AI Disclosure
Last updated: March 22, 2025
ASG is committed to transparent, explainable, and bias-aware AI. This page describes every AI feature we use, how it works, and your rights regarding AI-generated decisions.
Our AI Principles
Explainability
Every AI output comes with a human-readable explanation. The Why Engine is core to our ATS and matching features.
Bias Mitigation
Our mock interview and matching models include bias detection layers that flag potentially biased outputs for review.
Human Oversight
AI is a tool to assist humans — not replace them. All consequential decisions (hiring, application review) are made by humans.
AI Features Inventory
| Feature | Model / Method | Transparency | Human Override? |
|---|---|---|---|
| ATS Resume Scoring | Keyword extraction + heuristic scoring | Full — Why Engine explains every score | Yes — candidates can appeal via support |
| AI Resume Builder | Claude (Anthropic) — claude-sonnet-4-6 | Output shown in full; not submitted without review | Yes — user edits before use |
| Mock Interview AI | Claude (Anthropic) — claude-sonnet-4-6 | Scores and feedback shown per answer | N/A — practice tool only |
| Candidate Matching | Semantic embeddings (pgvector) + scoring | Ranking shown with match percentage | Yes — employers make final decisions |
| Auto-Apply | Rules-based matching on user criteria | Full audit log accessible by user | Yes — user sets criteria; can revoke |
The Why Engine
The Why Engine is ASG's proprietary explainability layer for ATS scoring. Instead of returning a black-box score, it produces:
- A list of matched and missing keywords
- A narrative explanation of the score in plain English
- Specific, actionable suggestions for improvement
- Separate subscores for keyword match, format quality, and experience level
The Why Engine does not use personal demographic data (age, gender, ethnicity) in scoring.
Bias Disclosure — Mock Interviews
Our AI interview evaluator is trained to detect and flag potential bias signals in questions or evaluations. Specifically:
- Questions that may imply assumptions about age, family status, or nationality are flagged
- Evaluation scores are normalized across demographic proxies where detectable
- Bias flags are logged and reviewed by our AI safety team quarterly
Despite these measures, AI systems can exhibit unexpected biases. Mock interview results are for practice only and should not be treated as authoritative assessments.
EU AI Act Compliance
Under the EU AI Act, our ATS scoring and candidate matching features may qualify as “high-risk” AI systems when used in employment contexts. ASG complies with the EU AI Act requirements including:
- Human oversight requirements for automated hiring decisions
- Transparency requirements (this disclosure page)
- Right to explanation for significant AI-assisted decisions
- Data governance and accuracy obligations
Your Rights Regarding AI
- Right to Explanation: Request a detailed explanation of any AI-generated score or decision affecting you
- Right to Contest: Challenge AI-assisted decisions through our support team
- Right to Human Review: Request human review of any significant AI output
- Right to Opt-Out: Disable AI features and use manual tools only (limited functionality)
Contact ai-ethics@americansoftwaregroup.com to exercise these rights.