Over this series I've argued that research needs an open, shared way to record how manual experiments are actually executed. It would help reproducibility, preserve expert knowledge, smooth handoffs from people to instruments, and give tool makers honest feedback, all without touching scientists' results.
Arguments only go so far. Now it's time to build it, and I'd like to do that with you.
Why this matters now
Instruments are getting a shared language. In August, Anthropic previewed the Model Hardware Standard, "a shared specification for AI agents to safely operate physical devices," covering things like microscopes, liquid handlers, and plate readers. Whatever you think of any one company's approach, the direction is clear: the machine side of the bench is being standardized, and fast.
The human side is not. The steps people perform by hand, the pipetting, the incubation that ran long, the substitution someone made because the buffer was out, still live in notebooks, memory, and email. If methods are going to move between people and instruments, the manual steps need a common description too.
There is a second reason to move early. If no open standard exists, the gap will still get filled, just by whichever vendor or platform gets there first, in a format only that product reads. Standards written after the fact tend to inherit those choices. A standard written now, by the people who run the experiments, builds the shared foundation before proprietary formats harden around it.
What we're forming
I'm forming a working group within the Research Data Alliance, an international community that develops shared approaches to research data. RDA working groups are time-limited teams of experts, expected to deliver their Recommendations within 12 to 18 months. That fits the job: write a focused version 0.1 and test it against real work.
Each group starts with a Statement of Work, which goes through community review before the group is formally recognized. That is a feature, not a hurdle. It means the scope gets challenged by people outside our circle before we begin.
We haven't chosen where the work will be hosted. My view is that a standard like this ought to live in a public open-source repository, so anyone can read the draft, comment, and propose changes. We'll settle that together with the group.
Who we need
- Lab automation teams working on handing methods from people to instruments
- Core facility staff who run standardized protocols for many clients
- Identifier and metadata experts working on RRIDs, PIDINST, and related standards
- Protocol repositories and publishers who could host and link execution records
- Tool manufacturers who want better insight into how their products are used
- Bench scientists who will tell us, bluntly, what's worth recording and what isn't
What we'd do first
Proposed first five steps
- Review existing standards and tools that touch execution records, and map where they overlap
- Agree on the minimum fields a useful record needs
- Map those fields to RRIDs, PIDINST, and protocol repositories
- Set privacy and consent principles
- Publish a version 0.1 draft with worked examples for open comment
This is a starting proposal. The group will shape it.
How to join
I'm looking for co-chairs and founding members, especially outside North America, to help shape the Statement of Work. You don't need to be an expert in everything on the list above. You need to care about how experiments actually get done. Individual membership in the Research Data Alliance is free.