5 Working with offenders

Core idea

Working with offenders is forensic psychology's most operationally consequential domain. Once someone has been convicted, the system has to make two linked decisions: how dangerous is this person now? and what, if anything, will reduce that danger? Both questions look clinical but are statistical at their core. A modern risk assessment combines static factors (history that cannot change — age at first offence, prior convictions, victim type) with dynamic factors (states that can change — substance use, employment, antisocial peers, attitudes toward the offence). Treatment then targets the dynamic factors, almost always through cognitive-behavioural programmes organised around the Risk-Need-Responsivity (RNR) model: match the intensity of intervention to the offender's risk level, target only their criminogenic needs, and deliver in a style they can actually engage with.

Author's argument: Risk assessment is not prophecy. It is a structured, evidence-anchored estimate that converts a clinician's hunch into a defensible probability — and treatment is the lever that bends that probability downward, but only when the right thing is delivered to the right person.

Why it matters

Every parole decision, every sentencing recommendation, every prison-programme allocation rests — implicitly or explicitly — on a risk judgement. Get it wrong in one direction and a dangerous person is released; get it wrong in the other and someone is kept inside who would not have reoffended. Both errors are visible in the press; only one is politically punished. Forensic psychology's contribution is to make the trade-off legible, replicable, and improvable over time. The same logic explains why intuitively appealing programmes like Scared Straight can be worse than doing nothing, and why apparently boring interventions (literacy, vocational training, structured CBT) outperform charismatic ones.

The asymmetry of error

A false negative — releasing someone who reoffends violently — generates a public scandal. A false positive — detaining someone who would not have reoffended — generates a quiet injustice that almost no one sees. This asymmetry pushes systems toward over-prediction of risk, which is itself harmful: it wastes resources, fills prisons, and (because dynamic factors decay without treatment) can manufacture the very risk it predicts.

Key takeaways

Mental model

Copy sourceZoom

Practical application

Risk assessment: from gut feel to structured estimate

A structured risk assessment proceeds in roughly the same order regardless of the instrument used.

Static vs dynamic factors — and why the distinction matters

What works — and what does not

The prediction-limits problem

Treatment that respects responsivity

The responsivity pillar is the most often ignored. A high-quality CBT manual is useless to an offender who reads at a sixth-grade level, finds group sessions humiliating, or has a learning disability the programme staff have not assessed. Responsivity divides into general (CBT is the most evidence-supported delivery style) and specific (this person needs visual aids, smaller groups, a different gender of facilitator, sessions sequenced around their work schedule). The latter is where most front-line clinical skill goes.

Example

Consider Marcus, a 26-year-old serving a four-year sentence for aggravated assault outside a bar. He has two prior assault convictions, the first at age 18. He drinks heavily on weekends, works intermittently as a labourer, and lives with a partner who has been encouraging him to stop drinking.

A naive risk assessment would weight his criminal history and predict high risk. A static-only score would in fact put him in a high-risk band. But a structured assessment that incorporates dynamic factors — a stable partner, recent stretches of sobriety, a job he wants back, expressed remorse toward the victim — would moderate that score and suggest a clear intervention plan.

Under the RNR framework, his treatment plan looks like this:

• Risk (moderate-high) justifies a structured programme of perhaps eighty hours, not a brief intervention and not the maximum-intensity programme reserved for the most prolific offenders.
• Need targets two specific criminogenic drivers — alcohol misuse and a habitual cognitive script that interprets minor slights as serious threats requiring violent response. It does not target his anxiety about his father's health, even though that is real distress, because anxiety is not driving his offending. Treating non-criminogenic needs may be humane but does not reduce recidivism and crowds out the things that do.
• Responsivity notes that Marcus is verbally fluent but distrustful of authority. The plan pairs him with a facilitator who himself has lived experience and uses concrete role-play rather than abstract worksheets. Sessions are scheduled in the late afternoon so he can keep his day-release labouring job.

A year later, Marcus has not reoffended. The plan did not prove he was at the moderate-high band the tool indicated — he might have been one of the seven in ten flagged people who would not have reoffended anyway. The honest reading is that the plan reduced the probability of reoffending in someone whose history made it elevated, and that the reduction is most defensibly attributed to changing the specific dynamic factors the plan targeted.

Related lessons