Episode 12.4: Types of Open Practices I: Preregistration & Registered Reports; Open Data (FAIR & CARE)
Last updated on 2026-09-08 | Edit this page
Estimated time: 19 minutes
Overview
Questions
- What’s the difference between preregistration and a Registered Report?
- What does it actually take for data to be “open” in a useful way?
- What are the FAIR and CARE principles, and why do we need both?
Objectives
Learners will be able to:
- Distinguish preregistration from Registered Reports, and explain what problem each solves.
- Apply the FAIR principles to evaluate whether a dataset is genuinely reusable.
- Explain what the CARE principles add that FAIR alone does not cover.
- Connect each practice to a real research example.
Think Like a Researcher
Picture five different teams, each convinced their vaccine hesitancy study is rigorous:
One team collects their data first, looks for whatever patterns turn up, then writes the hypothesis to match what they found. A second team locks in their hypothesis and analysis plan on a public registry weeks before touching any data. A third team goes further still: a journal reviews and provisionally accepts their entire study design before a single data point exists. A fourth team finishes their study and, when asked, says the data is “available upon request.” A fifth team uploads their dataset to a public repository, with every variable clearly labelled, in a format anyone can open.
All five believe they’ve done solid research. Only some of them have actually protected their work from a specific kind of bias, and only one has made their data something a stranger could genuinely reuse. This episode looks at exactly how and why.
Preregistration and Registered Reports
Preregistration: This means publicly and permanently recording a hypothesis and analysis plan before collecting data.
It exists to solve a specific, well-documented problem: it’s very easy, without intending to, to notice a pattern in your data and then write it up as though you predicted it all along. Preregistration draws a hard line between what you predicted and what you noticed afterwards, because once it’s timestamped, it can’t quietly change.
Example: Before surveying a single household, a research team registers their prediction (that fear of side effects is the strongest driver of vaccine hesitancy in their study area) along with exactly how they’ll measure and test it.
Registered Reports: This takes preregistration a step further. The entire study (the question, the design, the planned analysis) is submitted to a journal and peer-reviewed before any data exists. If reviewers approve the design, the journal commits, in principle, to publishing the results regardless of whether they turn out to be exciting, boring, or a flat contradiction of the hypothesis.
Registered Reports exist to fix a different problem: journals have historically been far more likely to publish “positive,” surprising results than clear null findings, which quietly pushes researchers toward chasing whichever result is publishable, rather than whichever result is true.
Example: A team wanting to test whether a short video reduces vaccine hesitancy gets their entire study design reviewed and accepted first. Whether the video works, doesn’t work, or makes no measurable difference, the result gets published either way.
Open Data: FAIR and CARE
Publishing a data file isn’t automatically the same as making data genuinely usable. Two sets of principles describe what “usable” actually requires.
FAIR: Findable, Accessible, Interoperable, Reusable
Findable: the dataset can actually be located, usually through a stable identifier and clear metadata (not buried in a personal folder no one else can find). Accessible: it can be retrieved through a known process, even if some access conditions apply for sensitive data. Interoperable: it uses standard, well-documented formats, so it can be combined with other datasets without extensive reformatting. Reusable: it’s documented clearly enough (units, variable meanings, collection method) that someone else can understand and reuse it correctly.
Example: The vaccine hesitancy team’s dataset has a permanent identifier and a full codebook explaining every column. A researcher on the other side of the country can locate it, download it, understand exactly what “hesitancy score” means, and reanalyse it without ever contacting the original team.
CARE: Collective Benefit, Authority to Control, Responsibility, Ethics
FAIR describes how usable data is. CARE, developed particularly around Indigenous and community data, addresses something FAIR doesn’t: who the data is about, and who gets a say in how it’s used.
Collective Benefit: the data should generate real benefit for the people or communities it describes, not just for outside researchers. Authority to Control: the people or communities the data comes from have a legitimate say in how it’s governed and shared. Responsibility: those handling the data are accountable for how they use it and what it’s used for. Ethics: the community’s rights and wellbeing come first, ahead of convenience for the research team.
Example: If the vaccine hesitancy study collected data from a specific community, CARE would ask the team to involve that community in decisions about how the data is shared, not just make it technically FAIR-compliant and call it done.
FAIR and CARE Together
FAIR without CARE can produce data that’s technically easy to reuse, but taken from communities who had no say in it. CARE without FAIR can produce data that’s ethically governed, but practically impossible for anyone to find or use. Well-designed open data practice treats them as complementary, not as a choice between one or the other.
Test Your Knowledge!
Challenge 1:
A researcher analyses their data first, notices an unexpected pattern, and then writes up the paper describing that pattern as their original hypothesis. What open practice would have prevented this, and how?
Preregistration. Recording the hypothesis and analysis plan before collecting data creates a timestamped record, so it’s clear afterward what was predicted in advance versus noticed only once the data was in.
Challenge 2:
What is the key difference between preregistration and a Registered
Report?
A. Registered Reports don’t require a hypothesis. B. In a Registered
Report, the study design is peer-reviewed and provisionally accepted
before data collection; preregistration alone does not involve journal
review. C. Preregistration is only used in qualitative research. D.
There is no real difference.
B. Both lock in a plan in advance, but a Registered Report adds formal peer review and provisional publication acceptance before any data exists.
Challenge 3:
A dataset has a permanent link and a detailed codebook, but the only way to request it is by emailing the original author, who left the institution two years ago and never replies. Which FAIR principle is most clearly failing here?
A. Findable B. Accessible C. Interoperable D. Reusable
B. Accessible. The data may be findable (it has a link) and well-documented, but there’s no working way to actually retrieve it.
Challenge 4:
True or False: FAIR data principles and CARE data principles are trying to solve the same problem.
False. FAIR is about making data technically usable and discoverable. CARE is about who has a say in how data (especially data about specific communities) is governed and used. They’re complementary, not interchangeable.
- Preregistration timestamps a hypothesis and analysis plan before data collection, separating genuine predictions from patterns noticed after the fact.
- Registered Reports go further, having the study design peer-reviewed and provisionally accepted before results exist, which removes the incentive to chase “exciting” results over accurate ones.
- FAIR (Findable, Accessible, Interoperable, Reusable) describes what makes a dataset genuinely usable by others.
- CARE (Collective Benefit, Authority to Control, Responsibility, Ethics) addresses who has a say in how data about people and communities is governed: a question FAIR alone doesn’t answer.
- Strong open data practice applies both FAIR and CARE together.
Inline instructor notes…
To-do: Add infographic.
💡 Preregistering a hypothesis doesn’t mean a study can’t explore anything unplanned. It just means the write-up has to be honest about which parts were predicted in advance and which were exploratory.