「こっちもかけてみようか」
segment 93 · 329.30s
Open questions · Human–AI collaboration
Return what the video could not tell us to the skill graph.
The AI formulates questions from the gap between what the source supports and what a learner still needs. An expert answer becomes a provenance-bearing graph node and updates the related process context.
Progress
0 / 9
01 · QUESTION FORMULATION
How an open question is formulated
The system does not plausibly fill gaps. It fixes the source-supported boundary, then asks only for learning-critical fields that remain unstated.
- 01Perceive video and speech
Ground speech, on-screen text, entities, and motion in time
- 02Structure the graph
Separate process, action, attention, and criteria
- 03Detect a missing field
Find absent criteria, values, exceptions, rationale, or identity
- 04Formulate a neutral question
Avoid leading the answer and retain source links
- 05Apply the human answer
Add accepted evidence and mark the question resolved
02 · OPEN QUESTIONS
Open questions in the skill graph
Select a question to inspect why it was generated, its evidence, and the exact graph patch an answer will create.
Tacit sensory cue
OpenWhat tactile difference tells the practitioner they are feeling the face correctly?
Why it was generated
The action is visible, but the source does not state which tactile, acoustic, or visual difference lets an expert judge it.
What the source establishes
Finish an L-shaped corner — Straight passes from two directions can leave the corner untouched. Present the file at 45°, connect the faces, and use the safe edge to protect the adjacent face.
Learning-critical field missing from the source
Tacit sensory cue
Evidence IDs2:32
speech-l-edge-reasonExpert answer
03 · GRAPH UPDATE
Live graph patch
Applying an answer immediately changes the question state and adds a human-verified evidence node.
LEARNING UNITFinish an L-shaped cornerRelated learning unit
RAISES_QUESTION
OPENQUESTIONWhat tactile difference tells the practitioner they are feeling the face correctly?Awaiting an answer
SUPPORTED_BY
EVIDENCEITEMAwaiting an answerHuman-verified evidence
04 · SOURCE CHECKS
Supporting source checks
Transcript, OCR, and neutral-ID checks are a separate provenance gate—not answers to the skill questions. They remain available below.
Progress0 / 5
1. Audio spot checks
Listen and confirm whether the displayed source-language text matches the recording.
「それがこれよさっきの」
segment 28 · 98.74s
「当ててこれかけとる こっち側傷つかないんだ こっち側しか削れない」
segment 45 · 158.12s
2. Confirm the visual correspondence once
This fixes a neutral tracking ID; it does not infer what the work means.
object_1May the highlighted part held in the hand be assigned the neutral ID object_1?
tool_1May the highlighted long, narrow vertical tool be assigned the neutral ID tool_1?
Continue with this source
One source, five connected views
Switching views preserves the same source hash across instruction, analysis, skill structure, evidence, and human review.
01Learning guideLearn the demonstration as a visual procedure02Vision labInspect model outputs, live graph, and teaching visuals03Skill mapExplore process, tools, judgment, and learning04EvidenceReturn to speech, source text, and observations
05Open questionsAnswer knowledge gaps and update the skill graphYou are here