How to brief a first-person data collection

Begin with the research question

Describe the intended use in plain language. A team studying task order may need complete sequences. A team studying object interaction may care more about the moment a hand meets a tool or material. Explain what your model or reviewer must be able to distinguish. First-person video records an observed human demonstration; it does not, by itself, provide robot joint commands, forces or executable actions. Identify any additional signals your project would require and discuss their feasibility separately.

Define one complete example

Choose a narrow task and write its beginning and end. For a packing sequence, the beginning might be an unpacked item beside suitable materials; the end might be the item secured in a closed package. Specify whether preparation, inspection, pauses and corrections belong in the recording. This prevents two collectors from interpreting the same task name differently.

• Name the objects, tools and materials needed to understand the task.
• Describe the starting state and the visible completion condition.
• State which interruptions or unsuccessful attempts should be retained.

Choose variation deliberately

Separate what should remain consistent from what should change. You might hold the task sequence constant while varying object shape, workspace layout or demonstrator. Avoid changing every factor in the first pilot: it becomes difficult to identify why an example is unusable. Write down the variation you need and the reason for it. If evaluation will compare unfamiliar people or environments, discuss how those groups should be identified and kept distinct in the delivery.

Specify the view and the handoff

Explain what must remain visible, especially hands, contact points and the objects being manipulated. Discuss viewpoint, resolution, frame rate, audio and lighting as requirements to validate, rather than selecting them from habit. Agree a file naming convention, task identifiers, metadata fields and any annotation format. Include participant permissions, location access and treatment of incidental personal information in scoping; these affect where and how recording can happen.

Use the pilot to settle open questions

Ask for a pilot specification that lists representative examples, review responsibilities and acceptance rules. Give your reviewers a small set of explicit questions: is the relevant interaction visible, is the sequence complete, and can the supplied labels be used in the intended workflow? Record changes to the brief before expanding collection. A short versioned document is more useful than scattered feedback that different teams interpret differently.