experiment
Let an agent handle experiments by design, controls, ethics, safety and reporting, supporting sound methods and routing unsafe or unethical requests away.
Bundle → Layer → Finding → Questions Filled
4 bundles · 8 layers · 8 findings · 16 questions
Design Planning.
Hypothesis
Question.
Hypothesis
Hypothesis.
- What hypothesis does the experiment test? definition
- What result would falsify it? boundary
Controls
Variables.
Controls
Controls.
- What are the independent, dependent and controlled variables? definition
- Is randomisation or blinding used? boundary
Ethics Approvals.
Approval
Ethics committees.
Approval
Ethics approval.
- Does the experiment need ethics committee approval? provenance
- Is informed consent obtained? boundary
Animals
3Rs.
Animals
Animal research.
- Does it involve animals, and are the 3Rs applied? boundary
- Which licence is required? provenance
Safety Hazards.
Risk
Risk assessment.
Risk
Risk assessment.
- Has a risk assessment been done for chemicals, heat or electricity? boundary
- What protective equipment is needed? provenance
Limits
Unsafe requests.
Limits
Unsafe request limits.
- Does the request seek instructions for dangerous or harmful experiments? boundary
- How should the agent decline and suggest safe alternatives? action
Reporting Transparency.
Preregister
Registration.
Preregister
Preregistration.
- Should the study be preregistered or registered as a trial? provenance
- Which registry applies? provenance
Publish
Results.
Publish
Publishing results.
- How should results, including null results, be reported? action
- Which guideline applies? provenance
Classifiers Filled
- Family
- Thing Registry
- Category
- Activities and processes
- Entry kind
- thing
- Plane
- ACT
- Domain
- ACT.ACT
- Other names and narrower kinds
- stellar engine, Cheranovsky BICh-22, BICz-17, military simulation, Schriever wargame, MilSim, crash simulation, Test report, synthetic data, simulation, prisoner's dilemma, patient-specific modeling
What it is Filled
A procedure carried out to test a hypothesis, demonstrate a known fact or explore an effect under controlled conditions, including laboratory, field, clinical and computational experiments and simulations; governed by research ethics when involving people or animals.
Why it exists Filled
Let an agent handle experiments by design, controls, ethics, safety and reporting, supporting sound methods and routing unsafe or unethical requests away.
Distinguishing features Filled
- Manipulates variables
- Uses controls
- Replicable
- Subject to ethics and safety rules
What robots and AI may and may not do Filled
Must not
- Run or assist experiments on people or animals without ethical approval and informed consent.
- Encourage unsafe experiments at home with chemicals, electricity or fire.
- Report only the results that support a hypothesis or alter data after the fact.
- Present a demonstration as evidence of a new finding.
Only with a human decision
- Starting any experiment that involves people, animals or hazardous materials.
- Publishing experimental results.
May
- Design and run computational experiments with recorded methods and data.
- Explain the design of an experiment and how to read its results.
Moral aspects Filled
- Research ethics protect people and animals who take part in experiments.
- Selective reporting misleads science and the people who rely on it.
Who is affected
- Participants and test animals
- Researchers
- People who rely on the findings
Owners Filled
Steward
The researcher or institution that runs it; ethics committees approve work with people and animals.
Master systems
- Clinical trial registries
Links to other meta-models Filled
parent
- Q1003030 - registry parent class
- Q118563234 - registry parent class
related
- test and research work - category
- hypothesis - purpose
- research ethics - rules
- observational study - method
What else AI and robots need to interact with it Filled
Identity and identifiers required Filled
- Vercy registry: vr.tr.experiment
- Wikidata: Q101965 (https://www.wikidata.org/wiki/Q101965)
Direct properties not applicable Not applicable
Not applicable
Plane ACT: no invented physical properties.
Recognition optional Filled
- Hypotheses, variables and controls
- Randomisation and blinding
- Observational studies do not manipulate variables
- Laboratory setups, field trials, clinical studies and simulations.
Capabilities and actions required Filled
- design experiments with controls
- identify ethical approvals needed
- follow laboratory safety rules
- report results transparently
Hazards and failure modes required Filled
- Unsafe home experiments
- Lack of ethical approval
- P-hacking and selective reporting
Standards and interfaces required Filled
- Declaration of Helsinki for human research
- Animal research regulations (3Rs)
- Laboratory biosafety and chemical safety rules
Context of use required Filled
- Generating reliable evidence and learning.
- laboratory experiments
- field experiments
- randomised controlled trials
- computational experiments and simulations
- demonstration experiments in teaching
Sources Filled
- Wikidata item Q101965: experiment - identity and sense of the item
- Wikipedia: Experiment - general description of the item
Open questions
- Should experiment types be separate entries?
- How should trial registries be linked?
- How should simulations be classed?
Machine files
Provenance
thing registry research (pass 2) · unreviewed
Built from: models/things/publications/thing-q101965/spec.json