Phase 11: Technical Interview Type

Time estimate: ~2-4 weeks per domain
Impact: Technical interview capability
Key decisions already made by Dr. Rigos:

Decided

• Rubrics stay system-wide (not per-discipline)
• Practical scoring uses regular competency-based scoring
• Disciplines are organizational only (workspace type, difficulty), not scoring constructs
• Sequential candidate flow: conversational interview → practical assessment

What We Need

1. Technical Competence Definition

Define "technical competence" psychometrically. What construct(s) are we measuring?

• Domain knowledge (does the candidate know the material)?
• Problem-solving approach (how do they think through technical problems)?
• Communication skill (can they explain their reasoning)?

Is "technical competence" one construct or multiple? If multiple, should they be separate competency entries in the framework?

2. Scoring Rubrics

Design scoring rubrics that distinguish:

• Knowledge from communication — a brilliant engineer who explains poorly should score differently than an average engineer who explains brilliantly
• Depth from breadth — deep expertise in one area vs. surface knowledge of many
• Practical skill from verbal reasoning — doing vs. describing

3. Legal Boundary

Establish the "interview" vs. "test" boundary:

• Structured technical interviews may qualify as employment tests under EEOC Uniform Guidelines
• This triggers stricter validation requirements
• Where is the line for our platform?

4. Construct Contamination Review

Are we scoring technical knowledge or verbal fluency? Risk: candidates who are great communicators score higher even if their technical knowledge is equivalent.

5. Domain-Specific Considerations

For each initial domain (Software Engineering, Data Science, Finance):

• Are there domain-specific anchor patterns?
• Do difficulty levels (easy/medium/hard) map to anchor levels?
• How should the AI calibrate question difficulty?

6. External Platform Scoring (Tier 3)

If we integrate HackerRank/CodeSignal scores, is it psychometrically defensible to combine platform scores with interview scores? Or should they remain separate data points?