What Can a Career Assessment Tell You?

The counsellor who left the choice with the person

Young people came to Boston's Vocation Bureau with a question that still troubles people today: how should they choose the work of their lives? Frank Parsons called it the choice of a “life-work”. When his colleagues published Choosing a Vocation in 1909, they preserved the method he had developed at the Bureau.

That method is often remembered as a matching exercise: understand the person, understand the occupation, then bring the two together. Parsons's account was richer. A wise choice required three kinds of inquiry. The first concerned the person's aptitudes, abilities, interests and ambitions, along with their resources and limitations. The second concerned the conditions, rewards, opportunities and prospects in different lines of work. The third was the reasoning needed to connect what had been learned about the person with what had been learned about the work.

Leave out any one of the three and the advice becomes thin. Self-knowledge alone is a portrait with nowhere to go. Occupational information alone is little more than a catalogue. Even when both are present, someone still has to weigh the relationship between them.

Parsons also drew a limit around the counsellor's role. The Bureau would not choose an occupation for an applicant. It would help the person investigate the question carefully enough to reach a conclusion of their own.

A modern career report can appear in seconds, complete with scores, types, occupational lists and a precise-looking match percentage. The technology has changed. Parsons's question has not: what has the assessment really learned, and how much should that evidence be allowed to decide?

The short answer

A career assessment can help describe a response or performance, support a carefully bounded prediction, point towards a developmental question, or contribute evidence to a decision. How much confidence any of those conclusions deserves depends on what was observed, how it was scored, who the evidence came from, the setting in which it was gathered and the decision someone hopes to make.

There is no direct reading of a career hidden inside a person. An interest profile, personality type, reasoning score, values exercise, achievement test or work sample cannot, by itself, establish which work someone will enjoy, enter, perform well, sustain or find worthwhile.

The term “career assessment” covers far more than a test. It may refer to a standardised inventory, a structured interview, a card sort, a work sample, a portfolio, observation, a skills audit, a guided story or a comparison with occupational information. The International Test Commission focuses on what the process is being used to infer. When a procedure samples behaviour and supports a serious conclusion about a person, evidence and competent use matter whatever label appears on the cover.

Before drawing a career conclusion, it also helps to be clear about what has been assessed. The companion construct map separates personality, interests, values, abilities, aptitudes and motivation, which are related but different sources of evidence.

One quick way to test a report is to look at its main verb:

The farther the report moves from description towards prediction or decision, the more evidence and care it needs. A responsible report makes that distance visible.

Career assessment · Inference distance

How far can this evidence travel?

An observation can support a bounded description before it can support prediction, diagnosis, development or a consequential decision. Open each claim to see what must be added.

Observed evidence 1

Interest response

A person reports liking outdoor investigation.

Boundary: It does not establish present skill, future learning, access or satisfaction.

Observed evidence 2

Timed reasoning task

A person solves spatial problems under a defined time limit.

Boundary: It does not establish interest, work performance or learning under support.

Observed evidence 3

Short work sample

A person completes a specified task under observed conditions.

Boundary: It does not represent every task, context, colleague or future opportunity.

1 · Claim distanceDescribeShow detailsHide details

Ask: What response or performance pattern was observed?

Add evidence: Clear scoring, representative observations and an honest account of conditions.

Safe next question: What does this result summarise, and what does it leave out?

2 · Claim distancePredictShow detailsHide details

Ask: Which later outcome, population and setting are proposed?

Add evidence: Relevant follow-up outcomes, uncertainty, comparable populations and transport evidence.

Safe next question: How often, for whom and under which conditions has this relationship held?

3 · Claim distanceDiagnoseShow detailsHide details

Ask: Is a defined condition or source of difficulty being distinguished?

Add evidence: Specialist competence, appropriate differential evidence and a defensible diagnostic framework.

Safe next question: Is diagnosis authorised here, or is further qualified assessment required?

4 · Claim distanceDevelopShow detailsHide details

Ask: What learning, reflection or experiment follows?

Add evidence: A developmental aim, opportunity to practise, feedback and evidence of change.

Safe next question: What small experience could test or develop this possibility?

5 · Claim distanceSupport a decisionShow detailsHide details

Ask: What action is proposed, with what consequences?

Add evidence: Alternatives, opportunity, fairness, error costs, human judgement, monitoring and recourse.

Safe next question: What other evidence and whose judgement should enter before anyone acts?

Exploration can begin with partial evidence. Placement, selection and exclusion demand a longer argument, stronger safeguards and a reviewable human decision.

Three illustrative evidence cards show an interest response, a timed reasoning task and a short work sample, each with a boundary on what it cannot establish. Five disclosure panels then move from description to prediction, diagnosis, development and decision support. Each panel states the question, added evidence and a safe next question. The further a claim travels, the more evidence it needs and the greater its possible consequences; no score, match percentage or verdict is produced.

Description begins with what happened

Every assessment begins with something a person does. They endorse an activity, rank a work value, solve a spatial problem, describe a turning point, complete a task or supply evidence from previous work. A scoring process then summarises that event, compares it with a reference group or places it within a model.

The safest first sentence stays close to what happened: “On these items, today, you reported this pattern.” Or: “Under these instructions and time limits, you completed this set of tasks.” In an interview, the equivalent might be: “These themes recurred in the account you gave.”

Starting there prevents a score from quietly becoming a description of the whole person. It also keeps today's evidence separate from tomorrow's decision. The psychometrics page follows that chain in detail. Reliability asks which kinds of repetition the result should survive, while validity asks which interpretations and uses the evidence can support.

Interest inventories provide a useful example. The current Standards for Educational and Psychological Testing describes them as measures of preferences, including likes and dislikes for activities, subjects, occupations and kinds of people. The US Department of Labor's O*NET Interest Profiler uses John Holland's six-domain RIASEC model for educational planning, career exploration and guidance.

An interest result can bring an otherwise scattered pattern into view. It might connect someone's preference for repairing, investigating, creating, helping, persuading or organising with occupational information arranged in the same way. Competence, access and future satisfaction remain open questions. Preferences have histories too: exposure, opportunity, confidence and knowledge all affect which activities a person has learned to enjoy or can imagine doing.

Interests can remain recognisable for a long time. A longitudinal meta-analysis by Douglas Low and colleagues found substantial stability from adolescence into middle adulthood, generally increasing through early adulthood and varying by interval and form of interest. That continuity is one reason an interest assessment can be useful. The movement that remains is a reason to resist treating it as a permanent instruction.

What Strong's interest blank could see

Edward Kellogg Strong brought empirical comparison into vocational assessment in 1927. His Vocational Interest Blank asked about occupations, amusements, school subjects, activities and personal characteristics. The directions, preserved by the Smithsonian National Museum of American History, made clear that it was not a test of intelligence or school learning.

Strong compared each respondent's interests with those reported by successful men in named professions. This gave counsellors an intriguing new question to ask: how closely did one person's pattern resemble the pattern found among people already doing a particular kind of work?

The comparison group placed a hard limit on the answer. “Successful men” were a selected group drawn from a labour market that had already restricted entry to many professions. A resemblance to their interests could not tell another person whether they would gain admission, perform well, enjoy the work or remain in it. People who had been excluded from those professions might have interests that the scale barely recognised.

Strong had turned vague impressions into observations that could be inspected and compared. That was a real advance. What the blank measured, however, was resemblance to a particular historical group. The questions, occupational categories and entry barriers of that world travelled with the score.

Prediction needs an outcome

A career prediction needs a named outcome. Course completion, speed of learning, later performance, persistence, satisfaction, entry and income are separate things. Each must be observed over some period, among a particular group of people and under conditions that may help produce the result.

Interest measures do have predictive relationships, although exploration remains their main use. Christopher Nye, Rong Su, James Rounds and Fritz Drasgow reviewed more than sixty years of research and found that vocational interests and interest–environment congruence relate to performance and persistence. This makes interests relevant to engagement. It does not make a high interest score a promise of later performance, because skill, teaching, working conditions, opportunity and selection all intervene.

Ability and aptitude measures start with performance on tasks. An aptitude interpretation then uses that present evidence to say something about learning or later performance in a specified domain. As the dedicated aptitude page explains, potential is inferred rather than observed directly. A useful prediction says what kind of instruction or performance it concerns and over what period.

Achievement tests face a different question: what has already been learned? They can support course placement when their content and the institution's policy fit that purpose. The College Board describes ACCUPLACER, for example, as a current-skills assessment for placement, and recommends local evidence alongside multiple factors. None of this amounts to a judgement about a person's general potential.

Work samples come closer to the activity itself. A person might draft a response to a client, diagnose a fault, edit a spreadsheet or assemble a component. Such a task can reveal something a preference inventory cannot, but a twenty-minute sample is still only a sample. It may omit sustained responsibility, teamwork, fatigue, unfamiliar cases, tools, supervision and the learning that follows entry.

For any predictor, the full claim should sound something like this:

This evidence has this relationship with this outcome, for people like those studied, under these conditions and over this period.

“Future success” is too vague to fill any of those blanks. Prediction is a relationship between measured evidence and a measured outcome, and limitations at either end affect the result. Even a well-established correlation cannot tell us by itself why the relationship occurred or what different training and opportunity might change.

Diagnosis is a restricted word

Career reports sometimes borrow the tone of diagnosis. A type is “revealed”, a hidden blocker is “identified”, or an algorithm announces why someone is stuck. Clinical language can make a tentative interpretation sound like a condition discovered inside the person.

Clinical and differential diagnosis is a specialised use. The Testing Standards say that when a professional needs to distinguish among diagnostic groups, the chosen evidence should support that distinction. The 2024 NCDA Code of Ethics requires appropriate training and care when diagnosis is warranted. It also applies assessment standards to both quantitative and qualitative methods.

Medical-sounding prose does not give an ordinary career category diagnostic force. A Holland code organises vocational interests and environments. “Low confidence about mathematics” may be an accurate summary of a self-report, while “difficulty choosing” may identify a problem worth discussing. Neither phrase establishes a disorder or explains the whole decision.

An assessment can identify useful gaps without crossing that line. A work sample may show that a procedure needs practice. A structured conversation might uncover missing occupational information, a financial threshold or a fear worth exploring. A current-skills assessment may indicate that someone would benefit from support before beginning a course. These findings can guide development and, when appropriate, referral.

A report becomes useful through what happens next

It is easy to treat an assessment as the end of a transaction: complete the questions, receive a result, leave with a list. Research on career interventions suggests that much of the value lies in what follows.

Susan Whiston and colleagues combined 57 studies involving 7,364 participants. Across several career-choice outcomes, people who received an intervention scored about a third of a standard deviation above comparison groups on average. The size of the effect varied. Career decision-making self-efficacy showed a larger average effect than vocational identity, and practitioner support was associated with larger effects.

These were studies of interventions, not proof that any inventory will help. Steven Brown, Nancy Ryan Krane and colleagues identified possible ingredients such as written exercises, individual feedback, information about work, modelling and support for the person's decision. An assessment can contribute to that process, but it seldom constitutes the whole of it.

Development begins when a result changes what someone understands or does. It may give a person better words for their preferences, reveal a conflict, bring an unfamiliar occupation into view or show which assumption needs testing next. The next step might be a conversation, a short course, an observation day, a project, a work sample or an application. Reading the report is only the beginning.

Stories can add evidence that a scale misses. Someone may explain the conditions under which they enjoyed teaching, when a once-valued occupation became unsustainable, or how family and migration shaped the use of a qualification. Mary McMahon and Mark Watson have shown how narrative inquiry and score interpretation can be combined. A later review of qualitative career assessment found promising ideas alongside fragmented definitions and a limited empirical base. A story can deepen an interpretation, though the feeling of coherence is not evidence enough on its own.

The current Australian professional standards make this a collaborative task. Practitioners are expected to review results with clients, understand validity, reliability and the relevance of norm groups, and take account of consent, culture, language, accessibility and digital delivery.

The person being assessed can supply context that the score does not contain. They can reject a poor description, correct an assumption and explain what a proposed option would demand of their life.

The institution changes the meaning

Put the same set of questions into another institution and its meaning can change. A result used to start a conversation is doing different work from one used to place, classify or reject someone.

Four settings show how the burden grows:

The ASVAB makes these differences visible. Parts of the assessment family contribute to US military eligibility and occupational classification. A school-based Career Exploration Program also uses results for career planning. Although some of the battery is shared, exploration, eligibility and assignment involve different claims and consequences.

The US Department of Labor's Work Importance Locator carried an unusually plain warning. It invited people to rank work values and explore occupations, while telling employers and education programmes not to use the results for job or training screening. O*NET retired the tool in June 2024, but the warning still shows what intended-use discipline looks like. Evidence that helps reflection may be wholly inadequate for gatekeeping.

Employment selection therefore needs its own validation programme. The SIOP principles require a connection to the work, relevant criteria, evidence for the procedure and attention to the local setting. An exploratory inventory does not gain that authority simply because an employer replaces a “suggest” button with a “reject” button.

Institutions shape the available options as well as the use of the assessment. A result may suggest compatibility while tuition, care, licensing, location, disability access, migration status, discrimination or weak labour demand blocks the route. Social Cognitive Career Theory places supports and barriers within the process that connects interests, confidence and goals. Psychology of Working Theory treats economic constraints and marginalisation as central to access to decent work.

If an institution makes a path unavailable, calling it a poor personal match puts the barrier in the wrong place.

A recommendation adds another assessment

A polished digital report can look more complete than the evidence beneath it. Colour, ranking and numerical precision create an impression that the difficult reasoning has already been done. In practice, automated interpretation adds several new layers.

Suppose an interest inventory produces six scores. Before it can recommend occupations, a system must:

1. choose an occupational database and its version;
2. decide which person and occupation variables to compare;
3. set weights and thresholds;
4. handle missing or conflicting evidence;
5. rank options;
6. turn ranks into prose; and
7. imply an outcome such as satisfaction, persistence or success.

Each choice brings its own assumptions. Evidence for the inventory cannot validate the occupation data, matching rule or outcome language. A “92 per cent match” means little unless the system explains what went into the calculation, what was left out and why the number should matter for a real outcome.

The Testing Standards require computer-generated interpretations to disclose their bases and limitations and to avoid implying an empirical link between a result, a prescribed intervention and an outcome when that evidence does not exist for a relevant population. The 2025 ITC/ATP technology-based assessment guidelines add interface, device, accessibility, automated scoring, security, privacy and system-version concerns.

AI can make the explanation more fluent, but it does not shorten the evidence chain. A language model is perfectly capable of producing persuasive prose from weak inputs, outdated occupational data or a rule that the reader cannot inspect. Research on algorithmic hiring shows how hard it can be to evaluate vendor claims about development, validation and bias mitigation from public disclosures alone.

The difference is visible from the reader's side. A transparent recommendation leaves its reasoning, inputs, priorities, alternatives and uncertainty open to view. An opaque one offers polished language while concealing the choices that produced it. Behind either interface, the intended use and later outcomes still have to be understood over time. The NIST AI Risk Management Framework treats these as continuing governance responsibilities, not just interface choices.

Whether a particular career platform is valid, reliable, properly normed, fair, suitable for an age group or predictively accurate cannot be settled by a general article. Those are claims about a specific product and version, and they need their own evidence.

What sits behind an assessment conclusion

A career-assessment conclusion can look like a single answer, but it is the end of a long chain. It begins with something a person said, chose, ranked or did. That observation might come from self-report, a performance task, a work sample, an interview or imported data. A scoring method then turns it into evidence about an interest, value, characteristic pattern, present ability, learned achievement, aptitude, confidence, skill or work behaviour.

The result also contains a comparison, whether or not the report makes it obvious. A person may be compared with a norm group, a standard, an occupational profile, their own other scores or a model-generated category. The decimals can hide how much the answer might move with different items, another day, a different rater or another device.

The most consequential step is often a verb. A report may describe, predict, diagnose, develop or recommend. Prediction needs a named outcome and a relevant time and population. Recommendation adds prerequisites, cost, location, care, accessibility, discrimination, labour demand and alternative routes. None of those facts lives inside a personality or interest score.

That is why a career assessment is best understood as a beginning rather than a verdict. Its conclusion can be compared with a conversation, a short project, a course unit, a work sample, a visit or a new experience. The point is not to audit a report like a specialist. It is to see that the neat answer on the page is an argument about a changing person in a changing world.

What a career assessment can tell you

That a result can describe something real without describing the whole person. Preferences, performances and stories all offer evidence, but their meaning depends on how they were gathered and what they are being asked to represent.

That prediction needs a named future. Learning, performance, persistence, satisfaction and entry are different outcomes, produced under different conditions.

That development happens after the score. Feedback becomes useful through inquiry, practice, new information and better-structured choices.

That the institution supplies part of the meaning. The same evidence might begin a conversation, place a student, classify a recruit or deny an applicant. Higher stakes demand stronger evidence and better safeguards.

That the decision still belongs to the person. An assessment can organise a difficult question. It cannot live with the choice that follows.

Parsons's Bureau relied on paper forms, interviews and occupational reports. Its tools would look slow beside a modern platform, yet the boundary it drew still feels fresh. The counsellor offered information, careful reasoning and help, then left the choice with the applicant.

That is a worthy ambition for career assessment now: clearer evidence, reasoning that can be inspected, and a next step the person can test for themselves.

---

Sources

• Parsons, Frank. Choosing a Vocation. Houghton Mifflin, 1909.
• Smithsonian National Museum of American History. Strong Vocational Interest Blank for Men. 1927 object and directions.
• Holland, John L. “A Theory of Vocational Choice”. Journal of Counseling Psychology 6 (1959): 35–45.
• American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing. 2014.
• Career Industry Council of Australia. Professional Standards for Australian Career Development Practitioners. 5th ed., 2026.
• National Career Development Association. Code of Ethics. 2024.
• International Test Commission. Guidelines on Test Use.
• International Test Commission and Association of Test Publishers. Guidelines for Technology-Based Assessment. Version 1.1, 2025.
• Society for Industrial and Organizational Psychology. Principles for the Validation and Use of Personnel Selection Procedures. 5th ed., 2018.
• National Academies of Sciences, Engineering, and Medicine. “Overview of Psychological Testing”. 2015.
• National Center for O*NET Development. ONET Interest Profiler Manual*. 2021.
• National Center for O*NET Development. Work Importance Locator Archived Materials. 2024.
• US Department of Defense. ASVAB Technical Bulletin No. 4.
• US Department of Defense. ASVAB Career Exploration Program.
• College Board. ACCUPLACER Overview.
• College Board. How Multiple Factors Improve Placement Decisions.
• Nye, Christopher D., Rong Su, James Rounds and Fritz Drasgow. “Vocational Interests and Performance”. Perspectives on Psychological Science 7 (2012): 384–403.
• Low, K. S. Douglas, Mijung Yoon, Brent W. Roberts and James Rounds. “The Stability of Vocational Interests from Early Adolescence to Middle Adulthood”. Psychological Bulletin 131 (2005): 713–737.
• Kristof-Brown, Amy L., Ryan D. Zimmerman and Erin C. Johnson. “Consequences of Individuals' Fit at Work”. Personnel Psychology 58 (2005): 281–342.
• Whiston, Susan C., Yue Li, Nancy Goodrich Mitts and Lauren Wright. “Effectiveness of Career Choice Interventions”. Journal of Vocational Behavior 100 (2017): 175–184.
• Brown, Steven D., Nancy E. Ryan Krane and colleagues. “Critical Ingredients of Career Choice Interventions”. Journal of Vocational Behavior 62 (2003): 411–428.
• McMahon, Mary, Mark Watson and Wendy Patton. “Qualitative Career Assessment: A Review and Reconsideration”. Journal of Vocational Behavior 110 (2019): 420–432.
• McMahon, Mary, and Mark Watson. “Telling Stories of Career Assessment”. Journal of Career Assessment 20 (2012): 440–451.
• Lent, Robert W., Steven D. Brown and Gail Hackett. “Toward a Unifying Social Cognitive Theory of Career and Academic Interest, Choice, and Performance”. Journal of Vocational Behavior 45 (1994): 79–122.
• Duffy, Ryan D., David L. Blustein, Matthew A. Diemer and Kelsey L. Autin. “The Psychology of Working Theory”. Journal of Counseling Psychology 63 (2016): 127–148.
• National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework 1.0. 2023.