Depending on whether a given experiment is for just learning the system, a small personal experiment, or a long-term collaborative experiment, different “experiment frameworks” will be necessary.
iSENSE will support these different types:
- Sandbox Experiment
- This is the experiment to use when you are first learning the system, or trying something new where you don't expect to ever share the results. Sandbox experiments have minimal set-up requirements (fewer options to configure) and the data captured are kept privately by default.
- Personal Experiment
- This is an experiment where you have some idea what you are planning to do. Only one user can contribute data to this experiment (the person who creates it). Data are published by default.
- Open Experiment
- This is a shared experiment where others can contribute to the data set. The experiment should have clear guidelines so that others who are participating can do so properly. Data are published by default, and anyone can contribute data to the experiment without explicit permission.
- Invitational Experiment
- This is a shared experiment where you need an invitation code to join. These should have very clear instructions that have been tested with users previously, so it is more likely to result in high quality data. Data are published, so anyone can view the results, but you do need the code in order to contribute data.
Notes from Dec 6 phone conf:
- what parameters can experiments define?
- experiments themselves can be shared objects with read/write permissions.
- read/write permissions to data.
- configuration of visualizations -- is that part of the experiment, or it is somewhere else?
- how are you going to save & represent each data artifact?
- can you take data from old experiments and incorporate into new ones?
- these 4 are presets from a bunch of configurable parameters.
Notes from jan 11 conc consortium mtg:
- peer review of experimentation procedures was really valuable (CLEO / global lab project)
- all the kids want to be a chief, but they have to collect data for others' experiments as well
