Chapter 14: Assessing Violence Risk

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement not replace the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

Imagine you're a psychologist.

You've got a clipboard in your hand, a patient is sitting right across the desk from you, and you have this impossible choice to make.

Right, it's a heavy situation.

Exactly.

Because if you check the box labeled low risk, this person walks out the door and they might severely hurt someone in your community.

But if you check high risk, you are literally stripping away their fundamental human right to freedom and locking them away.

It is arguably the most high stakes question in the entire legal system.

I mean, we are talking about life altering decisions based on essentially trying to foresee human behavior.

And for a really long time, the fields of law and psychology really struggled to get it right.

So if you are diving into forensic psychology for the first time, consider this the deep end.

Oh, absolutely.

It's complex stuff.

Consider this your one -on -one last minute lecture tutoring session.

We are taking a deep dive into chapter 14, assessing violence risk from the Handbook of Forensic Psychology, fourth edition.

It's a fantastic chapter, really lays out the groundwork.

Yeah, and our mission today is to walk you chronologically through the foundations of this field.

We're going to look at the notoriously flawed early research, explore how modern clinical tools evolved, and then finally examine how a judge and jury actually digest a psychological evaluation in a courtroom.

Sounds like a solid plan.

Okay, let's unpack this.

Because predicting human behavior isn't about, you know, looking into a in the 1970s, the scientific community thought they were actually terrible at this.

Yeah, the 1970s was just a period of massive self -doubt for the field.

Back then, they didn't even use the term violence risk assessment.

They evaluated something they called dangerousness.

Dangerousness.

I mean, that sounds so permanent, like having brown eyes or being left -handed.

It makes it sound like a personality trait.

You just can't escape.

That's spot on.

And the language shift is vital for that exact reason.

You see, dangerousness implies an either -or state like you are or you aren't.

Shifting the terminology to violence risk assessment acknowledges that risk is dynamic.

It fluctuates depending on, you know, environmental conditions, interpersonal triggers, and time.

Which makes way more sense.

But before that shift happened, the field was basically rocked by a series of studies claiming mental health professionals were completely inept at predicting future violence.

The most infamous one was a 1972 study by Kozol, Boucher, and Garofalo.

Oh, I recalled the headline statistic from that era.

It claimed clinicians were wrong two -thirds of the time when they predicted someone would be violent.

I mean, how did they even come up with a number that terrible?

Well, the researchers evaluated 592 males who had been convicted of assault of offenses.

The clinical team evaluated them and eventually classified 386 of them as not dangerous, which led to their release.

Okay, so a pretty big chunk.

Right.

But the crux of the study focuses on a specific group of 49 patients.

The clinicians classified these 49 individuals as still dangerous and recommended they stay confined.

Let me guess.

The legal system had other plans.

Exactly.

Legal authorities ignored the clinical staff and released those 49 patients anyway.

Over a five -year follow -up, 35 % of those 49 dangerous patients committed a serious crime, compared to only 8 % of the not dangerous group.

Okay, wait.

Doing the math there, if 35 % of the people the clinicians flagged as dangerous actually committed a crime, that means 65 % of them did not.

Yep.

That's it.

So that is where the wrong two -thirds of the time headline comes from.

But honestly, evaluating these clinicians based on patients who were released against their advice seems incredibly unfair.

It really is.

It is like calling a meteorologist a total fraud because they advised you not to go sailing in a massive hurricane, you took the boat out anyway, and you just happened to survive.

That doesn't mean the storm warning itself was scientifically wrong.

What's fascinating here is how the chapter totally deconstructs that flawed methodology.

First, those 49 patients were a highly unrepresentative sample.

The legal authorities didn't just randomly release the absolute worst offenders.

Oh, so they cherry -picked who to release.

Basically, yeah.

They released individuals whose dangerousness was already considered borderline or uncertain by the parole board.

If the truly high -risk, unequivocally violent patients had been released, the accuracy rate of the clinicians' predictions would likely have skyrocketed.

That is a huge caveat.

And second, we have to look at the reality of crime statistics.

Research suggests only about 20 % of serious crimes ever lead to an arrest.

Oh, wow.

Meaning a significant chunk of that 65 % who supposedly didn't offend might have actually been highly violent out in the community.

They just never got caught by the plate.

Exactly.

The data was missing the undiscovered true positives.

But really, the most crucial flaw was a fundamental misunderstanding of what the clinicians were actually doing in those evaluations.

What do you mean?

Well, they were not predicting violence with absolute psychic certainty.

They were making a conservative medical and legal recommendation not to release someone because that person posed a significantly elevated risk under certain conditions.

But the researchers tested them as if they were issuing an ironclad guarantee.

That makes a lot of sense.

Were there other studies piling onto this narrative?

Because I know cases where large hospitals were shut down or populations were moved can really skew the data.

Yeah, the Backstrom and Dixon studies are perfect examples of that.

These involve judicial decisions where hundreds of patients who had been deemed too dangerous for regular civil hospitals were suddenly transferred or released en masse.

Let me guess, a lot of them were fine.

Right.

Follow -up studies showed very few of them actually acted violently afterward.

Critics used this to claim psychiatrists couldn't predict danger.

But the catch is, those original confinements were never based on individualized clinical assessments in the first place.

Oh, so it was just a blanket label.

Exactly.

Yeah.

The literature calls them administrative or political predictions.

They were global institutional assumptions made by the system.

Not careful.

One -on -one scientific evaluations.

So if human intuition is perceived as too biased and courts demand absolute transparency when taking away someone's freedom, how do you remove the human element?

It seems like psychologists had to scramble to prove their worth here.

They absolutely did.

And it led to a massive evolution.

The field basically split into three distinct models of assessment to try and solve this problem.

The first model is essentially what they were already doing, which we now call unstructured clinical judgment.

Unstructured.

So just a standard interview.

Yeah.

It's an informal impressionistic conclusion reached by a human evaluator relying purely on their own professional experience and clinical intuition.

You know, on a human level, I can see the appeal of that.

You are sitting across from a unique individual with a complex life story.

Unstructured judgment allows you to highly individualize the assessment and look at the specific, perhaps rare, quirks of the person in front of you.

It definitely fosters individualized case conceptualization.

The fatal flaw, though, is that it lacks rules.

There is zero transparency into how the clinician weighed the information.

Like two doctors could look at the same person and see totally different things.

Precisely.

One evaluator might focus heavily on a patient's history of trauma, while another might completely ignore it and focus only on a past arrest.

They might even rely on factors that have no actual scientific association with violence.

Over time, this produces decisions with very low reliability.

I imagine the courts hate that.

From a legal standpoint, it is an absolute nightmare.

When an examinee's fundamental rights are at stake, an entirely unstructured, subjective decision just cannot be adequately reviewed or challenged by a court.

So the pendulum swings entirely in the opposite direction.

If the human mind is too messy, you turn to math.

That brings us to the second model, actuarial prediction.

This sounds like life insurance mathematics.

It's very similar.

You use a strict statistical equation, an actuarial table, and just punch in the numbers.

Tools like the VRAG or the STATIC -99 are prime examples.

It is highly reliable.

And honestly, courts initially love the transparency.

Because it's just a formula.

Right.

You score a five, you go into the moderate -risk bin.

You score a nine, you go into the high -risk bin.

But the actuarial approach has severe scientific drawbacks, the largest being sample dependence.

Sample dependence, meaning who they tested it on.

Exactly.

These instruments are built by finding risk factors that statistically predicted violence in one specific derivation sample.

Let me see if I understand this.

A derivation sample means the original group of people they studied to build the math formula.

So if you build an actuarial tool by studying only men in a maximum -security prison in the 1990s, the math is perfectly optimized for that exact demographic.

You nailed it.

And when you take those strict statistical weights and apply them to an entirely new group of people, say, female offenders in a modern community clinic, the predictive accuracy almost always degrades.

That makes sense.

That process is called cross -validation, and actuarial tools often really struggle with it.

Furthermore, strict actuarial tools forbid you from considering risk factors that aren't on the test.

Wait, really?

You just have to ignore them?

Yep.

For example,

if a tool doesn't include a variable for treatment noncompliance, and surprisingly,

the VREU does not urge there, you are mathematically forbidden from factoring it in.

You could have a patient sitting in front of you who violently attacks staff every time they refuse their medication, but if the formula doesn't ask about it, you literally can't score it.

Wow.

It is like following a strict baking recipe.

An actuarial recipe might work perfectly in a commercial kitchen at sea level, but if you take that exact same recipe, change absolutely nothing, and try to bake it in a humid home kitchen in the mountains, it completely flops.

You need the ability to adjust based on the environment.

That's a great way to put it.

Actuarial tools also completely ignore dynamic risk factors, things that change, like acts of substance abuse or acute psychotic symptoms.

They rely almost entirely on static historical facts.

Right, because you can't change the age at your first arrest.

Exactly.

Plus, researchers point out a massive statistical flaw here.

Group -level probabilities break down at the individual level.

Saying a group has a 44 % chance of recidivism comes with statistical confidence intervals so wide that it renders the specific percentage almost meaningless for the actual singular person sitting in your office.

Which naturally leads us to the modern gold standard, the third model, Structured Professional Judgment, or SPJ.

If unstructured is just guessing, and actuarial is a rigid recipe, SPJ sounds like a master chef using a flexible tasting menu tailored to the diner.

I like that analogy.

Structured Professional Judgment tools like the HCR20 for general violence, the SUVRY for youth, or the Sarov or Spousal Assault bridge the gap, they use logical item selection rather than strict mathematical formulas.

So they aren't just relying on one derivation sample anymore.

Right.

The creators reviewed all the scientific literature and selected risk factors that have broad empirical support across many different samples, not just one.

Crucially, SPJ does not use numeric cutoffs.

Wait, you don't add up points to get a percentage score at the end?

No, not at all.

The clinician assesses the presence of the risk factors.

But more importantly, they assess the relevance of those factors to the specific individual.

Based on that comprehensive picture, the clinician classifies the summary risk as low, moderate, or high.

So it brings the human back into the loop.

Exactly.

But a guided human.

SPJ heavily incorporates those dynamic risk factors we just mentioned, which allows the clinician to actually recommend ways to manage the person's future behavior rather than simply issuing a grim prediction.

Okay.

So what does this all mean?

We have these three models.

Unstructured, actuarial, and structured professional judgment.

When we look at the comparative research, which one actually wins in the real world?

A landmark study here is the 1993 LID study.

They looked at psychiatric emergency room patients.

The study found that when clinicians use their judgment to assess potential violence, they actually perform significantly better than chance for the male patients.

Okay.

But did they perform better than chance for the female patients, too?

No, actually.

They missed female violence entirely.

Clinicians drastically underestimated the risk posed by women, spending much less time asking them about potential violence compared to men.

Wow.

Yeah.

The study also highlighted an interesting tension.

A simple three -item actuarial screen outperformed the clinicians for predicting any community violence.

However, the clinicians matched the actuarial tools for predicting serious violence, which is the threshold that actually justifies civil commitment.

Wait.

If the actuarial screen outperformed them on general violence, doesn't that mean the math wins?

Not necessarily.

Because of how the data was gathered, the actuarial data, like recent heavy drug use, was collected by researchers after the patients were discharged into the community.

Oh, I see.

You cannot reliably gather that kind of verified data in a chaotic emergency room setting while the patient is actively in crisis.

That makes total sense.

The tool is only as good as the data you can punch into it in the moment.

Exactly.

And if we connect this to the bigger picture, the data today is incredibly clear.

Recent meta -analyses, like Guy in 2008 and Singh and colleagues in 2011, synthesize dozens of independent studies.

And what do they find?

The evidence proves that SPJ summary risk ratings, those low, moderate, high clinical judgments, are just as accurate, if not more accurate, than actuarial methods.

And both of them vastly outperform unstructured clinical judgment.

The text even mentions that the leading SPJ tool, the HCR20, adds unique predictive power over the PCLR.

Can you clarify what the PCLR is for our listeners and why that matters?

Sure.

The PCLR is the psychopathy checklist revised.

It is the gold standard for measuring psychopathic traits, things like grandiosity, lack of empathy, and manipulative behavior.

Historically, psychopathy was considered one of the strongest predictors of violence.

Right.

But the research shows that if you use the comprehensive HCR20, it improves your prediction model significantly.

The PCLR, on the other hand, doesn't add anything to the table that the HCR20 hasn't already caught.

Okay, so the data proves SPJ is the winner.

But theory is one thing, right?

Practice is another.

How does a clinician actually take this SPJ model and apply it to a human being sitting across from them?

Because I know the chapter preserves a very strict logical flow for how this comprehensive assessment is conducted.

It's a rigorous six -step process.

Let's imagine a hypothetical patient.

We'll call him John.

Step one is gather critical information.

You cannot just ask John if he plans to be violent.

He might lie or lack insight.

So you do some digging?

Exactly.

You need meticulous history from multiple collateral sources, police reports, victim statements, daily hospital logs, and previous psychiatric files.

Step two is identify risk factors.

You use the standard list provided by the SPJ tool you selected.

For the HCR20, you systematically check for historical, clinical, and risk management factors present in John's life.

Okay, and step three is evaluating relevance.

This sounds like the core of the clinical work.

How do you decide what is actually relevant to John?

This is where you apply the decision theory of violence.

You have to figure out the specific mechanics of why John chooses violence.

Think of a car.

You are looking for motivators, which act like the gas pedal.

Why did John want to hurt someone?

Maybe he feels deeply disrespected.

Okay, I like this analogy.

Then you look for disinhibitors, which are like cut brake lines.

Why doesn't he care about the consequences?

Maybe he has antisocial traits and doesn't care about going to jail.

And what's the third piece?

Finally, you look for destabilizers, which act like a blown tire.

This could be an active psychotic episode or severe intoxication that makes John lose control of the vehicle entirely.

That makes it so clear.

Once you understand the engine of John's violence, you move to step four, which is develop scenarios of violence.

Now, I have to push back here.

The text says the clinician develops a repeat scenario, a best case scenario, a worst case or doom scenario, and a twist scenario where the violence evolves.

Honestly, it sounds a bit like a Hollywood writer drafting alternate endings to a movie.

How does a clinician keep this grounded in science rather than just writing fiction?

This raises an important question.

And the key word the field uses here is plausible.

You aren't writing creative fiction.

You are forecasting based on hard evidence.

Plausible being the operative word.

You use the psychological theory, John's past behavior, and the specific risk factors you just identified to prune away the implausible scenarios.

If John has only ever committed violence against intimate partners when heavily intoxicated, a scenario where he soberly attacks a random stranger at a bank is entirely implausible.

You focus only on what reasonably could happen given his established history.

Got it.

So step five is consider management plans based on those plausible scenarios.

If we know the worst case scenario involves alcohol and his ex -partner, the management plan focuses on intensive monitoring, substance abuse treatment, and strict no -contact orders.

Spot on.

And finally, step six is communicate findings where you step back, look at the whole picture, and assign that final summary risk rating of low, moderate, or high risk.

Yes, and the chapter also highlights crucial additional clinical considerations during this For instance, when it is safe to do so, clinicians should conduct stress interviews.

Stress interviews.

Like try and make them angry.

Sort of, yeah.

A violent patient might seem perfectly calm when unchallenged in a quiet room.

You have to gently test their frustration tolerance to see how they will react to the stress of a less secure community environment.

That's fascinating.

The text also emphasizes not underestimating female risk, closely analyzing future living circumstances.

Like if John is going back to a highly discordant family situation and the absolute necessity of consulting with colleagues to avoid personal blind spots.

Okay, here's where it gets really interesting.

Once the clinician finishes this rigorous six -step assessment and signs the report, those findings are handed over to the legal system.

You would think the courts would demand absolute flawless scientific perfection before locking someone up or executing them based on a prediction.

You would think so, yes.

But historically, the courts have been far more accepting of risk predictions than the scientific community itself.

The 1983 Supreme Court case, Barefoot v Estelle, is the starkest example of this.

In a capital punishment case,

a testifying psychiatrist who had never even personally interviewed the defendant, by the way, took the stand and predicted that the probability of the defendant committing future violence was 100 % and absolute.

100%.

We just spent 10 minutes talking about dynamic factors and plausible scenarios.

No genuine scientist would ever guarantee human behavior with 100 % absolute certainty.

Surely the Supreme Court threw that testimony out.

Controversially, the Supreme Court allowed it to stand.

They basically argued that they trust the adversarial system.

They believe the cross -examination process, the presentation of opposing experts, and the common sense of the jury are sufficient to sort out the shortcomings of questionable scientific testimony.

And the court has consistently upheld the use of risk assessments to restrict liberty.

In cases like Kansas v Hendricks and Kansas v Crane, the Supreme Court upheld the civil commitment of sexually violent predators.

Even after prison.

Exactly.

Even after these individuals had served their full prison sentences, the state was allowed to keep them confined in psychiatric facilities based on their mental abnormality and the predicted risk of future dangerousness.

That brings up a mechanical question, though.

How does psychological evidence even get admitted into a trial in the first place?

Who acts as the bouncer for the courtroom?

It depends on the jurisdiction's legal standards, usually either Frey or Doebert.

Think of Frey as a popularity contest within the scientific community.

Under Frey, the assessment method must be generally accepted by a meaningful segment of the relevant experts.

Okay, and what about Doebert?

Doebert, on the other hand, is like a mechanic looking under the hood of a car.

Under Doebert, the judge acts as an active gatekeeper.

They don't just ask if it's popular, they ask if it is scientifically valid, whether it has been peer reviewed, and most importantly, what the actual known error rate of the test is.

Do they treat clinical and actuarial tools differently?

Interestingly, yes.

Clinical interviews are almost always admitted as standard medical expertise under either standard.

Actuarial tools are sometimes scrutinized more heavily under Doebert because they present themselves as pure math, but they are usually admitted, especially when combined with a clinical evaluation.

What about civil commitments?

Say you haven't committed a crime, but a doctor wants to lock you in a psychiatric hospital involuntarily because you might be dangerous.

Is the standard of proof beyond a reasonable doubt like in a murder trial?

No, it is a slightly lower bar established by the Supreme Court in Addington v.

Texas.

The court ruled that you need clear and convincing evidence to justify an extended civil commitment.

So it's kind of a middle ground.

Exactly.

It's a higher standard than the preponderance of the evidence used in a standard civil lawsuit, but it acknowledges that requiring beyond a reasonable doubt for medical predictions is practically impossible.

And does the violence have to happen tomorrow to justify locking someone up?

The legal definitions of imminence are actually shifting.

In the past, under decisions like Lassard,

there was a strict requirement that the violence be imminent and based on a very recent overt act.

But that's changing.

Yeah, newer decisions like Sahar and Seltzer show a legal trend toward flexibility.

Courts are increasingly recognizing that a substantial risk based on a genuine documented possibility of future harm can be enough to justify commitment, even if the violence isn't strictly imminent.

And we really can't talk about law, psychology, and risk without mentioning Tarasov.

This is the famous duty to protect case.

Oh,

absolutely.

It is the cornerstone of clinical liability.

If a clinician determines or crucially should have determined, based on professional standards, that a patient poses a serious danger to a foreseeable, identifiable victim,

they have a legal obligation to exercise reasonable care to protect that victim.

Like calling the person in danger.

Right.

That might mean warning the victim directly or notifying the police.

Tarasov absolutely cements why accurate, comprehensive SPJ risk assessment isn't just an academic goal, it is a profound legal duty.

It really highlights the whole journey this field has taken.

I mean, we've moved from the flawed, unstructured political predictions of the 1970s through the rigid context blind mathematics of actuarial tools to arrive at the nuanced, highly effective structured professional judgment models we use today.

And the text makes it clear the field is still pushing forward.

The future directions outlined in the chapter are incredibly promising.

The field is intensely focused on refining dynamic risk, measuring exactly how risk factors change week to week so we can adjust treatment in real time.

That sounds huge for rehabilitation.

It is.

Even more exciting is the shift toward protective factors.

Tools like the SEPARO -F instrument actually systematically evaluate a person's strengths.

It looks at their social support, their coping skills, and their buffers against violence rather than just tallying up their deficits.

So why does all this matter for you, the student diving into this material?

Because understanding this deep dive equips you to look past sensationalized true crime headlines that label people as permanently dangerous.

You now possess a grounded, scientific framework.

You know exactly how the justice system actually measures risk, balances public safety, and protects civil rights using structured, evidence -based tools.

If we focus on those new developments, it leaves us with a final thought to mull over.

If our ability to assess risk continues to improve by focusing on dynamic, changeable factors, and protective traits,

will the legal system ever shift its primary focus away from strictly punishing past behavior and toward actively managing and healing future trajectories?

Now that is the multi -million dollar question at the heart of forensic psychology.

Thank you for joining us for this session.

From the Last Minute Lecture team, we wish you the absolute best of luck in your forensic psychology studies.

Keep asking the hard questions.