Testlify's AI Ethics and Compliance Framework

Overview

Testlify is committed to the highest standards of AI ethics in talent assessment. Our AI systems are built to be fair, transparent, and compliant with applicable laws — including New York City's Local Law No. 144 of 2021, which mandates bias audits for Automated Employment Decision Tools (AEDTs).

This article explains how Testlify designs, audits, and governs its AI systems to protect candidates and support responsible, equitable hiring decisions for employers.

1. Testlify's commitment to ethical AI

Testlify recognizes that AI plays a critical role in talent skill assessment and that the technology carries real consequences for candidates' careers and employers' decisions. Our approach to ethical AI is built on three pillars:

• Fairness: Ensuring assessments reflect candidates' skills, not personal characteristics.
• Transparency: Being open about how our AI works, what data it uses, and what its limitations are.
• Accountability: Maintaining a rigorous, ongoing audit and improvement cycle so our systems remain equitable over time.

In line with NYC Local Law No. 144 of 2021, Testlify regularly evaluates our AI tools for potential disparate impacts on different demographic groups, ensuring they do not inadvertently perpetuate biases in hiring or promotion processes.

2. How Testlify's AI system works

Testlify's AI system is the core of our talent assessment platform. It uses advanced algorithms to evaluate and score candidate abilities across a wide range of tests, helping employers identify the best fit for their roles. The system processes assessment response data to generate informed predictions about candidates' skills and potential job performance.

Key design principles behind the system include:

• Assessments are scored objectively based on demonstrated ability, not subjective impressions.
• Data inputs are carefully controlled to exclude characteristics unrelated to job performance.
• Results are designed to be consistent and reproducible across candidate cohorts.

3. Bias prevention safeguards

Testlify's AI is designed with built-in safeguards to ensure every candidate is assessed on a level playing field:

• Exclusion of sensitive information: The system excludes all personally identifiable information (PII) — including race, gender, and other protected attributes — from AI algorithms. Assessments are based solely on demonstrated skills and abilities, not personal characteristics.
• Feature normalization: The AI normalizes scores across questions of varying difficulty, ensuring all candidates are evaluated consistently regardless of the specific questions they encounter.
• Human-in-the-loop oversight: Final hiring decisions are made with human review. AI insights inform — but never replace — human judgment. This approach captures AI's efficiency while preserving the empathy, context, and nuance that only humans can provide.

Tip for employers: Testlify's AI scores are one data point in your hiring process. Your team's judgment and interviews remain essential to well-rounded, fair hiring decisions.

4. The Process Audit framework

Testlify follows a structured Process Audit framework — inspired by financial auditing practices — to rigorously evaluate and manage AI risk. This impartial methodology provides a repeatable, standardized process for AI system audits.

The audit proceeds in three phases:

1. Scoping: A comprehensive survey of Testlify's AI system to map its capabilities, inputs, outputs, and limitations. This phase establishes the full scope of what will be evaluated in subsequent stages.
2. Evaluation and verification: Testlify submits documentation demonstrating compliance with established audit criteria. Auditors then verify the evidence by:
   • Reviewing submitted documents and external references
   • Interviewing personnel involved in bias testing processes
   • Observing the analytical procedures used in bias testing
   • Inspecting samples of bias testing data
3. Certification: Once all audit criteria are satisfied, the auditor drafts a final report and formally certifies the AI system. Certification confirms Testlify's adherence to ethical AI standards.

Testlify currently conducts rigorous internal audits and is committed to an independent external audit to provide an objective third-party perspective on our practices.

5. Testlify's responsibilities

To ensure the audit process is effective and credible, Testlify upholds the following responsibilities:

• Document accuracy: All materials, evidence, and documentation submitted for audits are accurate, complete, and free from misrepresentation of any kind.
• Resource allocation: Testlify provides all necessary personnel, data access, and technical resources to support a thorough and efficient audit process.
• Openness and collaboration: Auditors receive full cooperation and unrestricted access to all relevant documents and materials required for a comprehensive evaluation.

6. Auditor responsibilities and independence

The auditor's primary role is to deliver an objective, impartial opinion on Testlify's bias testing practices. Auditor responsibilities include:

• Evaluating assertions: Reviewing Testlify's claims about bias testing to confirm they satisfy established audit criteria and reflect actual system behavior.
• Ensuring no material misstatement: Verifying that all information provided by Testlify is free from significant errors or misrepresentation — whether unintentional or otherwise.

Auditor independence is maintained in accordance with standards set by ForHumanity and Sarbanes-Oxley. These frameworks ensure the audit outcome is not influenced by external pressures or conflicts of interest, preserving the integrity of the certification.

7. Audit findings and overall opinion

The following findings are based on Testlify's internal audits. An independent external audit will provide further validation and external verification of these results.

1. Disparate impact quantification: Testing confirms no significant bias across demographic groups in relation to race or gender. Testlify conducted sufficient testing to assess the AI tool's impact across different demographic groups.
2. Governance: Testlify has established effective internal governance structures with clearly defined roles and responsibilities for overseeing AI ethics and compliance.
3. Risk assessment: Potential risks contributing to bias have been identified and evaluated at each stage of the AI lifecycle, incorporating input from multiple stakeholder perspectives.

Overall audit opinion: Testlify's AI system adheres to required standards for bias and fairness, passing all evaluated criteria. This reflects our ongoing commitment to ethical AI practices and continuous system improvement.

8. Dataset limitations

Testlify is transparent about the inherent challenges of AI bias testing. Results must be interpreted with the following limitations in mind:

• Inferred demographic labels: Datasets do not always include self-declared demographic labels. Where labels are inferred rather than self-reported, the accuracy of bias assessments may be limited.
• Data effectiveness: The effectiveness of bias testing depends on the volume and quality of available historical data. Limited data can constrain the scope of testing and the robustness of conclusions drawn.

Internal audit results are promising. However, Testlify acknowledges these limitations and remains committed to continuously improving our data collection and analysis methods to strengthen the accuracy and reliability of our bias testing over time.

9. Our commitment going forward

Testlify reaffirms its commitment to ethical AI and responsible technology. A successful audit is a milestone — not a destination. Looking ahead, Testlify will:

• Regularly update and refine AI algorithms to improve fairness and accuracy
• Conduct annual audits to ensure ongoing compliance and continuous improvement
• Monitor evolving AI ethics standards and legislation, adapting our practices accordingly
• Engage clients, candidates, and industry experts for feedback to continuously raise our ethical standards

We believe responsible AI is a shared effort. Testlify welcomes collaboration with all stakeholders to uphold and advance our commitment to fairness in hiring.

Need help? Contact support.