Practice — Adaptive Mock Interview Sessions

ES Reference: Phase 3 (INTOS-ES-P3-001) — Adaptive Practice and Interview Outcome Learning

Overview

Practice mode transforms your preparation content into demonstrated interview capability. Phase 3 introduces consent-driven audio mock interviews, an adaptive evidence-aware AI interviewer, transcription with rubric-based evaluation, and practice analytics that feed into readiness and Next Best Action.

Practice is not a hidden real-interview answer feed, employer proctoring tool, or deception detector. It is a safe, private environment to build and demonstrate interview skills.

Key Principles

• Explicit consent before any recording — consent version and timestamp recorded per session
• User controls retention and deletion — audio, transcript, and feedback are independently deletable
• No facial, emotion, accent, personality, or protected-attribute analysis — ever
• Evaluation is transcript-linked and rubric-versioned — you can challenge and correct
• Adaptive, not adversarial — the interviewer probes, challenges, and follows up, but is never hostile
• Text-only fallback — practice is available even without a microphone (with disclosed limitations)

Practice Session Lifecycle

PLANNED → DEVICE_CHECK → CONSENTED → IN_PROGRESS → PROCESSING → EVALUATED
              ↓              ↓              ↓
          ABORTED        FAILED → RETRYING
EVALUATED → ARCHIVED → DELETED_CONTENT

Screens in This Feature Area

1. Practice Sessions List (/dashboard/practice)

Overview of all your practice sessions with status badges and performance summaries.

Buttons & Actions:

Button	Behavior
New Session	Configure and start a new practice session
Resume Session	Continue a paused session
View Report	Open evaluation and feedback for a completed session
Filter by Job/Kit	Filter sessions by associated job or kit
Archive	Archive an old session

2. New Practice Session Configuration (/dashboard/practice/new)

Configure session parameters before starting.

Buttons & Actions:

Button	Behavior
Select Job/Kit	Choose the job and kit to practise for
Select Round	Pick interview round: recruiter, hiring manager, technical, behavioural, leadership, case, board
Select Persona	Choose interviewer persona matching the round type
Set Duration	Configure session length (short/standard/extended)
Set Focus	Optionally focus on specific competencies or question types
Start Session	Proceed to device check and consent

3. Device Check & Consent (/dashboard/practice/[id]/setup)

Pre-session device verification and consent capture.

Buttons & Actions:

Button	Behavior
Test Microphone	Verify microphone access and audio quality
Test Speakers	Verify audio output
Troubleshoot	Guided troubleshooting for permission/device issues
Review Data Use	See what data is collected, how it's used, and retention policy
Give Consent	Explicit consent to recording (required before microphone access)
Start Text-Only	Skip audio and proceed with text-based practice
Cancel	Abort session setup

What's Explained Before Consent:

• Audio is recorded and transcribed
• Transcription provider and data processing
• You can delete audio while keeping feedback, or delete everything
• Retention policy and deletion rights
• No emotion, accent, or protected-attribute analysis

4. Active Practice Session (/dashboard/practice/[id])

The live adaptive interview interface.

How the Adaptive Interviewer Works:

1. Plan → Session plan generated from RoleModel, round type, prioritized questions
2. Opening Question → Calibrated to your level, round type, and focus area
3. Your Response → Recorded/typed, transcribed if audio
4. Response Analysis → Transcript + response analyzed for content, specificity, evidence alignment
5. Policy Decision → Interviewer chooses: follow-up (probe deeper), challenge (test robustness), new competency (breadth), or clarification
6. Next Question → Interviewer asks next question, building on context
7. Repeat → Within turn, duration, and cost budget
8. Closing → Interviewer wraps up; option to ask your own questions
9. Evaluation → Asynchronous rubric-based scoring and feedback

Buttons & Actions (During Session):

Button	Behavior
Start Recording	Begin audio capture for current turn
Stop Recording / Submit	End recording and submit response for analysis
Pause Session	Pause at any time (preserves media chunks and sequence)
Skip Question	Move to next planned question
End Session	Complete the session and trigger evaluation
Report Issue	Flag a problematic question (discriminatory, inappropriate, etc.)
Emergency Stop	Immediately stop recording and abort (no data retained beyond this point)

5. Interviewer Personas

Persona	Style	Typical Use
Recruiter	Screening, logistics, motivation, fit	Initial screens
Hiring Manager	Role fit, experience depth, team fit	Manager rounds
Technical/Architecture	Domain expertise, system design, problem-solving	Technical rounds
Behavioural/Leadership	People management, conflict, culture, decisions	Leadership rounds
Case/Presentation	Analytical thinking, structured communication	Consulting/strategy rounds
Board/Executive Panel	Strategy, governance, P&L, transformation	C-suite/board

6. Post-Session Evaluation (/dashboard/practice/[id]/report)

Comprehensive evaluation with transcript-linked feedback and rubric scores.

Evaluation Dimensions:

Dimension	What's Evaluated	Requires Audio?
Relevance	Directly addresses question and competency	No
Structure	Logical CAR/STAR/executive flow	No
Specificity	Concrete context, actions, and decisions	No
Evidence Alignment	Alignment with verified evidence; unsupported-claim flags	No
Ownership Clarity	Distinguishes candidate, team, and organization contributions	No
Business Impact	Outcome and metrics appropriate to seniority level	No
Technical/Strategic Depth	Round- and role-specific depth of insight	No
Clarity & Brevity	Understandability and appropriate duration	No
Delivery	Pace, pauses, filler indicators	Yes (audio)
Follow-up Handling	Responsiveness under challenge/clarification	Yes

Buttons & Actions:

Button	Behavior
View Feedback	See rubric scores and detailed feedback per dimension
View Transcript	Read the full session transcript with turn-by-turn breakdown
Correct Transcript	Fix transcription errors (preserves original with audit trail)
Challenge Evaluation	Request re-evaluation (triggers recalculation with audit)
Replay Audio	Listen to recorded responses
Copy Feedback	Copy feedback text for external use
Export Report	Download practice report (PDF)
Start New Session	Begin another practice round
Delete Audio Only	Remove audio recording but keep transcript and feedback
Delete Full Session	Remove all session content per retention policy

7. Practice Analytics (/dashboard/practice/analytics)

Longitudinal view of practice performance with session trends, competency trends, and readiness integration.

Dimensions Tracked:

Analytics	What It Shows
Session Trend	Rubric dimension scores across compatible sessions/versions
Competency Trend	Aggregated performance by competency with confidence and sample size
Question Readiness	Practised status, last score, next-practice recommendation per question
Answer Comparison	Approved answer intent vs. spoken response (without penalizing natural variation)
Time Use	Question/answer duration and session completion metrics

Buttons & Actions:

Button	Behavior
Time Range	Filter by date range
By Job/Kit	Filter analytics by job or kit
By Competency	Filter by specific competency
Export Data	Download analytics data

8. Outcome Learning (/dashboard/practice/learning)

AI-curated learning recommendations based on practice performance patterns, linked to readiness and NBA.

Buttons & Actions:

Button	Behavior
View Recommendation	Open detailed learning recommendation
Apply to Next Session	Focus next practice session on specific weak areas
Create Task	Add a preparation task based on the insight
Dismiss	Remove a recommendation

Privacy, Security & Fairness

• Explicit consent before any recording — consent version/timestamp stored per session
• No background recording — recording stops when you leave the session
• Audio is encrypted and owner-scoped — never exposed via stable public URLs
• User-controlled deletion — delete audio only (keep transcript/feedback), transcript+audio, or full session
• No facial analysis, emotion detection, accent penalty, personality diagnosis, or deception detection
• Delivery feedback focuses on intelligibility and user-selected goals, not accent conformity
• Adaptive questions exclude discriminatory, sexual, medical, or unrelated personal topics
• Stop/report available — users can flag inappropriate questions
• Practice data NOT visible to coaches/employers without explicit separate permission
• Text-only mode available when microphone is unavailable (discloses delivery scoring limitations)

Media Pipeline

1. Capture → Audio recorded in resumable bounded chunks in the browser
2. Upload → Encrypted chunked upload with retry and idempotency
3. Transcription → Provider transcription with configurable retention
4. Storage → Encrypted, owner-scoped with no stable public URLs
5. Deletion → Independently deletable: audio, transcript, or full session

Fallback Behavior

If adaptive generation fails during a session, the system falls back to a deterministic prioritized question sequence from the kit's question bank — so your practice session continues without interruption.

Missing Data Disclosure

When evaluation dimensions cannot be assessed:

Scenario	Disclosure
Text-only mode (no audio)	Delivery and follow-up handling dimensions marked "Not assessed — audio unavailable"
Short response	Some dimensions may have low confidence — explicitly labeled
Transcription failure	Turn marked "Transcription unavailable — not evaluated"

Related ES Requirements

ID	Requirement
P3-E01	Practice lifecycle: create, device check, consent, pause/resume/end, archive/delete
P3-MED-001	Explicit session consent before microphone access or recording
P3-MED-002	Device/permission check with troubleshooting
P3-MED-005	Users can delete audio while retaining feedback, or delete all session content
P3-MED-007	Text-only practice when microphone unavailable; disclose limited delivery scoring
P3-E03	Adaptive interviewer: plan, ask, listen, follow-up, challenge, conclude — multi-turn
P3-E04	Content and delivery rubric with evidence-linked feedback
P3-E05	Practice analytics: trend, competency, answer, and recommendation updates
AC-01	Users can create, run, pause, complete, archive and delete sessions safely
AC-02	Recording requires consent; audio/transcripts are secure and independently deletable
AC-03	Adaptive interviewer conducts bounded multi-turn role/round/persona simulations with fallback
AC-04	Evaluation is transcript-linked, rubric-versioned, transparent and excludes prohibited inference
AC-05	Transcript correction and re-evaluation work with audit history
AC-06	Practice analytics update readiness/recommendations using compatible samples

Related Docs

• [[interview-kits]] — Practice sessions use questions from your interview kits
• [[readiness]] — Practice performance replaces provisional readiness input with scored evidence (Phase 3)
• [[role-intelligence]] — Round hypotheses and competencies from RoleModel guide session plans
• [[debrief]] — Actual interview debriefs feed into outcome learning loop
• [[brief]] — Interview-Day Brief is separate and never activates recording
• [[nba]] — NBA generates practice recommendations for low-readiness competencies