From Open Education to Open Research: The Next Frontier

What is being shared

Research produces more than articles. Behind a published result there are usually measurements, survey responses, images, recordings, spreadsheets and code that cleaned and analyzed them. Research data sharing is the practice of making some of that material available to others, ideally with enough explanation that they can use it.

Open education encourages teachers to share lesson materials so that others do not start from zero. Sharing data and code applies the same logic to research. A dataset collected with considerable effort can answer questions its creators never asked, if other people can get hold of it.

Why sharing can help

Several benefits are commonly cited. They are real in many cases, though not automatic.

• Checking. Others can rerun the analysis and look for errors, which supports reproducibility.
• Reuse. A dataset may be combined with others, used for a new question, or used to test a new method without repeating the costly collection.
• Teaching. Real datasets give students material to practice on, which is more engaging than invented numbers. Many learners first meet open data in this way.
• Efficiency. Funders often prefer that expensive data collection be used as widely as possible.
• Recognition. A well-documented dataset can be cited, giving credit to the people who gathered it.

Evidence on how much shared data is actually reused is mixed. Some datasets are used many times; many are never downloaded. Claims of large savings or breakthroughs should be treated cautiously.

Making data usable

Putting a file online is not the same as sharing it well. A dataset that nobody can interpret is of limited use. Good practice generally includes:

1. A clear description. Say what the data are, when and how they were collected, and by whom.
2. A data dictionary. Explain each column or variable, the units, and how missing values are marked. A column called “score” is ambiguous without one.
3. Open, non-proprietary formats. Plain text formats such as CSV are readable by many programs and are likely to remain readable for years.
4. Consistent organization. Sensible folder structures and file naming make it easier for others to find things.
5. A license. Without a stated license, people do not know whether they may reuse the material. Questions of copyright and licensing apply to data too, and the rules differ between places, so check official guidance for your own situation.
6. A persistent identifier. Depositing data in a trusted repository, which gives it a stable address, is better than hosting it on a personal page that may disappear.

Many researchers use the phrase “findable, accessible, interoperable and reusable” to summarize these goals. It is a useful checklist, even if the details vary by field.

Sharing code

Analysis code is often as important as data. Two people given the same dataset can arrive at different numbers if they clean it differently, and the code records exactly what was done. Shared code is most helpful when it includes a short note about how to run it and what software it needs.

Version control tools record the history of changes to code, which helps others see what was altered and when. Releasing code under an open-source software license tells others what they may do with it.

Researchers sometimes worry that their code is untidy. Most readers prefer imperfect shared code to polished code that stays hidden, though it should be honest about its limits.

Barriers to sharing

There are good reasons researchers hesitate, and a fair account must include them.

• Privacy and consent. Data about people can be sensitive. Removing names is often not enough, because combinations of details can sometimes identify someone. Where consent or ethical approval does not allow sharing, the data should not be shared, or only through controlled access.
• Commercial or security limits. Agreements with partners can restrict release. Some data, such as the location of vulnerable species or sites, can cause harm if made public.
• Time and skill. Cleaning and documenting data takes effort that is not always funded or rewarded.
• Fear of being scooped or criticized. Some researchers worry others will publish first or find mistakes. Careful acknowledgment and citation norms are meant to address this, though they do not remove every concern.
• Ownership. Data may belong to an institution, a funder or a community, not just the researcher. Indigenous and other communities may have their own expectations about how information about them is handled, and these deserve respect.
• Unclear incentives. Hiring and promotion often prize articles over datasets.

A commonly used summary is “as open as possible, as closed as necessary”. It reminds us that open does not mean unconditional.

Finding and reusing open data as a learner

If you want to use shared data, look first for repositories maintained by universities, libraries or government statistics offices. Read the description and the license before you begin. Check the date, how the data were collected and what the limitations are. Credit the source in the way the license or the dataset’s own citation guidance requests.

Be a careful consumer. A tidy-looking dataset may contain errors, and a chart made from it can mislead. The skills of information literacy and of reading graphs apply just as they do to any other source. Remember also that correlations found in a dataset do not by themselves show cause.

A sensible starting point

If you are creating data, even for a small class project, start good habits early: keep raw files unchanged, record what you did to them, name files consistently, and write a short description as you go. Doing this while the work is fresh is far easier than reconstructing it later. Whether the final data can be shared openly is a separate decision that depends on consent, ethics and agreements, but well-organized work is easier to share whenever sharing is appropriate.