What Is Explainable AI in Candidate Scoring? Why It Matters for Compliance

A candidate applies for a software engineering role. Your ATS assigns an AI match score of 72 out of 100. The recruiter glances at the number and moves on to the next candidate. But what does 72 actually mean? Which skills matched? Which didn't? Was the candidate penalised for career gaps, for attending a less prestigious university, or for writing their CV in a non-standard format? Without an explanation, the score is a black box — and a black box that influences hiring decisions is both a legal liability and a missed opportunity for better recruitment.

Explainable AI (XAI) is the practice of designing AI systems that produce outputs humans can understand, interpret, and verify. In recruitment, this means AI candidate scores that come with clear, readable explanations of why a particular score was assigned — what matched, what didn't, and what the AI considered most important.

What Is Explainable AI and Why Does Recruitment Need It

Explainable AI refers to AI systems whose internal decision-making process can be understood by humans. This stands in contrast to “opaque” or “black box” AI, where the model produces an output but the reasoning behind that output cannot be meaningfully explained — not even by the engineers who built it.

Recruitment needs explainable AI for three interconnected reasons:

• Legal compliance. GDPR Article 22 and the EU AI Act both create rights and obligations around explanation of AI-assisted decisions. Unexplainable AI scoring in recruitment creates direct legal exposure.
• Recruiter trust. Recruiters who do not understand AI scores cannot use them effectively. Research shows that recruiters either over-rely on unexplained scores (automation bias) or ignore them entirely. Neither outcome improves hiring quality.
• Candidate fairness. Candidates affected by AI scoring decisions have a legitimate interest in understanding how they were evaluated. Unexplainable AI makes it impossible to identify and challenge discriminatory scoring patterns.

Black Box vs Glass Box: Two Approaches to AI Scoring

AI scoring systems in recruitment generally fall into two architectural categories:

The practical implication for HR teams: if your ATS vendor cannot show you exactly why a candidate received a particular score, you are using black box AI. After August 2026, this creates significant compliance risk for any organisation operating in or hiring from the EU.

The Legal Case for Explainability: GDPR Article 22

GDPR Article 22 provides individuals with the right not to be subject to “a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her.” Recital 71 of the GDPR further specifies that data controllers must provide “meaningful information about the logic involved” in automated decision-making.

For AI candidate scoring, this creates a clear obligation: if a candidate's application is significantly affected by AI scoring (which includes being ranked low and therefore not reviewed by a human), the candidate has the right to understand how the scoring logic works and to request human intervention.

A black box AI system that cannot explain its scoring logic fails the “meaningful information about the logic involved” test. This is not a theoretical risk. Data protection authorities across Europe have increasingly focused on automated decision-making in employment contexts, and the combination of GDPR Article 22 with the EU AI Act's transparency requirements creates dual enforcement pressure.

EU AI Act Article 86: The Right to Explanation

The EU AI Act reinforces and extends GDPR's explanation requirements. Article 86 establishes that “any affected person subject to a decision which is taken by the deployer on the basis of the output from a high-risk AI system” has the right to “clear and meaningful explanations of the role of the AI system in the decision-making procedure and the main elements of the decision taken.”

For recruitment specifically, this means:

• Candidates must be able to request an explanation of how AI scoring influenced the recruitment decision
• The explanation must cover the “main elements” — not a generic description but the specific factors that determined the outcome in their case
• The explanation must be “clear and meaningful” — understandable to a non-technical person
• The deployer (employer) is responsible for providing this explanation, even if the AI system was built by a vendor

This obligation is distinct from the transparency requirements under Article 13 (which apply to how the system is designed and documented) and creates a direct, individual right for each candidate affected by AI-assisted decisions.

What a Good AI Score Explanation Looks Like

An effective AI score explanation in recruitment should include the following elements:

1. Overall score with context. The numerical score plus a plain-language interpretation (e.g., “Strong match — this candidate meets most key requirements for this role”).
2. Key strengths. The specific skills, qualifications, or experience factors that contributed positively to the score, with relative weighting.
3. Key gaps. The specific requirements the candidate does not meet, and how these gaps affected the score.
4. Factor-level breakdown. A visual or tabular breakdown showing how each evaluated dimension (skills, experience, qualifications, certifications) contributed to the overall score.
5. Confidence indicator. An indication of how confident the AI is in the score — particularly relevant when CV parsing quality is low or when the job description is vague.

Why Recruiters Need Explanations Too

Explainable AI is not just about candidate rights. Recruiters who receive AI scores without explanations face a fundamental trust problem: should they follow the AI's recommendation or their own judgment?

Research on human-AI collaboration in high-stakes decisions reveals two dysfunctional patterns:

• Automation bias. Without understanding why the AI scored a candidate high or low, recruiters default to trusting the AI over their own assessment. This is especially dangerous when the AI has a systematic bias — the recruiter amplifies rather than catches the error.
• Automation distrust. Recruiters who distrust unexplained AI scores ignore them entirely, eliminating any efficiency benefit. The organisation is paying for AI features that recruiters do not use.

Explainable AI resolves both patterns. When recruiters can see that an AI score is driven by specific, relevant factors, they can calibrate their trust appropriately — following the AI when its reasoning aligns with the role requirements and overriding it when they identify contextual factors the AI missed.

The Business Case for Transparent AI Scoring

Beyond compliance, transparent AI scoring delivers measurable business benefits:

• Higher recruiter adoption. Studies show that AI tools with explanations achieve 60 to 80% higher adoption rates among recruiters compared to black box alternatives.
• Better hiring decisions. When recruiters understand AI scores, they combine AI analysis with human judgment more effectively. A 2024 study by the Kellogg School of Management found that teams using explained AI scores made 23% more accurate hiring predictions than teams using unexplained scores.
• Reduced legal exposure. Explainable AI provides a documented, defensible basis for hiring decisions. When challenged, you can demonstrate exactly how the AI contributed to the decision and how human judgment was applied.
• Improved candidate experience. Candidates who receive meaningful feedback about their application assessment — even when rejected — rate the process more favourably and are more likely to reapply or recommend the employer to others.
• Easier bias detection. When you can see what factors drive scores, you can identify problematic patterns without needing specialised data science expertise.

How Treegarden Delivers Explainable AI Scoring

Treegarden's AI scoring system is designed with explainability as a core architectural principle, not an afterthought:

• Score breakdown on click. Every AI match score in the candidate list is clickable. One click opens a detailed modal showing the complete scoring explanation with strengths, weaknesses, and factor-level contributions.
• Plain-language explanations. Score explanations are written in natural language that HR professionals can understand without technical training. No jargon, no model weights, no feature vectors — just clear statements about what matched and what didn't.
• Audit-ready logging. Every AI score, explanation, and subsequent human decision is logged with timestamps and user identification. These logs provide the documentation basis required by GDPR Article 22 and the EU AI Act for demonstrating compliant AI deployment.
• Per-job AI configuration. Recruiters can adjust AI scoring criteria per job, ensuring that the AI evaluates candidates against role-specific requirements rather than generic criteria. This transparency extends to the configuration level — you know exactly what the AI is looking for.
• Human override with reason logging. When a recruiter disagrees with an AI score and overrides the recommendation, the system captures the override reason. This creates a feedback loop that improves AI accuracy over time and demonstrates human oversight for compliance purposes.

FAQ

Explainable AI in candidate scoring is not a luxury or a future aspiration — it is a current legal requirement under GDPR and an imminent one under the EU AI Act. Organisations using black box AI scoring in recruitment need to transition to explainable alternatives before August 2026. Treegarden provides explainable AI scoring as a core feature, with score breakdowns, audit logging, and human oversight designed with EU AI Act requirements in mind. Request a free demo to experience transparent AI scoring firsthand.