Limits of Forensic Psychology: Prediction, Causation, and Expert Inference

4 days ago

21 min read

Author: Ukrainian Psychological Hub · Published: October 11, 2026 · Editorial Policy

Ukrainian Psychological Hub | October 11, 2026

Forensic psychology can clarify legally relevant behavior, mental functioning, risk, memory, and decision-making. Its conclusions are strongest when the question is precisely defined, the evidence is independently checked, the method has demonstrated validity for that question, and the expert explains the uncertainty that remains. The central limits of forensic psychology concern three forms of inference: predicting what someone may do, identifying what caused a particular act or outcome, and translating psychological evidence into a defensible opinion for a legal decision-maker.

A psychological observation, a diagnosis, a test score, a statistical association, an estimated probability, an expert opinion, and a court's legal finding each answer a different question. Moving from one to another requires additional evidence and reasoning. The American Psychological Association's Specialty Guidelines for Forensic Psychology place accurate methods, multiple information sources, alternative hypotheses, role clarity, and communication of limitations at the center of responsible forensic practice.

What are the limits of forensic psychology?

The limits of forensic psychology are the boundaries on what psychological science and forensic assessment can reliably establish about an individual, an event, a legal capacity, or a future outcome. Those boundaries arise from imperfect measurements, incomplete records, differences among people and settings, probabilistic rather than deterministic patterns, potential evaluator bias, and the distinct responsibilities of scientific experts and legal authorities. For a broader introduction to the field, see Forensic Psychology: What It Is and What Forensic Psychologists Do.

Forensic psychology is useful when it provides disciplined answers to questions for which psychological expertise has demonstrable relevance. A psychologist may assess functional abilities related to legal proceedings, evaluate symptoms and their possible relationship to a legally relevant time, identify empirically supported risk and protective factors, explain established limitations of memory, or assess how interview conditions may affect evidence. The strength of each answer depends on the actual procedure and its application, not the authority implied by the expert's job title.

The field cannot independently establish that a person committed a crime, supply a universal numerical probability of an individual's future violence, determine the legal meaning of criminal responsibility across jurisdictions, recover an unrecorded mental state with certainty, or infer a specific cause simply because a risk factor or diagnosis is present. Those are limitations of the corresponding inference, which should be stated where the inference is offered.

The inferential ladder: from data to a legal conclusion

Psychological constructs and behavioral observations

A construct is a defined theoretical attribute, such as impulsivity, cognitive functioning, suggestibility, or a tendency to respond defensively on a questionnaire. An observation is a behavior or statement recorded under particular conditions. An expert who observed disorganized speech during an interview has direct evidence of that speech in that interview; whether it reflects enduring cognitive impairment, intoxication, stress, language differences, or a mental disorder requires further investigation. Observation is the starting point of interpretation, not its endpoint.

Assessment results, scores, and diagnoses

An assessment result is produced by a specified method. A psychometric score compares responses or performance with a reference framework, subject to measurement error, administration conditions, validity evidence, and normative population. A clinical diagnosis describes a pattern meeting diagnostic criteria. A screening result indicates a reason for further evaluation and has its own false-positive and false-negative characteristics. These categories must remain distinct in forensic reports because the questions to which they apply differ.

A diagnosis alone does not determine a person's ability to understand proceedings, form an intention at an earlier time, tell the truth, act violently, or satisfy a jurisdiction's legal test. An elevated risk score neither proves past conduct nor establishes guilt. The APA guidelines specifically recommend attention to legally relevant functional factors and caution when clinical diagnoses are used in forensic settings.

Expert opinions and legal findings

A forensic opinion is a professional inference made from a stated evidence base for a defined psycholegal question. The opinion can be evaluated by examining the assumptions, data quality, methods, relevant research, and competing explanations. A legal finding, including guilt, liability, competency, admissibility, or the satisfaction of a statutory defense, is made by the authorized legal decision-maker under applicable law. The expert provides relevant science and evaluation; the tribunal applies the governing legal standard.

An expert's report should therefore distinguish four layers: what the records and examinations directly show; what those observations reasonably support; what remains uncertain or contested; and what legal question the opinion is intended to assist. The research-to-case translation must be explicit at each layer.

Prediction: why a risk estimate is not an individual forecast

Prediction in forensic psychology commonly concerns future violence, general or sexual recidivism, acute aggression in an institutional setting, or a change in legally relevant behavior. Predictive instruments use statistical models, structured professional judgment, or a combination of evidence-based factors and case formulation. Their usefulness depends on the defined outcome, assessment population, follow-up interval, interventions, and decision being considered.

A risk factor is a characteristic associated with an outcome in a specified research population. A protective factor may be associated with lower risk or support safer management, but the strength and relevance of that relationship must be evaluated. Neither kind of factor establishes a necessary or sufficient cause of criminal conduct. Some variables are historical and fixed; others may change with treatment, housing, supervision, social circumstances, or clinical state. A risk estimate made under one set of conditions may require revision when conditions change.

Discrimination, calibration, sensitivity, and specificity

Discrimination describes whether a model tends to assign higher scores to people who later experience the specified outcome than to those who do not. An area under the receiver operating characteristic curve, or AUC, near 0.5 indicates chance-level discrimination under the comparison; an AUC of 0.7 describes moderate rank ordering, not a 70% probability that a designated individual will commit violence. AUC does not by itself determine whether a model's absolute risk estimates are accurate.

Calibration asks whether predicted percentages agree with the frequencies observed in comparable cases. If people assigned 20% risk actually experience the outcome about 10% of the time, the model is miscalibrated for that population or context. Sensitivity is the proportion of actual cases detected by a chosen threshold, and specificity is the proportion of non-cases correctly classified. Positive predictive value asks how often someone classified positive actually experiences the outcome; negative predictive value asks the corresponding question for a negative classification. These measures answer different practical questions.

In a 2023 systematic review and meta-analysis of forensic psychiatric populations, Ogonah and colleagues examined 50 publications, 36 tools, and 10,460 participants. Pooled AUCs for selected violent-recidivism predictions were approximately 0.64–0.72. Forty-nine of the 50 studies were rated at high risk of bias, and only one reported calibration. The authors described mixed predictive performance and major shortcomings in validation and reporting. These are results for specific forensic psychiatric discharge populations and instruments; they are not general accuracy figures for all offenders or all psychological evaluations.

A 2026 systematic review of the Violence Risk Appraisal Guide–Revised, based on 15 studies and 3,932 participants, reported pooled AUCs of 0.71 for general recidivism, 0.72 for violent recidivism, and 0.65 for sexual recidivism. The authors also identified questions of generalizability and the need for calibration research. Differences in outcome definition and population are essential to interpreting those estimates.

The base-rate problem: a worked example

Imagine 1,000 people in a hypothetical population, with a specified adverse outcome occurring in 5% over one year. Assume a classification threshold with 80% sensitivity and 80% specificity. Of 50 people who experience the outcome, 40 are correctly flagged and 10 are missed. Of 950 who do not experience it, 190 are incorrectly flagged and 760 are correctly classified. Among the 230 positive classifications, only 40 correspond to the outcome: positive predictive value is approximately 17.4%.

This arithmetic is an illustration rather than a validation finding for any real forensic instrument. It shows why a tool can have seemingly respectable sensitivity and specificity while producing many false positives when the outcome is uncommon. Raising the threshold may reduce false positives but increase false negatives. The resulting trade-off has consequences for personal liberty, treatment allocation, public safety, and the fairness of legal decisions.

A 2012 systematic review and meta-analysis by Fazel and colleagues examined risk assessment instruments across 73 samples involving 24,827 people and emphasized that predictive performance must be understood in relation to the decision and outcome of interest. A separate systematic review of predictive-validity reporting by Singh, Desmarais, and Van Dorn described recurring problems in how accuracy statistics are selected and communicated.

Time horizons and shifting conditions

Prediction for violence within the next day in a secure inpatient unit is a different task from predicting reconviction over five years after community release. The outcome may be recorded incidents, arrests, charges, convictions, or self-reported aggression; these are related but different measurements. Follow-up intensity, policing practices, and administrative data quality can affect observed recidivism independently of the person's underlying behavior.

A 2018 systematic review of inpatient forensic instruments found that instruments designed for imminent aggression had higher median discrimination than tools aimed at longer-term outcomes. That finding supports a distinction among tasks, rather than a single judgment about whether violence can be predicted. An evaluator should state the horizon, setting, outcome, and conditions of any estimate.

Causation: why a statistical association does not explain an individual act

Causal explanation requires evidence that an exposure, condition, or mechanism contributed to an outcome in a way that survives plausible alternative explanations. Correlation can arise from genuine causal influence, confounding, reverse causation, selection effects, measurement artifacts, and chance. A factor associated with offending across many people may offer limited evidence about why one person acted in one particular way on a particular occasion.

Consider an observed association between substance use and a category of offenses. Interpreting the association requires information on timing, circumstances, social conditions, comorbid factors, opportunity, and possible alternative pathways. The presence of substance use in an individual case does not alone establish that it caused an offense or determine the individual's mental state. Likewise, childhood adversity can be relevant to developmental research without providing a single-person causal biography of a criminal act.

Strong causal arguments ordinarily identify a plausible mechanism, establish temporal order, test rival explanations, assess how evidence was selected, and indicate what would be expected if the favored explanation were mistaken. In many retrospective cases, the data required for decisive causal tests were never collected. That does not render all interpretation uninformative; it determines the appropriate level of the conclusion.

Group evidence and the individual-case inference

Research may show that one group has an elevated average risk relative to another, but the distributions often overlap extensively. Applying the group's average relationship directly to a person can produce an ecological inference error. The relevant individual may differ from the study group in critical respects, and treatment, supervision, age, opportunity, and social context may change what happens next.

A forensic expert should identify whether an assertion is population-level, subgroup-level, or person-specific; whether the evidence was developed for a similar population and decision context; and how the individual evidence modifies the general pattern. These distinctions are especially important when statistical results are presented as though they reconstruct past intent or identify an otherwise unknown offender.

Causation in legal proceedings

Psychological causation and legal causation involve different tasks. A psychologist may explain factors that plausibly influenced functioning or decision-making, while legal doctrines of causation, responsibility, intent, and foreseeability are defined and applied under the applicable law. Scientific support for an influence does not by itself settle proximate cause, diminished responsibility, insanity, or any other legal issue.

The Reference Manual on Scientific Evidence provides a broad scientific framework for evaluating statistical and causal reasoning in litigation. Forensic psychologists must also attend to the jurisdiction-specific question they were asked and explain the relationship between the empirical evidence and that question.

Retrospective inference: what can be known about a past mental state?

Some forensic evaluations concern a person's capacities or mental state at a time before the evaluation. Such work may draw on contemporaneous clinical notes, witness accounts, digital records, behavior near the event, interviews, and other documented material. Reliability often depends on the quality, timing, independence, and completeness of those sources. Later recollections may be altered by knowledge of the alleged event, litigation, repeated interviews, and incentives to present a particular account.

A present-day clinical examination may support an understanding of current symptoms or functioning. Reconstructing a person's mental condition on a historical date requires additional reasoning. Documented symptoms do not automatically establish what the person understood, intended, or could control at the legally relevant time. The evaluator should explain alternative interpretations and distinguish what contemporaneous evidence establishes from what is inferred afterward.

Competency, legal capacity, and criminal responsibility

Competency to stand trial generally concerns present abilities connected to participation in proceedings, subject to the controlling legal test. Criminal responsibility or an insanity defense typically addresses a person's mental state in relation to an alleged act, under jurisdiction-specific doctrine. Civil capacities can concern discrete decisions, time periods, and statutory requirements. These are separate legal questions; one answer does not determine another.

Psychologists may describe relevant functioning, limitations, and the support for a forensic opinion. The court determines whether the legal standard is satisfied. Empirical reliability of professional judgments is itself an appropriate topic of review: Guarnera and Murrie's 2017 systematic review and meta-analysis examined the field reliability of competency and sanity opinions and highlighted the importance of examining agreement in ordinary forensic practice. A single expert's apparent certainty is not a substitute for demonstrated reliability.

Clinical diagnoses, symptoms, and the problem of stigma

Mental disorders and criminal behavior are heterogeneous subjects. The presence of a psychiatric diagnosis supplies clinical information about diagnostic criteria, course, or functioning; it does not identify an offender, prove dangerousness, or establish responsibility for a particular action. Evaluation of risk may require a range of clinical, historical, situational, social, and environmental information, considered in context. A person can experience severe symptoms and present little risk of violence, just as harmful behavior can occur without a diagnosable mental disorder.

The relationship between observed behavior, reported symptoms, diagnosis, and legally relevant capacities must be argued independently. It is scientifically unsound to transform a diagnosis into an explanation of crime through stigma or to equate unusual behavior with deliberate deception. This distinction is explored further in the separate English Hub article Schizophrenia and Violence: What Research Really Shows, which retains ownership of the diagnosis-specific evidence.

The limits of tests, scales, and professional judgment

Psychometric validity is purpose-specific

A test's validity is about evidence supporting particular interpretations and uses of its scores. Reliability, norms, measurement invariance, and criterion validity may vary by age, gender, culture, language, disability, and setting. A scale validated for clinical screening may not be validated for competency opinions, legal causation, offender identification, or violence prediction. Changes in how a tool is administered or interpreted can invalidate an apparently precise result. The differences in purpose and standards are explored in Forensic vs. Clinical Psychological Evaluation: Purpose, Roles, and Evidence.

Score cutoffs create classification errors. Reports should identify the instrument and version, target construct, comparison population, limitations of available norms, relevant reliability and validity evidence, and how the particular result contributes to the ultimate opinion. More decimal places in a result do not create more empirical certainty.

Actuarial methods, structured professional judgment, and unstructured opinions

An actuarial approach combines predictors according to a specified rule. Structured professional judgment organizes evidence about risk and management needs while allowing a formulated account of the case. An unstructured judgment depends primarily on the assessor's unaided appraisal. Research does not license treating all three as interchangeable or assuming that an expert's personal experience can always improve a statistical estimate.

The review by Garb and Wood on methodological advances in statistical prediction discusses the comparative role of statistical prediction and structured professional judgment and warns that clinician overrides can reduce predictive validity. In practice, a structured process should make its decision rules, assumptions, uncertainties, and grounds for any departure from a score transparent. Where an intervention is recommended, its goals and the evidence that it changes outcomes should be distinguished from predictive performance.

Response style, possible exaggeration, and malingering

Forensic evaluations may involve incomplete disclosure, misunderstanding, defensive responding, genuine distress, exaggerated symptoms, or intentional production of symptoms for an external incentive. A validity indicator is evidence about response patterns under specific conditions, not a direct measurement of motive. An unusual test pattern does not establish malingering without an appropriately supported analysis of alternatives and the full case context.

Interviews, standardized procedures, collateral material, and documentation may each supply information, but the expert must state when evidence conflicts or remains unverified. In particular, neither a diagnosis nor an instrument's validity scale should be used alone to pronounce a person truthful or deceptive.

The limits of expert inference, expertise, and confidence

Forensic opinion involves several possible points of failure: selecting which records to review, defining the referral question, choosing instruments, interpreting scores, weighing contradictory statements, deciding which explanatory model fits, and summarizing the conclusion. A well-validated test does not automatically validate every claim later made from its results. A specialist may be appropriately qualified in one forensic task while lacking expertise in another.

The APA guidelines call for adequate scientific foundations, appropriate methods, disclosure of limitations, consideration of rival hypotheses, and differentiation of observations, inferences, and conclusions. They also address opinions about people who have not been examined. When an examination or independent corroboration is unavailable, that limitation belongs in the opinion itself.

Contextual bias and adversarial allegiance

Knowledge of incriminating records, retaining-party expectations, or the supposed facts of a case can shape how ambiguous data are interpreted. In an experimental study, Murrie and colleagues (2013) found evidence that forensic experts' scores were influenced by the side that ostensibly retained them. The study establishes a measurable vulnerability to adversarial allegiance in its research conditions; it does not show that every professional opinion is biased or quantify error in an unrelated case.

Procedural protections include an explicit referral question, systematic review of alternative hypotheses, source-by-source attribution, documentation of discordant evidence, independent consultation when indicated, and careful separation of facts supplied by advocates from facts independently established. Bias cannot be neutralized merely by declaring impartiality.

Reproducibility and agreement

An opinion is easier to evaluate when another qualified professional can see the evidence base, method, rationale, and limitations. Agreement between experts on the same facts is an empirical question; legal systems should not assume perfect agreement merely because each professional is trained. Studies of field reliability, including Guarnera and Murrie (2017), demonstrate why ordinary-practice reproducibility matters alongside research conducted under more controlled conditions.

Disagreement may arise from different source information, different definitions of the legal question, uncertainty in assessment, competing reasonable models, or questionable methods. A responsible report identifies the substantive source of disagreement where possible and avoids recasting every disagreement as professional incompetence.

Scientific reliability and legal admissibility are separate questions

In U.S. federal courts, the amended Federal Rule of Evidence 702, effective December 1, 2023, requires the party offering expert testimony to demonstrate that it is more likely than not that the rule's admissibility requirements are met. The testimony must help the factfinder, rest on sufficient facts or data, result from reliable principles and methods, and reflect reliable application of those principles and methods to the case. The gatekeeping framework discussed in Daubert is relevant to this federal rule.

A scientifically plausible proposition may still fail to meet an evidentiary requirement because the methods do not support the specific asserted conclusion. Conversely, admission of expert testimony is a procedural judicial determination, not a scientific certificate of perfect accuracy. The legal rule does not transfer automatically to every U.S. state, every administrative proceeding, or another country; other jurisdictions may apply different evidentiary and substantive standards.

The U.S. Department of Justice's Uniform Language for Testimony and Reports illustrates an institutional emphasis on limiting testimony to appropriately supported statements in forensic disciplines. Its documents govern covered DOJ forensic examiners and are not a general licensing or competency standard for all psychologists. The broader principle is that the strength of a statement must match the evidence behind it.

Investigative claims: the danger of converting a plausible story into proof

An investigator might infer an unknown offender's traits from the scene of an offense, infer deception from demeanor, or infer memory accuracy from a witness's confidence. Such claims require validation for the exact task. Some research can identify broad statistical regularities, but turning them into a reliable diagnosis, identity, motive, or conclusion about guilt is a distinct scientific challenge. The vividness of a profile or retrospective narrative does not establish its accuracy.

Experts examining eyewitness reports, interrogations, or forensic interviewing can describe experimentally supported influences on memory and reporting. The National Research Council's 2014 report on eyewitness identification demonstrates the importance of evidence-based procedure and the limits of translating experimental findings into individual identification accuracy. An expert may explain how a procedure could increase error risk without being able to determine from psychology alone whether a particular witness's identification is correct.

Laboratory findings, field validations, actual investigative procedures, and outcomes in the specific case should be presented separately. A common scientific finding about an influence on memory should not be relabeled a determination that a witness lied or that a defendant is guilty.

Cultural, demographic, and jurisdictional limits

A prediction model developed on a predominantly male sample of forensic psychiatric patients from high-income countries may perform differently with women, young people, other populations, or people subject to different policing and service systems. The 2023 Ogonah review found limited evidence for women and no studies from low- or middle-income countries in the reviewed forensic psychiatric validation literature. The resulting uncertainty should be explicit when an evaluator wishes to generalize beyond the data.

Cultural fairness concerns language access, culturally sensitive interviewing, measurement equivalence, accurate interpretation of behavior, historical inequalities in data collection, and the consequences of acting on uncertain classifications. A model can display similar AUC values between groups while differing in calibration or false-positive rates. Likewise, a numerical parity measure does not settle the legal or ethical appropriateness of a decision rule.

Jurisdictional differences also affect what the expert is being asked to establish. Definitions of insanity, trial competence, legal capacity, liability, and evidentiary standards differ in law. A psychologist's empirical finding should be mapped to the controlling local test; it should never be framed as a universally applicable legal conclusion.

Algorithmic prediction and machine learning in criminal justice

Machine-learning methods may combine many variables, learn nonlinear patterns, or produce a risk score efficiently. Their use introduces requirements for external validation, transparent outcome definitions, representative training data, recalibration over time, error analysis, and meaningful human scrutiny. A complex algorithm does not overcome selection bias in the records used to train it.

For example, a model trained on arrests or convictions may reproduce variation in enforcement or prosecution practices alongside any variation in actual offending. A model that ranks people fairly well on historical data may nonetheless overestimate absolute risk or degrade in a new jurisdiction. Decisions affecting liberty require particular attention to error costs, contestability, documented thresholds, and whether the score changes outcomes for the better. Evidence about the performance of a named tool must be assessed separately from promotional claims about artificial intelligence in general.

How a defensible forensic opinion is constructed

An effective forensic opinion begins with the question asked by the retaining party or tribunal and specifies the relevant jurisdiction, time, person, and legally relevant function. The evaluator then collects information proportionate to that question, records the sources and their limitations, checks relevant records where possible, selects methods with appropriate validation, and distinguishes clinical, statistical, and legal categories.

The evaluator should test plausible rival explanations rather than build a one-sided story around the first hypothesis. For each important conclusion, the report should state the supporting observations, how the method connects them to the opinion, the counterevidence, the status of corroboration, the population and time limits of any research, and an appropriate expression of uncertainty. The ultimate statement should remain within the evaluation's scope and the practitioner's expertise.

A useful practical test is whether a reader can reconstruct why an opinion follows from its evidence, what would change that opinion, and what the report does not establish. The APA forensic specialty guidelines support these practices through their provisions on methods, multiple sources, legally relevant factors, transparent testimony, and communication of conclusions.

Questions to ask about a forensic report

The most informative review questions are whether the referral question and legal standard are identified; whether factual premises have reliable sources; whether the evaluation involved direct examination or only records; whether tests were validated for this person and this purpose; whether alternative explanations were considered; whether risk statistics include their denominators, time frames, and error characteristics; and whether the final language exceeds the demonstrated evidence.

Other essential questions concern who commissioned the report, how conflicting data were handled, what population any predictive model represents, what safeguards addressed bias, and whether the expert has explained the likely impact of missing information. These questions assess the strength of the inference without presuming which side of a legal dispute is correct.

Three examples of the boundary between evidence and conclusion

Example 1: Elevated risk score before a release decision

A person receives a high score on a recidivism assessment. The score may summarize patterns observed in a particular validation population. Its meaning depends on the instrument, score distribution, predicted outcome, follow-up period, local baseline rate, and intervention context. A decision-maker still needs to understand the false-positive risk, the available protective factors, the person's circumstances, and the applicable release rules. The score alone does not establish an inevitable future offense, much less a prior offense.

Example 2: Diagnosis during a criminal-responsibility evaluation

Clinical records show that a defendant had symptoms of a mental disorder near the date of an alleged act. The evidence may support diagnostic and functional questions. The evaluator must still examine the temporal record, legal criteria, evidence concerning relevant knowledge or functioning, and plausible alternatives. A diagnosis alone does not decide whether a jurisdiction's legal responsibility standard is met. The authorized decision-maker applies the legal rule.

Example 3: Strongly expressed eyewitness confidence

A witness expresses high confidence after an identification. Research may help evaluate the timing and conditions under which confidence was collected, including procedural influences. The witness's statement is evidence to be assessed within the entire case. Psychology cannot translate confidence alone into a guarantee of accuracy, nor infer deliberate deception from a later inconsistency. The correct level of conclusion follows from the details of the identification procedure and the evidence taken together.

Myths and misconceptions about forensic psychologists

The first recurring myth is that an experienced psychologist can identify a criminal by personality or read concealed intentions from behavior alone. Identifying an offender requires case evidence and a lawful investigation; generalized behavioral patterns do not uniquely identify a person. A second myth is that a validated test produces a verdict. Validation is always tied to a specified score interpretation and use, and error remains possible.

A third myth treats high predictive AUC as a direct personal probability of future offending. AUC primarily measures comparative ranking and supplies no standalone individual probability. A fourth myth assumes mental illness explains criminal action. Diagnostic status, functional impairment, legal responsibility, and criminal behavior have distinct meanings. A fifth myth holds that confident expert testimony is the same as strong evidence. Confidence, especially when presented without methods and uncertainty, may conceal the gaps an informed reader most needs to inspect.

Frequently asked questions

Can forensic psychologists predict who will commit a crime?

They may assess risk using evidence from specified populations and decision settings, but the resulting estimates have measurable error and do not identify an individual's future conduct with certainty. Accuracy depends on the outcome, time horizon, population, base rate, and instrument. A prediction should be communicated as conditional and probabilistic rather than as a verdict.

Can a forensic psychologist prove that a mental disorder caused a crime?

Psychological evaluation can examine symptoms, functioning, chronology, and possible contributing mechanisms. Establishing a specific causal explanation requires evidence and assessment of competing accounts. Whether those findings satisfy a criminal-responsibility defense is a legal matter determined under the controlling jurisdiction's rules.

How accurate are forensic psychology assessments?

There is no single accuracy rate for the discipline. Different tasks measure different outcomes and use different methods. Published risk-tool performance is often moderate and varies by setting; research on agreement in competency or sanity opinions addresses another kind of reliability. Assessment quality must be judged against the actual question, method, population, and evidence base.

Are violence-risk assessment tools objective?

A structured algorithm can apply a defined calculation consistently, but the input records, variable choices, model design, validation samples, threshold, and interpretation all involve human decisions. Structured professional judgment adds explicit case formulation and management context. Both require external validation, error analysis, and fairness review.

What is the difference between a risk factor and a cause?

A risk factor is statistically associated with a greater likelihood of an outcome in defined data. A causal factor contributes to producing an outcome under a credible causal account. An association alone does not establish a particular causal pathway, and even a genuine population-level cause does not uniquely explain an individual's action.

Can a psychologist testify about someone they have never examined?

In some circumstances an expert may give a records-based or research-based opinion when legally permitted and scientifically appropriate. The scope depends on the available records, referral question, and jurisdiction. The APA specialty guidelines recommend identifying the information base and clarifying limitations when direct examination is unavailable.

Does an expert decide whether a defendant is competent, insane, or guilty?

Experts may assess and explain legally relevant psychological functioning and offer appropriately supported opinions. Courts and other authorized legal actors make findings under the governing standards. The allocation of responsibilities differs by jurisdiction, and guilt cannot be established by a psychological test score.

Does scientific peer review make a forensic opinion admissible in court?

Peer review is relevant to evaluating a method, but admissibility depends on the applicable evidentiary rules and the way the method was used in the case. Under U.S. federal Rule 702, sufficient facts, reliable principles and methods, reliable application, and helpfulness are central requirements.

What should lawyers, journalists, and readers look for in a forensic claim?

Look for a precisely stated question, explicit sources, validated methods, well-defined outcome measures, evidence of corroboration, treatment of rival explanations, disclosed uncertainty, and clarity about what belongs to scientific inference versus the court's legal determination. Claims expressed as unconditional certainty warrant especially close scrutiny when the underlying research is probabilistic.

References

American Psychological Association. (2013). Specialty guidelines for forensic psychology. American Psychologist, 68(1), 7–19. APA official guideline.

Au, V., et al. (2026). A systematic review of the predictive validity of the VRAG-R. Journal of the American Academy of Psychiatry and the Law, 54(1), 59–70. https://doi.org/10.29158/JAAPL.260001-26.

Fazel, S., Singh, J. P., Doll, H., & Grann, M. (2012). Use of risk assessment instruments to predict violence and antisocial behaviour in 73 samples involving 24,827 people: Systematic review and meta-analysis. BMJ, 345, e4692. https://doi.org/10.1136/bmj.e4692.

Federal Judicial Center & National Research Council. (2011). Reference manual on scientific evidence (3rd ed.). National Academies Press. https://nap.nationalacademies.org/catalog/13163/reference-manual-on-scientific-evidence-third-edition.

Federal Rule of Evidence 702. (2023 amendment). Testimony by expert witnesses. https://www.law.cornell.edu/uscode/text/28a/courtrules-Evid/article-VII/courtrule-702.

Garb, H. N., & Wood, J. M. (2019). Methodological advances in statistical prediction. Psychological Assessment, 31(12), 1456–1466. https://doi.org/10.1037/pas0000673.

Guarnera, L. A., & Murrie, D. C. (2017). Field reliability of competency and sanity opinions: A systematic review and meta-analysis. Psychological Assessment, 29(6), 795–818. https://doi.org/10.1037/pas0000388.

Murrie, D. C., Boccaccini, M. T., Guarnera, L. A., & Rufino, K. A. (2013). Are forensic experts biased by the side that retained them? Psychological Science, 24(10), 1889–1897. https://doi.org/10.1177/0956797613481812.

National Research Council. (2014). Identifying the culprit: Assessing eyewitness identification. National Academies Press. https://doi.org/10.17226/18891.

Ogonah, M. G. T., Seyedsalehi, A., Whiting, D., & Fazel, S. (2023). Violence risk assessment instruments in forensic psychiatric populations: A systematic review and meta-analysis. The Lancet Psychiatry, 10(10), 780–789. https://doi.org/10.1016/S2215-0366(23)00256-0.

Ramesh, T., Igoumenou, A., Montes, M. V., & Fazel, S. (2018). Use of risk assessment instruments to predict violence in forensic psychiatric hospitals: A systematic review and meta-analysis. European Psychiatry, 52, 47–53. https://doi.org/10.1016/j.eurpsy.2018.02.007.

Singh, J. P., Desmarais, S. L., & Van Dorn, R. A. (2013). Measurement of predictive validity in violence risk assessment studies: A second-order systematic review. Behavioral Sciences & the Law, 31(1), 55–73. https://doi.org/10.1002/bsl.2053.

U.S. Department of Justice. (2024). Uniform language for testimony and reports. https://www.justice.gov/olp/uniform-language-testimony-and-reports.

Psychological Encyclopedia