Open Data and Open Science in Education Research

Education research often influences teaching methods, school policies, assessment systems, and public investment. Yet the evidence behind those decisions is not always easy to inspect.

A study may report that a new teaching intervention improved student performance. Readers can see the final tables, but they may not have access to the original protocol, measurement instruments, data-cleaning decisions, analysis code, or excluded observations. Another research team may be unable to reproduce the analysis or test the same intervention in a different setting.

Open science aims to make research more transparent, accessible, and reusable. It can include preregistration, open-access publication, shared code, documented data, public research materials, replication, and clearer reporting of methods and limitations.

Education research also requires caution. Student records may contain information about children, disabilities, family backgrounds, school performance, disciplinary actions, or small communities. Openness must therefore be balanced with privacy, informed consent, legal requirements, and community trust.

The goal is not to make every file public. It is to make research as open as possible and as restricted as necessary.

What Open Science Includes

Open science is broader than free access to journal articles. It covers several parts of the research process.

Open Access

Open access allows readers to view a research article without a subscription. This may occur through an open-access journal, an institutional repository, a funder archive, or an accepted manuscript shared by the author.

Open Data

Open data refers to research data made available for verification, secondary analysis, or reuse. The level of access may range from public download to controlled access in a secure environment.

Open Materials

Research materials may include surveys, interview guides, lesson plans, observation protocols, intervention manuals, scoring rubrics, and recruitment documents.

Open Code

Analysis code records how researchers cleaned, transformed, analyzed, and visualized data. It may include scripts, notebooks, package information, and instructions for reproducing reported results.

Open Methods

Transparent methods explain how participants were recruited, how variables were measured, which observations were excluded, and how analytical decisions were made.

Open Engagement

Teachers, students, families, schools, and communities can also participate in identifying research questions, interpreting findings, evaluating privacy risks, and deciding how results should be shared.

Open science should not be confused with Open Educational Resources. OER are openly licensed teaching and learning materials. Open science addresses the broader research process.

Why Education Research Needs Transparency

Education studies can contain many analytical decisions. Researchers may choose among several outcomes, models, subgroup definitions, exclusion rules, and measurement approaches.

When those decisions are not visible, readers may struggle to understand how strongly the evidence supports the conclusion.

Greater transparency can reduce problems such as:

• Selective reporting of positive outcomes
• Unclear changes to the original analysis plan
• Missing descriptions of interventions
• Unavailable research instruments
• Poorly documented datasets
• Difficulty reproducing published results
• Repeated collection of similar data
• Limited comparison across studies

Transparency does not guarantee that a study is methodologically strong. It allows other people to evaluate the study more accurately.

Open Science Begins Before Data Collection

Research teams should not wait until publication to think about openness. Data sharing becomes more difficult when files are disorganized, consent forms do not mention reuse, and analytical decisions have not been documented.

Planning should begin before the first participant is recruited.

1. Define the data and materials that will be collected.
2. Identify sensitive and identifying information.
3. Prepare consent language that reflects the sharing plan.
4. Select file formats and naming conventions.
5. Decide which outputs may be public.
6. Identify materials requiring controlled access.
7. Plan documentation and analysis code.
8. Select an appropriate repository.
9. Budget time for data curation.
10. Define retention and destruction procedures.

This process is usually recorded in a data management plan.

Preregistration Before Results Are Known

Preregistration creates a time-stamped record of a study plan before researchers examine the main results.

A preregistration may describe:

• Research questions
• Hypotheses
• Sample and recruitment strategy
• Primary and secondary outcomes
• Exclusion criteria
• Statistical models
• Planned subgroup comparisons

Its purpose is not to prevent researchers from learning from unexpected patterns. It helps distinguish between confirmatory and exploratory analysis.

A confirmatory analysis tests a question defined before the main results were known. An exploratory analysis investigates patterns discovered during or after data analysis.

Both forms of research can be valuable. Problems arise when an exploratory result is presented as though it had been predicted in advance.

Platforms such as the Open Science Framework Registries allow researchers to create time-stamped study records.

Registered Reports

Registered Reports move part of peer review to an earlier stage.

A journal evaluates the research question, sampling plan, methods, and proposed analysis before the results are available. If the design is accepted and the researchers follow the approved plan, the journal may offer publication regardless of whether the findings are positive, negative, or inconclusive.

This format can reduce pressure to produce a surprising result. It also directs attention toward the quality of the research question and design.

Registered Reports are not appropriate for every project. Rapid-response research, emergent qualitative work, and participatory studies may require more methodological flexibility.

Open Data Does Not Mean Unrestricted Data

Different datasets require different levels of access.

A controlled dataset can still support open science when access rules, metadata, and application procedures are clear.

Why Education Data Need Special Protection

Education research frequently involves minors and small social groups. Removing names is not always enough to prevent identification.

A student may still be recognizable through a combination of:

• Age or grade level
• School or district
• Rare disability or medical condition
• Race, language, or demographic category
• Dates of participation
• Unusual academic history
• Small classroom membership

Longitudinal data can create additional risks because records from several years may be linked with public or administrative information.

Researchers must consider whether data release could expose a student, stigmatize a school, reveal private family information, or allow research data to be reused for surveillance or punishment.

Privacy rules vary by country and institution. Research teams should follow the laws, ethics requirements, and data-governance policies that apply to their project.

Consent Must Anticipate Data Sharing

Consent to participate in a study does not automatically mean consent for unrestricted public data sharing.

Consent materials should explain:

• What information will be collected
• Whether data will be de-identified
• Who may receive access
• Where the files will be stored
• How long they will be retained
• Whether secondary research use is possible
• Whether withdrawal remains possible after sharing

Consent does not remove the researcher’s duty to reduce risk. Participants may agree to sharing without understanding how data could later be linked with other sources.

Data Management Plans

A data management plan should function as an operational guide rather than a document written only for a grant application.

It can address:

• Data types and formats
• File names and folder structures
• Version control
• Access permissions
• Encryption and secure storage
• Backups
• Metadata and documentation
• De-identification procedures
• Licensing
• Repository selection
• Preservation period
• Data destruction

A useful planning question is:

What would another qualified researcher need to understand and reuse this dataset five years from now?

Making Education Data FAIR

The FAIR principles describe data that are Findable, Accessible, Interoperable, and Reusable.

Findable

The dataset should have a clear title, searchable metadata, keywords, creator information, and a persistent identifier.

Accessible

The access process should be explained. Accessibility does not always mean immediate public download.

Interoperable

Files should use documented formats, clear variable names, and shared standards where appropriate.

Reusable

The dataset should include sufficient documentation, provenance, licensing information, and descriptions of known limitations.

FAIR does not mean fully open. Sensitive data can follow FAIR principles while remaining available only through controlled access.

Documentation Makes Data Reusable

A dataset without documentation may be technically available but practically unusable.

Researchers should consider providing:

• A README file
• A codebook
• Variable names and labels
• Units of measurement
• Missing-value codes
• A description of the sample
• A data-cleaning log
• Transformation rules
• Relationships between files
• Known limitations

Intervention studies may also need documentation of treatment conditions, implementation measures, assignment procedures, comparison groups, and deviations from the original protocol.

Sharing Analysis Code

A published table does not show every decision made between raw data and final results.

Analysis code may reveal:

• Which observations were excluded
• How variables were recoded
• How scales were created
• How missing data were handled
• Which statistical model was used
• Which robustness checks were completed
• How figures were produced

Shared code should include comments, software versions, package dependencies, execution order, and any configuration required to reproduce the main outputs.

The code does not need to resemble a commercial software product. It should be understandable enough for another researcher to inspect and run.

Open Materials and Measurement Instruments

Sharing research materials can reduce duplication and make replication easier.

Useful materials may include:

• Surveys
• Interview protocols
• Observation rubrics
• Lesson plans
• Intervention manuals
• Scoring guides
• Recruitment scripts
• Consent templates

Some materials cannot be shared freely. Standardized assessments may be protected by copyright or lose validity if secure items become public. Culturally sensitive materials may also require restrictions.

When full release is not possible, researchers can still provide a detailed description, scoring logic, sample items, and a clear access procedure.

Reproducibility and Replication

These terms describe different forms of verification.

Replication is particularly important in education because an intervention may work only with a specific age group, school system, instructor, or implementation model.

Open data and materials make replication easier. They do not guarantee that findings will repeat in every setting.

Reporting Null and Mixed Results

Studies with positive findings often receive more attention than studies with null or mixed results. This can create an exaggerated impression of how consistently educational interventions work.

Open-science practices can reduce this problem through:

• Preregistration
• Registered Reports
• Public study records
• Complete reporting of planned outcomes
• Repositories for reports and supporting materials

A null result does not prove that an intervention can never work. Researchers must still examine statistical precision, sample size, implementation quality, measurement, and the comparison condition.

Open Access for Educators and Policymakers

Teachers, school leaders, and policymakers may not have subscriptions to academic journals, even when those journals publish research directly related to their work.

Researchers can improve access through:

• Open-access journals
• Institutional repositories
• Accepted manuscripts
• Preprints
• Funder archives
• Plain-language practitioner summaries

Free access alone does not guarantee that research is understandable. Practitioner resources should explain the study context, implementation conditions, limitations, and practical relevance.

Research-Practice Partnerships

Open science should not operate only as file sharing among researchers.

Teachers, students, districts, and communities may help:

• Select useful research questions
• Define meaningful outcomes
• Interpret findings
• Identify privacy risks
• Establish responsible reuse rules
• Create practical resources

Partnerships should define roles, governance, ownership, publication rights, and procedures for disagreement before the project begins.

Equity in Open Science

Openness can create benefits, but its costs are not distributed equally.

Potential inequalities include:

• Repository and publication fees
• Unfunded data-curation work
• Limited technical infrastructure
• English-language dominance
• Unequal access to statistical expertise
• Communities contributing data without receiving benefits
• Limited recognition for data stewards and software developers

A responsible open-science policy should ask who receives value from shared research and who carries the financial, technical, and privacy risks.

Data Ownership and Stewardship

The idea that researchers simply own the data can be misleading.

Education data may involve students, families, teachers, schools, districts, governments, and communities. Legal control, ethical responsibility, and community interests may not belong to the same party.

Researchers should distinguish among:

• Legal ownership
• Data custody
• Stewardship responsibilities
• Participant rights
• Institutional interests
• Collective community interests

Even legally permitted data use may require ethical consultation when reuse could affect a recognizable community.

Risks of Open Data

Re-Identification

Removing direct identifiers may not prevent linkage with other datasets.

Misinterpretation

A secondary user may misunderstand how a variable was measured or apply the data outside the original context.

Stigmatization

Poorly framed comparisons may harm the reputation of a school, district, or community.

Surveillance

Research data collected to understand learning may later be used to monitor, rank, or punish participants.

Unequal Credit

Researchers who collect and curate data may receive little recognition when others publish secondary analyses.

Responsible openness therefore requires governance, not access alone.

Controlled Access and Data-Use Agreements

A data-use agreement can define:

• The approved research purpose
• Authorized users
• Security requirements
• Prohibited data linkage
• Disclosure-review procedures
• Publication conditions
• Retention and destruction rules
• Responses to security breaches

Controlled access should not be treated as a failure of openness. It can enable valuable secondary research while protecting participants.

Academic Incentives and Research Culture

Researchers may hesitate to share because curation takes time, institutional support is limited, or academic promotion systems reward articles more than datasets, software, replications, and public engagement.

Research assessment should recognize:

• High-quality datasets
• Reusable software
• Research protocols
• Replication studies
• Registered Reports
• Data stewardship
• Community engagement

Expecting openness without funding and recognition transfers additional labor to researchers without changing the system that shapes their decisions.

Teaching Open Science

Open-science skills should be included in doctoral training, research methods courses, research assistant preparation, and faculty development.

Useful assignments include:

• Writing a data management plan
• Creating a codebook
• Preregistering a small study
• Reproducing a published table
• De-identifying a practice dataset
• Preparing a repository README
• Reviewing a consent form for data-sharing language

These tasks turn transparency from an abstract principle into a practical research skill.

A Practical Open-Science Workflow

Before the Study

• Involve relevant stakeholders.
• Define planned research outputs.
• Assess privacy risks.
• Prepare a data management plan.
• Preregister where appropriate.
• Include clear consent language.

During Data Collection

• Use consistent file structures.
• Separate direct identifiers.
• Maintain secure backups.
• Document changes to the protocol.
• Record data-quality problems.

During Analysis

• Preserve raw data.
• Use scripted transformations where possible.
• Document exploratory decisions.
• Version analysis code.
• Run disclosure-risk checks.

Before Publication

• Verify reported results.
• Prepare code and research materials.
• Create documentation.
• Select the appropriate access level.
• Write limitations and a plain-language summary.

After Publication

• Maintain repository records.
• Correct errors transparently.
• Respond to controlled-access requests.
• Track reuse and citations.
• Update documentation when necessary.

What Can Be Shared?

Questions Before Releasing Education Data

• What public benefit can sharing create?
• Did participants consent to the proposed use?
• Can individuals, classrooms, or schools be recognized?
• Could linkage with other datasets increase risk?
• Which variables are necessary for reuse?
• Would controlled access be safer?
• Is the documentation sufficient?
• Who will review future access requests?
• How will data creators receive credit?
• Could reuse harm a represented community?
• What happens if an error or privacy problem is discovered?

What Open Science Cannot Solve Alone

Transparency cannot automatically repair:

• A weak research question
• A biased sample
• An invalid measurement instrument
• Poor intervention implementation
• Unfair research relationships
• Limited practical relevance
• Institutional pressure for positive findings

Open data can make weak research easier to inspect. It does not transform weak research into strong evidence.

Openness improves the conditions for scrutiny. It does not replace methodological quality, ethical judgment, or respect for participants.

Conclusion

Open science can make education research easier to verify, reproduce, compare, and apply. Preregistration clarifies what was planned before results were known. Shared code and documentation reveal how findings were produced. Open access makes research more available to educators and communities.

Education data cannot be opened mechanically. They often represent children, small classrooms, schools, and communities that may remain identifiable even after names are removed.

The goal is not maximum openness at any cost. It is responsible openness: research questions declared clearly, methods documented honestly, findings made accessible, and data shared at a level that supports verification and reuse without sacrificing privacy, consent, or public trust.