Evidence-grounded session-end coaching
A planned trainer that checks observations and recommendations before giving feedback.
My contribution
Design a session-end workflow for behavioral analysis, reference retrieval, feedback drafting, and verification.
From observation to justified advice
The proposed trainer starts after a complete simulator session. It connects verified observations with applicable task guidance and expert examples, drafts feedback, and checks whether each recommendation is supported. A session report should prioritize at most two practice targets and support an insufficient-evidence outcome.
Evaluate the feedback itself
Correct phase labels alone do not establish useful coaching. The planned evaluation examines factual accuracy, specificity, actionability, task appropriateness, and the ability to withhold unsupported advice. Template-based feedback is a comparison condition. Experienced-operator reviews provide reference judgments alongside independently computed measures.
Test learning prospectively
Repeated novice practice sessions are a proposed follow-up. They would assess whether evidence-grounded feedback changes subsequent performance. Existing single-session recordings support retrospective analysis, not a claim that the trainer has improved learning or field safety.
Proposed architecture
A planned design. It shows what the study would test, not a result.
Advice only where the evidence supports it
A planned trainer. Verified observations meet applicable guidance, a draft is checked for support, and the report gives at most two priorities or says the evidence is insufficient.
- Proposed step, no result
Verified observationsProposed
- Input
- Observations from a complete simulator session, validated in the prior stage.
- Computation
- None.
- Output
- Evidence for feedback.
- Source
- Doctoral research plan, Paper 3
Applicable guidanceProposed
- Input
- Verified observations.
- Computation
- Selects task guidance and expert examples that apply.
- Output
- Relevant guidance.
- Source
- Doctoral research plan, Paper 3
Draft feedbackProposed
- Input
- Observations and guidance.
- Computation
- Drafts session-end feedback.
- Output
- Draft recommendations.
- Source
- Doctoral research plan, Paper 3
Support verificationProposed
- Input
- Draft recommendations and their evidence.
- Computation
- Checks whether each recommendation is supported.
- Output
- Supported recommendations, or none.
- Source
- Doctoral research plan, Paper 3
Report with up to two prioritiesProposed
- Input
- Supported recommendations.
- Computation
- Prioritizes at most two practice targets.
- Output
- Session report.
- Source
- Doctoral research plan, Paper 3
Insufficient evidenceProposed
- Input
- No supported recommendation.
- Computation
- Withholds advice.
- Output
- An explicit insufficient-evidence outcome.
- Source
- Doctoral research plan, Paper 3
Retrospective feedback reviewProposed
- Input
- Generated and template-based feedback.
- Computation
- Rates factual accuracy, specificity, actionability, task fit and withholding, with experienced-operator reviews.
- Output
- Feedback quality evidence. Not yet conducted.
- Source
- Doctoral research plan, Paper 3
Prospective novice practiceProposed
- Input
- Repeated novice practice sessions.
- Computation
- Tests whether feedback changes later performance.
- Output
- Proposed follow-up. No learning gain is claimed.
- Source
- Doctoral research plan, Paper 3
Adapted from PhD Research Summary, Asif Faisal Chowdhury, Proposed feedback workflow and repeated-practice follow-up.
Existing single-session recordings support retrospective analysis only. No learning or safety improvement is claimed.
Animated pulses show conceptual signal flow only. They do not show measured execution time or a live inference run.
Evaluation Scope & Methodology
Samples, denominators, uncertainty, sources and limits
Approach
Proposed retrospective feedback review followed by a separate prospective novice-practice study; no learning outcomes reported.
Limitations
- Proposed workflow and evaluation, not an achieved training intervention.
- Existing single-session recordings cannot establish learning gains.
- Experience-group differences do not by themselves identify useful training targets.
- Prospective repeated novice practice and expert review remain to be conducted.
Sources
- PhD Research Summary, Asif Faisal Chowdhury, Proposed feedback workflow and repeated-practice follow-up (Supplied manuscript reviewed November 2026). Public release pending
- Portfolio website content and coding-agent brief, Paper 3 plan; feedback and prospective evaluation (Supplied manuscript reviewed November 2026). Public release pending