Self-Report vs Observer Ratings in Big Five Personality Assessment

7 days ago

21 min read

Author: Ukrainian Psychological Hub · Published: October 9, 2026 · Editorial Policy

Self-report and observer ratings are two established ways of assessing the Big Five personality traits. In a self-report, a person describes their own characteristic thoughts, feelings, and behavior. In an observer report, someone who knows that person—or, in some research designs, a stranger—rates the same person's tendencies. Both methods can measure the same broad personality constructs, yet they draw on different information and are affected by different sources of error.

The central finding is that the two perspectives overlap substantially, especially when the observer knows the person well, but they are not interchangeable. Whether one perspective is more informative depends on the particular trait, what can be observed, the observer's relationship with the person, the instrument, and the criterion used to evaluate accuracy. A 2025 scientific review by Ashton and Lee summarizes evidence that close acquaintances often provide ratings that converge with self-reports, while cooperation-related traits such as Agreeableness can show weaker agreement.

For readers new to personality measurement, start with our Big Five personality traits guide and the Big Five personality test and interpretation guide. This article focuses specifically on source of information: how self- and observer-reports compare, what agreement means, why results diverge, and how to combine reports responsibly.

What is the difference between self-report and observer ratings?

A Big Five self-report questionnaire asks a respondent to rate descriptions of their own typical behavior or dispositions. A statement about being talkative, dependable, considerate, emotionally reactive, or interested in new ideas is answered from the person's perspective. The resulting responses are combined into scales defined by that particular questionnaire. The relevant time frame is usually a person's general tendencies, although instructions and instruments differ.

An observer-report questionnaire asks another respondent to describe the target person. The observer might be a spouse, close friend, family member, coworker, roommate, teacher, or trained research rater. An observer format may reword an item from first person to third person, but wording changes alone do not establish that the two forms are psychometrically equivalent. The observer must understand the target, instructions, time frame, and scale anchors.

The Big Five comprise broad dimensions commonly labeled Extraversion, Agreeableness, Conscientiousness, Neuroticism, and Openness to Experience. These dimensions organize variations in personality; they do not sort people into fixed types. Different instruments operationalize their content differently. A useful overview of the hierarchy appears in our Big Five domains, aspects, and facets article.

What each rating source can—and cannot—observe

A person ordinarily has access to internal experiences that no observer can directly witness: habitual worry, private emotional reactions, the reasons for a decision, internal curiosity, or intentions that were never translated into action. This access is valuable, but introspection is also selective. A person may overlook recurring interpersonal effects of their behavior, remember unusual episodes more vividly than typical ones, or interpret an item through an idealized self-image.

An observer sees actions, interaction patterns, and their consequences in contexts shared with the target. A colleague may notice punctuality and follow-through, while a close friend may know whether someone initiates conversations and follows up on commitments outside work. Observers infer motives and feelings rather than directly access them. They can also mistake role requirements, stress, unfamiliarity, or social expectations for enduring personality.

Observers therefore differ in what they have sampled. A roommate's knowledge of domestic organization and a supervisor's knowledge of work habits are partly nonoverlapping. A strong design asks which observer had an opportunity to see behaviors that express the trait being rated. Frequency of contact and psychological closeness both matter, and neither guarantees unbiased judgment.

Self–other agreement, observer consensus, and accuracy are three different questions

Self–other agreement asks whether people who score relatively higher on a given self-report scale also tend to score relatively higher when assessed by an observer. Researchers commonly quantify this with a correlation across a sample. It measures convergence of two sources of ranking, not literal equality of each person's numerical score.

Observer consensus asks how consistently two or more observers describe the same set of targets. Two friends may agree with each other while disagreeing with a target's self-description. Accuracy is a stronger, criterion-dependent claim: how well a judgment captures the attribute or predicts independently observed behavior. A study must specify what counts as an external criterion and whether it is independent of the ratings.

These distinctions were articulated in classic work on personality judgment by Funder and West. Correlations, means, agreement for one individual, consistency between observers, and prediction of an outcome answer separate methodological questions. Conflating them is a major source of overconfident interpretations of personality test discrepancies.

How strongly do Big Five self- and observer-ratings agree?

A meta-analysis by Connolly, Kavanagh, and Viswesvaran estimated correlations between self- and observer-ratings for the Big Five, correcting its estimates for specified reliability components. The resulting corrected means were .62 for Extraversion, .59 for Openness to Experience, .56 for Conscientiousness, .51 for Emotional Stability, and .46 for Agreeableness. These are meta-analytic, correction-adjusted associations across studies, not the expected raw correlation for any particular two questionnaires, population, or pair of raters.

The figures show both meaningful convergence and substantial nonshared information. A correlation of .62 does not mean that two raters agree on 62% of items, that an observer is 62% accurate, or that 62% of one person's personality has been measured correctly. Scale reliability, range restriction, sample composition, observer familiarity, and the particular subtraits represented all influence the observed association.

In a broader meta-analytic integration, Connelly and Ones synthesized evidence concerning 44,178 targets in 263 independent samples, examining observer consensus, self–other correlations, and prediction of behavior. Their findings highlighted the value of interpersonal intimacy for improving observer accuracy, particularly for traits that are harder to see. This supports gathering well-informed perspectives when the goal warrants the additional burden.

A 2025 review finds that with close acquaintances and longer Big Five or HEXACO inventories, mean levels are generally similar across self- and observer-reports, while agreement varies by trait. The review also cautions through its discussion of assumed similarity and cooperation-related traits that convergence cannot be interpreted apart from the social relationship and measurement context.

Does one group generally rate itself more positively than others do?

Research does not support a simple rule that self-reports are always more favorable than informant reports. In a meta-analysis of 152 samples and 33,033 people, Kim, Di Domenico, and Connelly found little average difference between self- and informant-report means on the Big Five. Their average standardized mean discrepancy was approximately −.038. Moderate mean differences appeared when self-ratings were compared with reports by strangers, underscoring how unfamiliarity changes the comparison.

An average difference near zero does not imply that every participant receives identical scores from every observer. It also does not rule out self-enhancement in particular populations or domains. Opposing individual discrepancies can cancel one another in a group mean. Mean-level agreement and rank-order convergence must be evaluated separately.

Population averages should never be used as individual verdicts. Even if a sample shows that self and close-informant scores are generally similar, a particular disagreement requires examination of the items, the observers' evidence, the situations involved, and possible measurement error.

How agreement differs across the five domains

Extraversion frequently produces relatively strong self–other agreement because several of its common expressions—sociability, conversational participation, and outward assertiveness—are visible in interpersonal settings. That pattern does not make every facet equally visible. A person may experience considerable social effort or private reserve while behaving outgoingly at work. The Connolly and colleagues' synthesis places Extraversion highest among its corrected domain-level correlations.

Conscientiousness is often observable through deadlines, organization, follow-through, and dependability, but judgments depend strongly on setting. A coworker who sees exceptionally reliable project management may never see the target's household routines. It is useful to distinguish a broad scale from specific habits and to avoid interpreting one informant's limited observations as a complete profile.

Agreeableness poses particular challenges because evaluations of kindness, trust, cooperation, respect, and conflict behavior are partly relational and value-laden. An observer may judge the way a person treats them rather than the person's cross-situational tendencies. John and Robins found the lowest agreement for Agreeableness among Big Five domains in their samples, and Ashton and Lee's review identifies cooperation-related traits as a recurring area of comparatively lower self–other agreement.

Neuroticism describes tendencies concerning negative emotional experience in many traditional Big Five measures, but aspects of emotional distress are private. Others may observe irritability, nervous behavior, or signs of tension while missing internal rumination. The self's greater access to internal states can matter, although close acquaintances may learn patterns over time. This page does not interpret Neuroticism-specific scales; that subject has its own dedicated English Hub articles.

Openness to Experience includes forms of intellectual curiosity, imagination, aesthetic interest, and willingness to explore ideas or experiences, whose visibility varies. A person may think expansively without routinely expressing those interests in a shared environment. The 2025 review reports a tendency for self-ratings of Openness to be somewhat higher than observer ratings in studies involving close acquaintances. Differences also depend on how particular measures define the relevant domain.

The individual domains are wider than single behaviors. Comparisons become more informative when item content is matched at the relevant level and scores are interpreted according to the measure's actual structure. There is no universal thirty-facet Big Five instrument: the NEO inventories, BFI-2, and Big Five Aspect Scales define and score narrower components differently.

Why familiarity, intimacy, and the observer's role matter

Observers gain evidence by witnessing a person across repeated situations. Yet mere frequency of contact can amount to seeing the same narrow role many times. A coworker might see the target daily but only during structured meetings, whereas a close friend might observe spontaneous behavior, coping under stress, and private opinions. The two observers can have different strengths without either becoming universally superior.

In the Connelly and Ones meta-analysis, greater interaction frequency was associated with improved observer judgments, while interpersonal intimacy was particularly important for substantial gains in accuracy. Watson, Hubbard, and Wiese compared married couples, dating couples, and friends and found substantial agreement, with higher convergence among married partners in their samples.

Observer selection should match the research or assessment question. If the aim concerns behavior across family, social, and work environments, a combination of informants may reveal more than one supervisor or one romantic partner. Choosing several observers from the same network, however, can create shared blind spots or correlated biases.

The self–other knowledge asymmetry model

The self–other knowledge asymmetry, or SOKA, model predicts that the value of a rating source depends partly on a trait's observability and evaluativeness. Traits involving internal experiences give the person a possible informational advantage, whereas traits with visible behavioral consequences are accessible to informed observers. For attributes that are socially desirable or undesirable, motives to preserve a favorable image can affect self-judgment.

In Vazire's 2010 study, 165 participants were rated by friends and strangers and completed behavioral tests used as criteria. The self was especially informative for Neuroticism-related characteristics, friends for intellect-related characteristics, and sources performed similarly for Extraversion-related characteristics under the study's procedures. These results test a mechanism of source-specific accuracy rather than establish a universal ranking of all self-reports and observer reports.

The SOKA framework is most useful as a question generator: Is the rated attribute readily observable? Is it especially evaluative? Which source had access to relevant information? Is the supposed gold-standard criterion itself independent and trustworthy? The answers can differ across items within the same broad Big Five domain.

Why two people can disagree about the same trait

A mismatch can arise because the self and observer have sampled different contexts, reference groups, and periods. The same person might be meticulous in paid work and improvisational at home, unusually talkative among friends and quiet at unfamiliar gatherings. If the measure asks about typical behavior without specifying context, raters may silently answer different questions.

Memory and inference contribute additional differences. Respondents often summarize many episodes into an overall judgment, relying on salient experiences or an implicit comparison standard. Someone who compares themselves with highly organized colleagues may feel relatively disorganized, while a friend compares that person with a very different social circle. Both are making relative judgments that can diverge even when they know many of the same facts.

Motivations also matter. Self-presentation, defensiveness, modesty, and aspirational self-description can influence self-report. Observers may idealize a loved one, resent an acquaintance, infer personality from one conflict, or project their own characteristics onto a target. Ashton and Lee discuss assumed similarity—the tendency for observers' own traits to be reflected in their ratings of another person—as a particularly relevant phenomenon for some domains.

Disagreement is therefore an observation about two reports. It does not, by itself, identify dishonesty, pathology, lack of self-awareness, or a defective questionnaire. It becomes useful evidence when its sources can be investigated and the practical purpose of the assessment is specified.

Are self-reports or observer ratings more accurate?

There is no context-free winner because accuracy is defined relative to an external criterion. If the outcome is privately experienced worry, a self-report may have a measurement advantage. If the outcome is observable frequency of social participation, observers may contribute important evidence. If the outcome is a complex performance assessment, validity depends on whether the outcome itself is measured fairly and independently.

In a classic study, Kolar, Funder, and Colvin compared personality judgments with videotaped behavior. Ratings by one acquaintance slightly outperformed self-ratings for the behavioral criteria studied, and aggregated judgments from two acquaintances performed better than self-ratings. This finding applies to the studied criteria and sample; it does not mean that two friends always know someone's entire personality better than that person does.

Research on ordinary behavior also documents complementary information in the two sources. Vazire and Mehl tested self- and other-perspectives against behavioral evidence and found that the common assumption of universal self-superiority merits scrutiny. A meta-analysis of observer ratings and job performance by Oh, Wang, and Mount shows why informant perspectives can be useful for specific applied outcomes, although a performance prediction never yields certainty about an individual.

The best source is the one whose observation opportunities and measurement properties suit the question. In many research designs, the strongest approach is to treat self and observer reports as two imperfect, partly independent indicators rather than assume one is a permanent ground truth.

How to compare scores fairly

First verify that the ratings assess the same target, domain, instrument version, and reporting period. A general Big Five self-report and a questionnaire focused on socially observable behavior are not guaranteed to have interchangeable scores. Even two instruments using identical domain labels can select different content, weight subcomponents differently, and use different response categories.

Next determine whether there are validated self and informant forms. Their comparability should be evaluated empirically, including language version and the population in which they are used. For example, the 60-item BFI-2 developed by Soto and John measures five broad domains and fifteen instrument-defined facets. Its terminology includes Negative Emotionality and Open-Mindedness; those scales relate to traditional Neuroticism and Openness to Experience, but their names and contents should not be mechanically swapped when comparing instruments.

Compare the score scales. A raw item mean is not a percentile; a percentile is dependent on the reference group used to calculate it. A score derived from self-report norms should not automatically be compared with observer-report norms as though both had identical distributions. See our Big Five test scores and percentiles guide for score interpretation principles.

Review uncertainty. Short inventories contain fewer items and may be less informative for fine-grained individual comparisons. Differences smaller than the instrument's measurement uncertainty should not be dramatized. Where available, consider reliability, standard errors, confidence intervals, measurement invariance, and known informant effects, while keeping the analysis grounded in the particular instrument's technical manual.

Finally, read any discrepancy as a question to explore. A low observer rating of a work-related behavior may reflect a specific context or a disagreement over the underlying behavior. Check concrete examples before inferring that someone's personality is fundamentally different from their self-description.

Choosing and combining multiple observers

An observer should know the target well enough to have seen relevant behavior, be able to follow the instrument's instructions, and provide a reasonably independent evaluation. It is valuable to select people who see different parts of the target's life when the goal is a cross-context description. An observer who has just entered into a serious dispute with the target may have experiences worth understanding but may also have heightened response bias.

When multiple ratings are collected, average or latent-score methods can sometimes reduce idiosyncratic rater error and improve estimates. The decision to aggregate requires a plan. If observers differ systematically because they see different contexts, an average may hide a real person-by-context pattern. Researchers may report both the consensus score and informative disagreements.

The number of raters needed depends on the precision required, the instruments, the diversity of contexts, and the intended analysis. There is no defensible universal rule that exactly two, three, or five observers are sufficient. In high-stakes assessment, the burden of additional reports must be balanced against privacy, consent, fairness, and the validity evidence for the method.

Treat an observer report as that person's perspective on the target, not as ownership of the target's identity. The subject of the assessment should understand who may see the data, why the reports are collected, and how discrepancies will be handled.

How observer reports are used in Big Five instruments

The Five-Factor Model is a conceptual organization of personality dimensions; a questionnaire is one operationalization. Some major questionnaire families provide both self and observer versions, but their content and norms depend on the specific instrument. The NEO personality inventory family has an extensive self/informant research history, including McCrae and Costa's early validation across instruments and observers.

The Big Five Inventory family, including BFI-2, uses a different facet arrangement and scoring approach. Research using BFI-2 forms should retain its own labels and technical assumptions. A recent study of French-language short informant forms illustrates that translated and shortened observer versions need psychometric evaluation in their own populations; findings from one sample do not automatically establish equivalence everywhere.

One practical mistake is combining ratings from BFI-2, NEO, and unrelated quick online quizzes in a single five-column score table as if matching domain names meant matching scales. For defensible comparison, the evidence for linking scales must be established rather than assumed. The assessment purpose may justify domain-level interpretation while not supporting fine-grained facet comparisons.

Domain scores, facets, and the limits of precision

Broad-domain agreement can hide narrower differences. Two people can agree that a target is relatively extraverted while disagreeing about whether that extraversion reflects sociability, assertiveness, or energy. A single high-level correlation says little about which subcomponents are shared and which are source-specific. Where a validated measure permits it, facet-level analyses can clarify what is being rated.

However, facet taxonomies are instrument-specific. NEO inventories conventionally organize thirty facets, BFI-2 contains fifteen defined facets, and the Big Five Aspect Scales focus on ten aspects. These structures are not universal subdivisions of an invariant five-factor taxonomy. Facets should be named and interpreted in the terms of the instrument that generated them.

Finer scores also demand more measurement information. A brief scale designed to screen broad domains may not reliably distinguish subtle individual facet patterns, particularly in one-time high-stakes contexts. Consider a full validated version if the research question genuinely requires narrower distinctions.

Reliability, validity, and source-specific method effects

Reliability concerns the consistency or precision of scores under specified conditions; validity concerns evidence supporting the interpretation and intended use. Both matter in comparing self and other reports. Two reliable rating sources can disagree because they capture different situations or use different reference standards. Two correlated scores can still share method-related bias.

A strong research design can evaluate several traits across several methods and compare their convergence and discrimination. Independent behavioral observations, longitudinal outcomes, and more than one observer can help, when the criterion is appropriate. This article applies those principles to the Big Five; general psychometric theory, test construction, and formal assessment standards belong to the dedicated measurement coverage elsewhere in the Hub.

Observer agreement is not the same as independent validation when observers influence one another, share a group norm, or know the self-report result. Blind ratings and separate collection procedures may therefore matter. Similarly, asking only close friends who were selected because they agree with the target can produce a biased sample of informants.

Measurement invariance can also matter when comparing informants across language groups, age categories, workplaces, or cultural settings. Matching translations and response formats is useful but insufficient on its own. Scale properties must be investigated in the groups and versions under comparison.

Observer reports and cross-cultural interpretation

Self-description and social judgment occur within cultural expectations about modesty, emotional expression, sociability, responsibility, and acceptable behavior. Informants may have different opportunities to observe a trait in cultures where some behaviors are more private, socially constrained, or role-dependent. Agreement patterns obtained in one language and setting should not be treated as universal score-conversion rules.

Cross-cultural studies of trait structure provide important support for Big Five concepts across many samples, but they do not show that every lexical study in every language reproduces precisely the same five-factor pattern. The same restraint applies to self–observer correlations, informant norms, and the meaning of particular items. Research with specific populations is preferable to unexamined generalization.

Comparing people from different communities, occupational roles, ages, or linguistic backgrounds requires attention to how descriptors are understood. A mismatch may reveal translation problems, social norms, or the observers' restricted context as well as individual personality differences.

Are observer ratings better for work, school, or relationships?

Observer reports can be informative about visible work behavior, classroom participation, and interpersonal routines. Coworkers and supervisors may see dependability or cooperation that colleagues experience directly, while teachers or peers may be able to describe classroom behavior. Yet the outcome being studied must be distinguished from the trait score itself, and the observer's opportunity to sample behavior remains limited.

For workplace questions, the meta-analysis by Oh, Wang, and Mount provides evidence that observer-rated five-factor traits can relate to job performance. Such statistical relationships vary by criterion, role, measurement method, and population. A favorable Big Five rating cannot guarantee performance, suitability, or leadership potential; any consequential personnel decision requires appropriate job-related validity evidence and safeguards against discrimination.

In personal relationships, perceived personality may influence interaction, but a partner's description also reflects the relationship. Disagreement can be a starting point for respectful dialogue about specific experiences. It cannot serve as a clinical judgment or definitive adjudication of character. Information obtained in relational conflict may be especially vulnerable to mutual attribution biases.

For education or personal development, reports are most useful when framed as context-sensitive feedback. A student's quiet classroom behavior may coexist with sociability elsewhere. A single observer's report should not be treated as a global label.

Can self–other disagreement reveal personality change?

A single discrepancy cannot establish that a trait has changed. Differences in current stress, social roles, instructions, memory, and which situations observers saw can shift ratings without a corresponding shift in a stable disposition. To study change, researchers need repeated assessments, comparable instruments, and a design that separates retest fluctuations from systematic changes in characteristic behavior.

Longitudinal studies have used multiple sources to examine temporal patterns. For example, Costa and McCrae collected self-reports and spouse reports in a study of adult personality over six years. Multi-source designs can strengthen interpretation, though they require careful attention to changes in both the target and the observers' relationship with that target.

At an individual level, a disagreement discovered today is best interpreted against specific episodes and repeated observations. Important questions include whether the source of behavior changed, whether the same observer rated the target at both times, and whether the difference exceeds the expected variability of the scales.

What about AI, digital footprints, and automated personality ratings?

Some models estimate Big Five scores from text, social media, or other behavioral traces. Such systems produce statistical inferences from available data; they are neither traditional self-report questionnaires nor equivalent to a knowledgeable human informant. Their accuracy depends on the target measure, training data, population, modality, opportunity to observe behavior, and external validation.

A correlation between an algorithm's output and a questionnaire does not prove that the algorithm directly observes the full personality construct, establishes psychological ground truth, or knows a person better than friends do. Models can learn demographic, linguistic, platform, and social-context signals that do not generalize. Privacy and consent are significant concerns when someone is profiled without voluntarily providing an assessment.

Human observer ratings also raise privacy issues, but algorithmic inference introduces different data-control and bias risks. A responsible comparison must specify exactly which criterion, prediction task, dataset, and subgroup were assessed. The full Big Five-specific digital inference topic is a distinct editorial owner in this knowledge network.

A practical procedure for comparing your own ratings with a friend's

Choose a reputable Big Five inventory with suitable self and informant forms and read its instructions. Complete the self-report independently, then ask a trusted person to fill in the informant version based on direct experience rather than guesses about private thoughts. State the purpose clearly and allow the other person to decline. The goal is to gather perspectives, not to stage a test of loyalty or honesty.

Compare broad patterns first, using the proper scoring instructions. Check whether both results use the same scale or a justified linkage. If scores differ, look at the behavior-specific items underlying the disagreement. Ask what situations the observer had in mind, what the self-report respondent had in mind, and how each interpreted the item.

Use neutral, specific questions: Which situations led to this rating? Does this happen only at work? Would someone who knows me in a different setting see the same thing? What did we each understand by the word organized or sociable? This conversation can reveal contextual information that a summary graph conceals.

Interpret the numbers as estimates on continuous dimensions. Avoid imposing universal high/low cutoffs, turning a domain score into a personality type, or using one person's lower rating as evidence of a mental disorder. For important decisions, consult an appropriately qualified professional using tools validated for that decision.

Ethical and clinical boundaries

Big Five scores describe normal-range individual differences and are not clinical diagnoses. Neither a self-report nor an informant report of high Negative Emotionality, low Agreeableness, or any other domain constitutes evidence that an individual meets diagnostic criteria for a mental health condition or personality disorder. A professional clinical assessment requires a separate question, evidence base, and process.

Informant data are personal information about both the target and the relationship. Obtain consent when appropriate, minimize identifiable data, protect access, and explain who will receive findings. Avoid quietly recruiting coworkers or intimate partners to rate someone in ways that could expose private feelings or create interpersonal harm.

High-stakes applications deserve the most scrutiny. An organization cannot justify a selection decision merely by saying that observer ratings were shown to be predictive somewhere in the literature. The specific instrument, intended use, relevant population, fairness assessment, and independent evidence have to support the particular use. A score provides probabilistic evidence, never a deterministic forecast.

Frequently asked questions

Are observer Big Five ratings more objective? They are additional judgments made from a different perspective. They can reveal visible behavior that a person underestimates, but observers also have limits, preferences, and biases. Objectivity requires a sound design and appropriate independent evidence, not merely a third-person questionnaire.

How much self–other agreement is normal? There is no universally applicable individual cutoff. Meta-analytic correlations characterize groups and vary by domain, instrument, familiarity, and methodology. A particular pair's score difference should be examined against the instrument's scaling and precision.

Which Big Five trait tends to show the strongest agreement? Extraversion often ranks relatively high in comparative studies because aspects of social behavior are visible. For example, the corrected domain correlation in the 2007 Connolly meta-analysis was .62, while other broad traits were lower. The expected pattern can vary with measures and samples.

Which trait tends to show weaker agreement? Agreeableness frequently shows lower convergence, plausibly because it is interpersonal and evaluative. Differences between judges can still be informative, and a specific person's disagreement does not identify who is correct.

Can a stranger complete an observer form? Some research designs use stranger ratings to study first impressions and visible cues, but strangers have limited behavioral information. Their reports should not automatically be treated as interchangeable with those of spouses or close friends.

Does a spouse always know someone better than a friend? No. Intimacy often increases the range of behavior visible to an observer, but it can also introduce idealization, conflict-related judgments, and assumed similarity. The relevant trait and observation setting matter.

Can self and observer scores be averaged? Sometimes, when the instrument, study design, and score interpretation support combining them. A simple average is not automatically justified, especially when norms or response processes differ. Researchers may instead model shared and source-specific variance.

What if my friend rates me as less conscientious than I rate myself? Review what behaviors and situations each person considered and whether the questionnaires were comparable. A disagreement does not establish deception, a hidden personality type, or a clinical condition.

Can I compare BFI-2 self-ratings with NEO observer ratings? Both may concern Big Five domains, but they are different operational measures. Score equivalence and facet comparisons require evidence; matching labels alone are insufficient.

Do observers assess internal emotions accurately? They can infer habitual emotional patterns from behavior and develop informed judgments with greater familiarity. They cannot directly observe private experience. Some internal characteristics may be more accurately captured by self-ratings under appropriate conditions.

Are observer ratings useful for hiring? Research supports predictive value in certain occupational settings, but use in selection requires job-relevant validation, fair procedures, attention to social desirability and rater bias, privacy protections, and appropriate professional oversight.

Can a disagreement indicate a personality disorder? No. Big Five rating discrepancies are part of normal personality measurement research. Diagnosis requires separate clinical assessment using appropriate criteria.

Does AI infer personality the same way an informant does? No. Algorithmic estimates from digital traces are model-based predictions contingent on training and evaluation, with important validity, generalizability, bias, and privacy limits.

What is the best way to understand conflicting ratings? Compare comparable instruments, examine specific items and contexts, invite more than one informed perspective when justified, and preserve uncertainty. The aim is a better evidence-based description, not a verdict about who knows the person best.

References

Ashton, M. C., & Lee, K. (2025). Self- and observer reports of personality. Annual Review of Psychology, 76, 771–795. https://doi.org/10.1146/annurev-psych-020124-115044.

Connelly, B. S., & Ones, D. S. (2010). An other perspective on personality: Meta-analytic integration of observers' accuracy and predictive validity. Psychological Bulletin, 136(6), 1092–1122. https://doi.org/10.1037/a0021212.

Connolly, J. J., Kavanagh, E. J., & Viswesvaran, C. (2007). The convergent validity between self and observer ratings of personality: A meta-analytic review. International Journal of Selection and Assessment, 15(1), 110–117. https://doi.org/10.1111/j.1468-2389.2007.00371.x.

Costa, P. T., Jr., & McCrae, R. R. (1988). Personality in adulthood: A six-year longitudinal study of self-reports and spouse ratings on the NEO Personality Inventory. Journal of Personality and Social Psychology, 54(5), 853–863. https://doi.org/10.1037/0022-3514.54.5.853.

Funder, D. C., & West, S. G. (1993). Consensus, self-other agreement, and accuracy in personality judgment: An introduction. Journal of Personality, 61(4), 457–476. https://doi.org/10.1111/j.1467-6494.1993.tb00778.x.

John, O. P., & Robins, R. W. (1993). Determinants of interjudge agreement on personality traits: The Big Five domains, observability, evaluativeness, and the unique perspective of the self. Journal of Personality, 61(4), 521–551. https://doi.org/10.1111/j.1467-6494.1993.tb00781.x.

Kim, H., Di Domenico, S. I., & Connelly, B. S. (2019). Self–other agreement in personality reports: A meta-analytic comparison of self- and informant-report means. Psychological Science, 30(1), 129–138. https://doi.org/10.1177/0956797618810000.

Kolar, D. W., Funder, D. C., & Colvin, C. R. (1996). Comparing the accuracy of personality judgments by the self and knowledgeable others. Journal of Personality, 64(2), 311–337. https://doi.org/10.1111/j.1467-6494.1996.tb00513.x.

McCrae, R. R., & Costa, P. T., Jr. (1987). Validation of the five-factor model of personality across instruments and observers. Journal of Personality and Social Psychology, 52(1), 81–90. https://doi.org/10.1037/0022-3514.52.1.81.

Oh, I.-S., Wang, G., & Mount, M. K. (2011). Validity of observer ratings of the five-factor model of personality traits: A meta-analysis. Journal of Applied Psychology, 96(4), 762–773. https://doi.org/10.1037/a0021832.

Soto, C. J., & John, O. P. (2017). The next Big Five Inventory (BFI-2): Developing and assessing a hierarchical model with 15 facets to enhance bandwidth, fidelity, and predictive power. Journal of Personality and Social Psychology, 113(1), 117–143. https://doi.org/10.1037/pspp0000096.

Vazire, S. (2010). Who knows what about a person? The self–other knowledge asymmetry (SOKA) model. Journal of Personality and Social Psychology, 98(2), 281–300. https://doi.org/10.1037/a0017908.

Vazire, S., & Mehl, M. R. (2008). Knowing me, knowing you: The accuracy and unique predictive validity of self-ratings and other-ratings of daily behavior. Journal of Personality and Social Psychology, 95(5), 1202–1216. https://doi.org/10.1037/a0013314.

Watson, D., Hubbard, B., & Wiese, D. (2000). Self–other agreement in personality and affectivity: The role of acquaintanceship, trait visibility, and assumed similarity. Journal of Personality and Social Psychology, 78(3), 546–558. https://doi.org/10.1037/0022-3514.78.3.546.

Psychological Encyclopedia