01 / CHOOSE A QUESTION
One study at a time.
A defined species, observable behavior and short protocol. Know what to observe, what to leave unknown and when to stop.
FIELD NOTES · IN DEVELOPMENT
A companion for observing animals carefully, following a study protocol and turning everyday moments into evidence researchers can examine.
01 / CHOOSE A QUESTION
A defined species, observable behavior and short protocol. Know what to observe, what to leave unknown and when to stop.
02 / OBSERVE FIRST
Keep the animal, session and observation connected. Record visible behavior before seeing any model prediction. Short, reviewed clips are a later step.
03 / LEARN TOGETHER
Check rights, privacy and label quality. Separate permission for private review, research training and possible publication. Make withdrawal part of the workflow.
A USEFUL FIRST EXPERIMENT
A reviewed research lead asks whether calls predict video-observed social context in Egyptian fruit bats. Current grouping checks leave no balanced strict holdout at this annotation level, so no first ML benchmark is selected and no Talk2Nature model has been run.
This is separate from a possible phone-based budgerigar observation study, which depends on household access and avian review. A context prediction would not establish what an animal meant.
Read the bat-candidate assessment → See the separate first-animal plan →BUILT AROUND A RESEARCH RECORD
The planned research database connects papers, code, datasets, study protocols, sessions, observations, permissions and review decisions. Recordings will need separate private storage, linked to those records by checksums and access rules.
We start with a phone-friendly web app. Native background recording, spare-phone stations and a paired TV display are later options, each with its own device and welfare testing. Initial studies use passive observation; no automatic playback.
HELP SHAPE THE FIRST STUDY
Research protocols, annotation workflows and careful code reviews are useful contributions now. Please keep private recordings and personal information out of public issues.
Help build Field Notes ↗