Evidence-Based Assessment and Selection

Introduction

Hiring is one of the highest-leverage decisions in a startup. A strong appointment can accelerate product development, customer learning, fundraising, and team capability; a poor appointment can consume scarce managerial time, damage morale, and create legal and reputational risk. Yet startup hiring is often informal. Founders may rely on intuition, “culture fit,” unstructured conversations, referrals, or confidence displayed during an interview. These approaches can feel efficient while producing inconsistent and biased decisions.

Evidence-based selection replaces impressionistic judgment with a systematic process. It identifies the capabilities required for a role, gathers job-relevant evidence through multiple assessment methods, evaluates candidates against predetermined criteria, and documents decisions. The aim is not to eliminate managerial judgment. Rather, it is to make judgment more disciplined, transparent, comparable, and defensible.

This chapter examines screening, work samples, structured interviews, validated tests, reference checks, scoring rubrics, validity, reliability, utility, adverse impact, bias control, candidate experience, and governance. The central principle is simple: assess what the job requires, use methods that provide dependable evidence, and make decisions consistently.

Learning Objectives

By the end of this chapter, readers should be able to:

1. Translate a job analysis into job-related selection criteria.
2. Design a structured selection process using multiple assessment methods.
3. Distinguish reliability, validity, utility, fairness, and adverse impact.
4. Construct scoring rubrics and structured interview guides.
5. Evaluate work samples, tests, references, and screening tools critically.
6. Identify and reduce common sources of bias in startup hiring.
7. Interpret selection data and improve a process over time.
8. Balance predictive quality, legal defensibility, cost, speed, and candidate experience.

1. Selection as an Evidence System

Selection is the process of gathering and interpreting information to predict future work performance and related outcomes. The outcome being predicted is the criterion, such as sales performance, software quality, customer retention, safety behavior, or effective leadership. The assessment method is the predictor, such as a work sample, interview score, cognitive-ability test, or structured reference rating.

A defensible process has four linked elements:

1. Job analysis: What tasks, responsibilities, knowledge, skills, abilities, and other characteristics does the role require?
2. Assessment design: Which methods can provide relevant evidence of those requirements?
3. Decision rules: How will evidence be scored, combined, and interpreted?
4. Evaluation and governance: Does the process predict performance, treat candidates fairly, and comply with applicable law and policy?

A startup should resist the temptation to copy assessment practices from large organizations without adaptation. A small venture may need a shorter process, but not an arbitrary one. A two-stage process can still be rigorous if its criteria, questions, scoring, and decision rules are clear.

Job-relatedness and competency models

Selection begins with a current job analysis. A useful analysis identifies:

• major outcomes expected in the first six to twelve months;
• recurring tasks and decisions;
• technical and interpersonal requirements;
• conditions under which work occurs;
• behaviors that distinguish effective from ineffective performance; and
• requirements that are genuinely necessary rather than merely customary.

Competencies should be expressed behaviorally. “Strategic thinker” is too vague to score reliably. “Can identify the three highest-impact customer segments using limited market evidence and explain the trade-offs” is more assessable.

A competency should not be included merely because a founder personally values it. Requirements such as attendance at a particular university, similarity to existing employees, willingness to work socially after hours, or “executive presence” may be weakly related to performance and can exclude qualified candidates.

2. A Structured Selection Architecture

A practical startup process may contain the following stages:

The stages should be proportional to the role. A low-risk junior role may require a résumé screen, a short work sample, and one structured interview. A senior financial, safety, or leadership role may justify more extensive validation, references, and background procedures.

Assessment methods can be combined through a multiple-hurdle model, in which a candidate must meet minimum standards at successive stages, or a compensatory model, in which strong performance on one predictor can offset weaker performance on another. Multiple hurdles are appropriate when a requirement is essential—for example, a legally required certification. Compensatory scoring is useful when strengths can reasonably offset weaknesses. The choice should be made before reviewing individual candidates, not after observing a preferred applicant.

3. Screening Without Over-Screening

Screening is useful for efficiently identifying applicants who meet essential requirements. It becomes problematic when it relies on vague impressions or proxies for social similarity.

Screening criteria

Criteria should be:

• directly connected to documented job requirements;
• observable from application materials or verifiable evidence;
• applied consistently to all candidates; and
• limited to requirements that are necessary at entry or can reasonably be learned.

A screening matrix might classify each criterion as “meets,” “does not meet,” or “unclear.” “Unclear” should not automatically mean rejection; it may trigger a clarification step.

Résumé and application review

Résumé review has limited reliability because résumés vary in formatting, self-presentation, and access to coaching. Reviewers should therefore focus on evidence of relevant accomplishments rather than prestige signals. Blind review—removing names, addresses, photographs, and other unnecessary identifiers—may reduce some forms of bias, although it cannot eliminate all bias because experiences themselves may reveal identity or opportunity differences.

Application questions should ask for job-relevant information. For example, a product-management application might ask candidates to describe a product decision made under uncertainty and the evidence used. The question should not demand unpaid work that is disproportionate to the role.

4. Work Samples and Simulations

Work samples ask candidates to perform tasks similar to those they will perform on the job. They often have strong face validity because candidates can see the connection between the exercise and the role. Examples include:

• debugging a small codebase;
• analyzing a customer-retention dataset;
• drafting a response to a dissatisfied customer;
• conducting a sales discovery conversation;
• prioritizing a product backlog;
• preparing a short investment or market memo; or
• facilitating a team decision.

A good work sample has a clear purpose, realistic constraints, standardized instructions, and an analytic rubric. It should be difficult enough to reveal differences in capability but not so broad that candidates are assessed on time, money, or support unavailable to others.

Designing a work sample

A five-step design process is useful:

1. Identify the competency or task to be assessed.
2. Select a representative, not unnecessarily elaborate, task.
3. Specify materials, time limits, permitted tools, and submission requirements.
4. Define observable dimensions of quality.
5. Pilot the task and check whether assessors interpret it consistently.

Work samples can create inequity when they require extensive unpaid labor, proprietary information, expensive software, or unpredictable scheduling. A startup should provide reasonable time limits, accessible formats, and accommodations. If artificial intelligence tools are permitted, the instructions should state whether the assessment evaluates unaided skill, tool-supported performance, or judgment in supervising AI output.

5. Structured Interviews

Unstructured interviews are common but weakly standardized. Different candidates may receive different questions, interviewers may overweight first impressions, and confident storytelling may be mistaken for competence. Structured interviews improve comparability by asking candidates the same core questions, using job-related prompts, and applying anchored scoring criteria.

Interview question types

Behavioral questions ask about past behavior:

• “Describe a time when you had to change a project direction after receiving new evidence. What did you do, and what happened?”

Situational questions present a hypothetical job problem:

• “A major customer requests a feature that conflicts with the product roadmap. How would you analyze and respond?”

Behavioral questions can provide evidence of actual experience, while situational questions are useful when candidates have limited experience or when future scenarios are central to the role. Both should be linked to competencies.

Structured interview practices

A structured interview should include:

• the same core questions for all candidates;
• probes that clarify the answer without leading the candidate;
• defined scoring dimensions;
• behavioral anchors for low, moderate, and high performance;
• independent scoring before panel discussion; and
• a documented record based on evidence rather than personality impressions.

Interviewers should avoid questions about protected or private matters unless a lawful, job-related process specifically requires them. Questions about family plans, age, disability, religion, ethnicity, health, or political affiliation are generally inappropriate and may create legal risk. Where physical or scheduling requirements are essential, ask whether the candidate can perform the specified job functions, with reasonable accommodation where applicable—not about a medical condition.

6. Tests: Reliability, Validity, and Appropriate Use

A test is not automatically valid because it is popular, technologically sophisticated, or branded as “scientific.” Its usefulness depends on the evidence supporting its interpretation and use for a particular job and population.

Reliability

Reliability concerns the consistency of measurement. A measure may be reliable across raters, across occasions, or across items. Reliability is necessary for validity: if scores are highly unstable, they cannot provide dependable evidence. Reliability does not, however, prove that the measure assesses the right construct.

For example, two interviewers may consistently agree that a candidate appears charismatic. Their agreement may be high, but charisma ratings may still have little relevance to the job. Conversely, a relevant assessment scored inconsistently by interviewers may be useful in principle but poorly implemented.

Validity

Validity is best understood as the degree to which evidence and theory support the interpretation and use of assessment scores for a stated purpose. Common sources of validity evidence include:

• Content-related evidence: The assessment reflects important job tasks or requirements.
• Criterion-related evidence: Scores are associated with relevant performance criteria.
• Construct-related evidence: The assessment behaves as expected for the construct it claims to measure.
• Consequential and fairness considerations: The use does not produce unjustified consequences or systematically disadvantage groups without sufficient job-related justification.

Validity is contextual. Evidence for a coding test used with experienced backend engineers may not automatically support its use for entry-level customer-support hiring.

A correlation between predictor and criterion is often summarized as r, where r ranges from −1 to +1. A positive correlation indicates that higher predictor scores tend to be associated with higher criterion scores; it does not establish causation. The squared correlation, r², is sometimes interpreted as the proportion of criterion variance statistically associated with the predictor in a simple linear model, but this interpretation requires caution. It does not mean that a predictor “explains” an identical percentage of every individual’s performance.

The standard error of measurement describes expected uncertainty in an observed score. In classical test theory:

where SEM is the standard error of measurement, SD is the standard deviation of observed scores, and rₓₓ is the reliability coefficient. This relationship assumes a classical test-theory model and a reliability coefficient appropriate to the score. It illustrates why a small score difference should not be treated as a precise ranking.

Criterion contamination and deficiency

A performance criterion can be contaminated by factors unrelated to the job, such as unequal territory, manager favoritism, or access to better tools. It can also be deficient when important aspects of performance are omitted. For instance, measuring a salesperson only by revenue may omit customer retention and ethical conduct. Assessment design should therefore consider the quality of both the predictor and the criterion.

7. Scoring Rubrics and Decision Rules

A rubric converts qualitative evidence into consistent judgments. It should identify dimensions, define performance anchors, and specify how scores are combined.

Suppose a startup assesses a product analyst using four dimensions:

If each dimension is scored from 1 to 5, the weighted score is:

where S is the overall score, wᵢ is the weight for dimension i expressed as a proportion whose weights sum to 1, and xᵢ is the candidate’s score on that dimension. The scale must be defined consistently, and weights should reflect job importance rather than interviewer preference.

For a candidate scoring 4, 5, 3, and 4 respectively, the score is:

The result is on the original 1-to-5 scale. A weighted average is not a substitute for judgment: a candidate may need to meet a minimum on an essential dimension even if the total score is high.

Rubrics should use behavioral anchors. For example, a score of 5 for analytical reasoning might mean: “Identifies the relevant evidence, distinguishes correlation from causation, states assumptions, tests alternatives, and reaches a proportionate conclusion.” A score of 2 might mean: “Uses isolated facts, overlooks key assumptions, and reaches a conclusion unsupported by the evidence.”

Interviewers should record evidence before assigning a rating. “I liked the candidate” is not evidence; “the candidate described testing two competing hypotheses and changed the recommendation after the results” is evidence.

8. Combining Evidence and Evaluating Utility

Selection decisions involve trade-offs among predictive quality, cost, time, speed, candidate burden, and legal risk. Utility is the practical value obtained from a selection method after considering these trade-offs.

A highly predictive assessment may have low utility if it is prohibitively expensive, takes weeks to administer, or causes strong candidates to withdraw. A modestly predictive method may have high utility when it is inexpensive, quick, job-relevant, and scalable. Utility should be evaluated at the process level rather than by asking whether one tool is universally “best.”

A simplified expected-value logic is:

This is a conceptual framework, not a universal accounting formula. “Value” may include productivity, retention, reduced error, customer outcomes, and avoided replacement costs. These quantities are often uncertain and should be represented as ranges rather than false precision.

Selection ratio also affects decision quality. If many applicants are available for a limited number of positions, the organization can be more selective; if the applicant pool is small, even a strong method cannot create evidence that does not exist. Selection should therefore be interpreted alongside applicant quality, labor-market conditions, and the consequences of a false positive or false negative.

9. Bias, Fairness, and Adverse Impact

Bias can enter selection through job requirements, sourcing, screening, interviewer behavior, assessment content, scoring, or final decision-making. Common mechanisms include similarity bias, confirmation bias, anchoring, halo and horns effects, affinity bias, overconfidence, and availability bias.

Bias-control practices

Effective controls include:

• structured job analysis;
• predefined essential criteria;
• standardized questions and work-sample instructions;
• diverse and trained assessors where feasible;
• independent scoring before discussion;
• evidence-based notes;
• removal of unnecessary personal information;
• calibrated scoring sessions;
• accommodations and accessible assessment formats; and
• periodic analysis of outcomes by demographic group where lawful and ethically appropriate.

Bias control is not achieved by assuming that a diverse panel is automatically objective. A diverse panel can still use an ambiguous process. Structure and accountability remain essential.

Adverse impact

Adverse impact refers to a substantially different selection rate for a protected group compared with a comparison group, even when the procedure is facially neutral. In U.S. employment practice, the four-fifths or 80 percent rule is commonly used as a screening indicator, not as a complete legal determination.

The selection-rate ratio is:

For example, if 40% of Group A applicants are selected and 24% of Group B applicants are selected, the ratio is 0.60. This may signal adverse impact because it is below 0.80. The result should prompt investigation of sample size, statistical uncertainty, job-relatedness, alternative methods, and applicable law. It does not by itself prove unlawful discrimination.

Organizations should consult qualified legal and HR professionals because requirements vary by jurisdiction. They should also avoid treating the 80 percent rule as a safe harbor: a ratio above 0.80 does not establish fairness, and a ratio below 0.80 does not automatically establish illegality.

10. Reference Checks and Background Information

Reference checks can verify employment facts and gather evidence about past behavior, but they are vulnerable to positivity bias, restricted information, inconsistent questions, and personal relationships. A structured reference check should:

• obtain candidate authorization where required;
• ask consistent, job-related questions;
• request specific examples rather than general praise;
• distinguish direct observation from hearsay;
• avoid seeking protected or irrelevant personal information; and
• interpret references as one source of evidence, not as an unquestionable verdict.

Useful questions include: “What type of work did the candidate perform directly?” “What were the candidate’s strongest contributions?” “What support or structure helped the candidate succeed?” and “What developmental area would be important for a manager to understand?”

Background checks must be accurate, proportionate, and legally compliant. A criminal record, credit history, or other information should not be used reflexively when it is unrelated to the role. Candidates should have a fair opportunity to respond to potentially disqualifying information where required.

11. Candidate Experience and Ethical Governance

Candidate experience is not merely a public-relations concern. It affects completion rates, employer reputation, offer acceptance, and the credibility of the organization’s values. A fair process should communicate stages, expected time, evaluation dimensions, contact points, and decision timelines. Rejections should be timely and respectful.

Ethically governed assessment includes:

• informed communication about what is being assessed;
• confidentiality and secure handling of applicant data;
• limited access to sensitive information;
• retention and deletion practices consistent with law and policy;
• reasonable accommodations;
• a process for correcting factual errors; and
• human review of consequential decisions.

Artificial intelligence can assist with scheduling, structured note organization, skills-taxonomy mapping, or generating consistent interview prompts. It should not be treated as an objective judge. Automated résumé ranking, emotion recognition, facial analysis, voice analysis, and inferred personality systems may encode historical inequities, rely on weak constructs, or lack transparency. Before using AI, an organization should document the purpose, data sources, validation evidence, human oversight, monitoring plan, and candidate notice obligations. A vendor’s claim that a tool is “bias-free” is not evidence of validity or fairness.

12. An Applied Case: Selecting a Founding Customer-Success Manager

The following is an illustrative case.

Background

Northstar, a 22-person software startup, needs its first dedicated customer-success manager. The founder initially proposes hiring someone “energetic, naturally persuasive, and a great culture fit.” Previous hires were made after informal conversations, and two employees who interviewed well struggled with customer follow-through.

Problem

The role requires onboarding customers, diagnosing adoption barriers, coordinating with product and engineering, documenting issues, and retaining trust when the product cannot immediately meet a request. The founder’s proposed criteria are broad and difficult to score. They may also favor extroversion over listening and disciplined follow-up.

Analysis

A short job analysis identifies five important competencies:

1. customer diagnosis and listening;
2. written and verbal communication;
3. prioritization and follow-through;
4. cross-functional collaboration; and
5. ethical judgment under commercial pressure.

The team decides that “energy” is not a selection criterion unless translated into observable behavior, such as sustained follow-up during a complex onboarding process.

Approach and solution

Northstar designs a four-stage process:

1. Application screen: evidence of customer-facing or problem-solving work, without requiring a particular degree.
2. Work sample: candidates review a short fictional customer account and prepare a 15-minute onboarding plan plus a written customer update.
3. Structured interview: all candidates answer six core questions, including a behavioral question about recovering a damaged customer relationship and a situational question about a feature request that cannot be delivered.
4. Structured reference check: two references are asked about follow-through, listening, judgment, and response to feedback.

The work sample is scored on diagnosis, prioritization, communication, and judgment. Two assessors score independently before discussing differences. The startup provides the same data, time limit, and permitted tools to each candidate. Candidates are told that the exercise evaluates reasoning and communication, not unpaid production work.

The decision rule requires a minimum score on ethical judgment and customer diagnosis, plus a satisfactory overall result. The founder’s opinion is considered, but it cannot override documented evidence without a written rationale.

Outcome

The process takes slightly longer than the old informal interview but reduces repeated interviews and clarifies the final decision. A candidate with less prestigious prior employment performs strongly on the work sample and provides specific examples of customer recovery. The startup hires that candidate. After six months, the manager shows strong onboarding completion and documentation behavior. The company does not claim that the process caused the outcome; instead, it reviews the evidence and continues monitoring performance.

Learning points

• Vague cultural language should be translated into observable behavior.
• Work samples can reveal relevant capability more directly than conversational confidence.
• Minimum standards protect essential requirements.
• Structured references are more informative than “Would you rehire this person?” alone.
• A selection process should be evaluated after hiring rather than assumed to be effective.

13. Monitoring and Continuous Improvement

Selection systems should be treated as organizational processes that require evaluation. Useful indicators include:

• stage-by-stage completion and withdrawal rates;
• time from application to decision;
• candidate experience feedback;
• inter-rater agreement;
• offer acceptance;
• early performance and retention;
• adverse-impact indicators where lawful and statistically meaningful; and
• relationships between assessment scores and later job outcomes.

Monitoring does not mean optimizing only for hires who remain. Turnover can reflect compensation, management, workload, or labor-market conditions rather than selection quality. Likewise, high early performance may reflect unusually favorable assignments. Evaluation should therefore use multiple indicators and avoid overclaiming causality.

A startup should periodically ask:

1. Are the criteria still connected to the current job?
2. Are assessors using the rubric consistently?
3. Are any stages duplicative or unnecessarily burdensome?
4. Do candidates from different groups experience materially different outcomes?
5. Does each assessment add information beyond earlier stages?
6. Are successful employees succeeding for reasons the process does not measure?

The process should be revised when evidence shows that a predictor is weak, redundant, inequitable, or no longer relevant.

Best-Practice Checklist

Before launching a selection process, the hiring team should be able to answer “yes” to most of the following:

• Have essential tasks and outcomes been documented?
• Is every assessment clearly job-related?
• Are the same core questions and instructions used for all candidates?
• Are scoring anchors written before candidates are evaluated?
• Are interviewers trained to record evidence rather than impressions?
• Are minimum requirements distinguished from preferred qualifications?
• Are candidate time, accessibility, privacy, and accommodation considered?
• Are decision rules established before reviewing results?
• Is there a plan to monitor fairness and predictive usefulness?
• Has the organization obtained appropriate legal and professional advice?

Chapter Summary

Evidence-based selection is a structured process for predicting job performance from relevant, reliable, and fairly interpreted evidence. It begins with job analysis and translates requirements into observable competencies. Screening should use necessary, job-related criteria rather than prestige or similarity signals. Work samples and simulations provide direct evidence of performance, while structured interviews improve comparability through common questions, anchored ratings, and evidence-based notes.

Reliability concerns consistency; validity concerns the justified interpretation and use of scores. Neither is guaranteed by technology or popularity. Rubrics improve decision consistency, but weighted scores should be combined with minimum standards for essential requirements. Utility requires balancing predictive value against cost, speed, candidate burden, and organizational consequences.

Bias control depends on structure, documentation, trained assessors, accessible methods, and ongoing monitoring. Adverse-impact analysis can identify potential group differences, but the four-fifths rule is only a screening indicator and not a complete legal test. Reference checks and background information should be relevant, authorized, consistent, and proportionate. Finally, candidate experience, privacy, transparency, and human accountability are central elements of ethical selection. Startups gain advantage not by making hiring informal, but by making it fast, disciplined, evidence-based, and continuously improvable.

Key Terms

Adverse impact: A substantially different selection outcome for groups produced by a facially neutral procedure.

Behavioral interview: An interview asking candidates to describe past behavior in relevant situations.

Candidate experience: The applicant’s perceptions and treatment throughout the selection process.

Competency: A job-relevant capability expressed through knowledge, skill, ability, or behavior.

Construct validity: Evidence that an assessment measures the construct it claims to measure.

Criterion: The work outcome or performance standard a selection method is intended to predict.

Criterion contamination: Irrelevant influences included in a performance measure.

Criterion deficiency: Important aspects of performance omitted from a performance measure.

Four-fifths rule: A commonly used adverse-impact screening rule comparing selection-rate ratios.

Job analysis: Systematic identification of a job’s tasks, requirements, and performance outcomes.

Multiple-hurdle model: A selection design requiring candidates to meet minimum standards at successive stages.

Predictor: An assessment measure used to predict a criterion.

Reliability: The consistency or dependability of a measurement.

Rubric: A scoring framework defining dimensions, performance levels, and evidence anchors.

Selection ratio: The proportion of applicants selected from an applicant pool.

Situational interview: An interview asking how a candidate would respond to a hypothetical work situation.

Structured interview: An interview using standardized questions, procedures, and scoring criteria.

Utility: The practical value of an assessment after considering benefits, costs, speed, burden, and risk.

Validity: The degree to which evidence supports an intended interpretation and use of assessment scores.

Work sample: A job-related task or simulation used to observe candidate performance.

Review Questions

Short Answer Questions

1. Why should selection begin with job analysis?
2. Distinguish reliability from validity.
3. What is one advantage of a structured interview over an unstructured interview?
4. Why should a work sample use standardized instructions and a scoring rubric?
5. What does the four-fifths rule indicate, and why is it not a complete legal determination?

Long Answer Questions

1. Explain how a startup could design a complete selection process for a technical or managerial role. Justify the assessment methods selected.
2. Discuss the relationship among job analysis, competencies, predictors, and criteria.
3. Explain how scoring rubrics improve selection decisions and identify their limitations.
4. Analyze the ethical and governance issues involved in using artificial intelligence in hiring.
5. Compare multiple-hurdle and compensatory selection models. Give an example of when each would be appropriate.

Analytical / Application Questions

1. A startup selects 30% of applicants from Group A and 18% from Group B. Calculate the selection-rate ratio and explain what the result should prompt the organization to investigate.
2. Design a four-dimension rubric for a founding sales manager. Assign weights, define observable evidence, and specify one essential minimum standard.
3. A founder prefers a candidate because the candidate is confident and attended a prestigious university, but the candidate performs poorly on a standardized work sample. Analyze the decision using validity, bias-control, and evidence-integration principles.

Multiple Choice Questions with Answers

Question 1

Which feature most directly distinguishes a structured interview from an unstructured interview?

A. It is conducted only by senior managers.

B. It uses common core questions and predefined scoring criteria.

C. It lasts longer than an unstructured interview.

D. It focuses exclusively on hypothetical situations.

Correct Answer: B. It uses common core questions and predefined scoring criteria.

Question 2

A selection method produces highly consistent scores but measures a characteristic unrelated to job performance. Which conclusion is most accurate?

A. It is reliable but may lack validity.

B. It is valid but lacks reliability.

C. It has both high reliability and high validity.

D. It cannot be evaluated scientifically.

Correct Answer: A. It is reliable but may lack validity.

Question 3

Which assessment is generally the most direct observation of a candidate’s ability to perform a job-related task?

A. A prestige-based résumé screen

B. An informal conversation

C. A work sample

D. A personality impression

Correct Answer: C. A work sample

Question 4

A rubric uses weights of 0.30, 0.25, 0.25, and 0.20. A candidate receives scores of 4, 5, 3, and 4 respectively on a five-point scale. What is the weighted score?

A. 3.25

B. 3.75

C. 3.95

D. 4.25

Correct Answer: C. 3.95

Question 5

Group A has a selection rate of 40%, while Group B has a selection rate of 24%. What is Group B’s selection-rate ratio relative to Group A?

A. 0.40

B. 0.60

C. 0.76

D. 1.67

Correct Answer: B. 0.60

Question 6

Which practice best reduces confirmation bias during assessment?

A. Allowing interviewers to change criteria for each candidate

B. Discussing candidates before each interviewer records a score

C. Recording independent evidence and ratings before panel discussion

D. Selecting the candidate who seems most enthusiastic

Correct Answer: C. Recording independent evidence and ratings before panel discussion

Question 7

When is a multiple-hurdle selection model most appropriate?

A. When every weakness should be offset by another strength

B. When candidates must meet essential minimum requirements at successive stages

C. When no job requirements can be defined

D. When the organization wants to avoid documenting decisions

Correct Answer: B. When candidates must meet essential minimum requirements at successive stages

Question 8

Which statement best describes validity?

A. The popularity of an assessment among employers

B. The consistency of scores across raters only

C. The evidence supporting an intended interpretation and use of scores

D. The speed with which an assessment can be administered

Correct Answer: C. The evidence supporting an intended interpretation and use of scores

Question 9

Which reference-check question is most job-related and evidence-oriented?

A. Did you personally like the candidate’s personality?

B. What political views did the candidate express at work?

C. What specific examples demonstrate the candidate’s follow-through?

D. Would the candidate fit in with our founders socially?

Correct Answer: C. What specific examples demonstrate the candidate’s follow-through?

Question 10

Why should a startup be cautious about using an AI résumé-ranking tool?

A. All automated tools are necessarily illegal.

B. AI tools cannot process written information.

C. The tool may encode historical bias and lack evidence for the intended use.

D. Human review is never needed when technology is used.

Correct Answer: C. The tool may encode historical bias and lack evidence for the intended use.