency includes describing what information is collected and how it is collected, stored, and processed. Ethical and privacy principles for learning analytics
A research team studies patterns of engagement in a fictional online postgraduate course. Their learner-event dataset contains pseudonymous learner IDs, timestamps, resource views, discussion-post counts, assessment-submission events, and weekly engagement measures. The team has removed direct identifiers, but the combination of detailed timestamps and unusual activity patterns could still be sensitive.
The data are available on request.
This statement leaves major questions unanswered. A reader does not know what the dataset contains, whether a public record exists, whom to contact, what access criteria apply, why access is limited, what documentation is available, or whether the data are raw or processed.
The learner-event dataset for this study is available through controlled access. The public repository record includes a persistent identifier, a metadata description, a README, a data dictionary, a collection description, and a processing history for the analysis dataset.
Direct download is not provided because the event-level records contain detailed learner activity traces that may create re-identification concerns when combined with other information. Researchers seeking access may submit a request through the access route described in the repository record; approved users receive the documented processed dataset and must follow the stated conditions of use. Aggregated study outputs and the documentation are available from the public record.
The improved statement does not promise fully open data. Instead, it makes the dataset findable, explains why direct access is restricted, identifies an access route, and tells readers what documentation they can inspect before deciding whether the data are relevant.
This approach reflects the distinction between access to data and access to metadata. FAIR guidance specifies that metadata should remain accessible even when the data are no longer available, supporting discovery and assessment without exposing restricted content. The FAIR Guiding Principles for scientific data management and stewardship