Partner violence: three blind spots that keep the system in place
The words we use to name violence decide whether we fight it or sustain it. That is not a poetic claim. It is the hardest lesson from thirty years of research into decision-making and framing[1]: language is not the wrapping of a problem, language is what activates the problem in our heads. And when you stand in a room full of professionals working with partner violence every day, you notice how often we are still blind to that.
I am not an expert in partner violence. I am a behavioural designer. I study how language, framing and system incentives decide what people actually do. I recently spoke to a Belgian partner-violence centre and put this to them: the way we talk about partner violence has itself become part of the problem. We say "domestic violence" and it sounds tame. We say "things got out of hand" and the perpetrator disappears from the sentence. We say "crime of passion" and it turns into a love story.
Those words activate scripts in our heads. In perpetrators, victims, professionals and bystanders. Scripts that normalise violence rather than break it. Anyone serious about this problem has to recognise three kinds of blindness that quietly keep the behaviour in place.
The three blind spots: blindness to how our brain actually thinks and decides (System 1 structurally beats System 2), blindness to what really drives behaviour (behaviour is always an attempted solution to an underlying need) and blindness to the system (rules, goals and feedback loops decide what individuals can actually do). Anyone intervening on only one layer gets the other two as opponents.
The first blind spot: how our brain thinks about violence
Daniel Kahneman simplified it into two systems[2]. System 1 is fast, automatic and associative: it labels what it hears and instantly attaches a feeling and an action route to it. System 2 is slow, analytical and energetically expensive: it can override System 1, but only if we consciously force it to. Under pressure, System 2 almost always loses. And in the world of partner violence, we are continuously under pressure.
What that looks like in practice was visible years ago in the Dutch public health body Pharos and its work on female genital cutting. The first campaigns spoke of "barbaric practice", "crime", "complicity". The moral position was clean. The effect was disastrous. Professionals trying to work with the affected communities lost access entirely, because the framing had condemned their conversation partners in advance. Only when Pharos shifted the framing to "our professionals must be able to do their work to protect as many girls as possible" did doors open. Nothing about the problem changed. Everything about access to the problem changed.
The same mechanism plays out in partner violence. The moment the label "partner violence" is dropped, System 1 jumps to red alert. The professional feels: if I take a step here I walk straight into a storm. The neighbour feels: this is none of my business. The GP feels: this is beyond my consultation. Precisely in that second, when System 1 takes over, the brain reaches for the words that soften the discomfort.
That is why "relationship problem" stays popular. It activates the symmetrical couples-therapy script: there are always two sides, they need to work it out together. That is why "domestic violence" persists. It sounds familiar, domesticated. That is why "things got out of hand" is so dangerous. It turns people into passive objects of an event nobody actually carried out.
The linguist Julia Penelope made this pattern famous[3]. You can peel a sentence about violence back step by step until the perpetrator has vanished. "John beat Mary." Then: "Mary was beaten by John." Then: "Mary was beaten." Then: "Mary is a battered woman." And finally: "Mary has a history of battering." Five sentences in, John is gone from the mind of anyone reading the sentence. The grammar has done his work for him.
The same erasure happens on a different axis with "crime of passion" and "tragic relationship". Violence becomes a tragedy in which two equals are stuck, driven by emotion nobody could channel. Romantic and dramatic. Legally and politically completely disarming.
A better System 1 framing is possible. Not as a slogan, but as a precise description that makes the brain run the right action script. Instead of "a man with anger problems": "a man who systematically controls his wife. Who she sees, what she spends, where she is allowed to go. The hitting is the part you see." Instead of "she has fled to the women's refuge": "he has turned her home into a place where she is not safe. He is the one who has to leave." Instead of "luckily she is now away from him": "she has left him. The next six months are the most dangerous of her life."
None of these sentences is more dramatic than the old one. They are simply more accurate. They let System 1 run the right film.
The second blind spot: behaviour is always an attempted solution
The second blind spot is more dangerous because it disguises itself as compassion. It sits in our assumption that we can change perpetrators of partner violence by confronting them with the awfulness of what they do. That is inside-out thinking: trying to change someone else's behaviour from inside our own moral framework. We want him to understand the damage. We want him to feel shame. We want him to face the worst part of himself.
In most cases the opposite happens. The perpetrator's judgemental brain goes on the defensive. He minimises. He externalises. He walks out of the conversation. The system shrugs: that's how they are.
Outside-in thinking asks for something uncomfortable. It asks you to suspend your moral judgement long enough to ask: which underlying need is this person trying to meet through this behaviour? Which better solution to the same need could move him? This is not excusing. It is the same logic we apply to anorexia. We know the eating disorder is, at the symptom level, harming the patient. We also know that at a deeper level it is an attempt to keep control in a life that feels uncontrollable. Anyone intervening only at the symptom level (eat more, or you'll die) is working against, not with, the patient's brain.
The same principle applies to perpetrators of partner violence. The hitting is the symptom. The underlying need is almost always an attempt at control: over the life, over the partner, over a threatened loss of status or certainty. The question then is not whether that is acceptable — it is not — but which better solution to the same need actually works.
Three entry points consistently work in practice.
The first is being a better example to the children. Almost every perpetrator says, at some point: "I don't want my kids to grow up to be like me." That sentence opens a door no moral reproach ever opens.
The second is bringing down the chronic stress that is making him ill. Many perpetrators live in a permanent state of physical and mental dysregulation. The violence is the release valve, not the cause. Anyone who can help relieve the pressure cooker gains access to the pressure cooker itself.
The third is preventing him from losing everything he is desperately trying to hold on to. This may be the strongest. The 2022 case of the former Ajax football director Mark Overmars, who lost his job overnight after sexually inappropriate behaviour came to light, illustrates it sharply. The moment the club's board confronted him, what happened was not what public outrage demanded — a moral self-examination. What happened is what always happens: a man who realised in a fraction of a second that everything he had built was about to slip out of his hands. Career, club, family, reputation. At that moment his Job To Be Done stopped being "carry on" and became "stop myself losing everything".
The same mechanism works in the workplace and at home. "If this is true, you'll be in the national press for the wrong reasons. You're putting your club, your family, your whole life at risk. I'm not going to tell you whether it's true or not. What I am telling you is that if it's true, you need to fix this now, and we can help you with that." That is not a moral verdict. It is a precise framing that makes the costs and benefits of the behaviour visible to the perpetrator himself. That framing decides whether someone shuts down or picks up the phone.
You can read about it. But what if you could apply it yourself, to clients, colleagues, citizens or stakeholders?
As Europe's #1 academy in Behavioural Design, we train working professionals in analysing, predicting and influencing behaviour, live, online or in-house. Based in Amsterdam, with more than 10,000 alumni from London to Sydney and from Singapore to New York.
The third blind spot: the system that keeps violence indoors
Anyone who looks only at the brains of perpetrators and victims misses the most important thing. Partner violence is not just a behaviour problem. It is a system. A system that, unintentionally but effectively, does precisely what we don't want.
Three sub-systems operate inside that system at the same time.
System A is the relationship itself. An intimate dyad in which violence functions as an instrument to maintain control, status or emotional regulation. A closed system with its own logic, its own feedback loops (escalation, reconciliation, hope, isolation) and its own private language outsiders cannot read.
System B is the social environment. Family, friends, neighbours, work, sports clubs, places of worship. Its dominant goal is social cohesion and the avoidance of discomfort. It sees more than it admits to seeing. Its mechanisms for looking away are not malicious; they are protective of the group.
System C is the institutional field. Police, domestic violence services, GPs, women's refuges, child protection, prosecutors, probation, insurers. This system claims protection and recovery as its goal, but in practice often optimises for something else: throughput, file closure, inter-agency agreements, lawfulness.
The system works extremely well for everyone with an interest in keeping violence private. It works for legislation that places family autonomy above safety, because then the state does not have to intervene structurally. It works for media outlets that frame partner violence as relationship drama, because then it is a personal story rather than a public problem. It works for policy that funds prevention as a temporary project, because responsibility can be handed off to a grant that runs out in two years. And it works, unintentionally but effectively, for all of us. Because as long as it is a private matter, we don't have to get involved.
The system is not doing too little. It is doing exactly what it was designed to do: keep violence indoors. Literally and figuratively.
Four feedback loops keep this equilibrium in place.
Loop 1: isolation reinforces dependence. The more violence, the more isolation from family and friends. The more isolation, the more dependent the victim becomes on the perpetrator for confirmation and reality-testing. The greater that dependence, the easier it becomes to normalise the violence. This loop runs entirely inside System A and is enough on its own to hold someone for ten years.
Loop 2: looking away normalises looking away. When the neighbours do nothing, that confirms for each individual neighbour that it cannot be that bad. Pluralistic ignorance as collective self-deception. Every neighbour reads the silence of every other neighbour as evidence that nothing is wrong.
Loop 3: report-without-consequence undermines reporting. A victim reports, something half-happens (a conversation, a warning, a file) and then the perpetrator goes home. The violence escalates, because the perpetrator now knows that reporting is harmless to him. Next time, the victim does not report. This loop is catastrophic: every weak intervention makes the next report less likely and the next escalation more likely.
Loop 4: paperwork crowds out contact work. Every new incident produces more registration requirements. More registration means less time for direct contact. Less contact means worse visibility into what is actually going on. Worse visibility means more incidents that escalate. More incidents lead to more registration requirements.
These are not anecdotal problems. They are structural patterns that override every individual professional, no matter how well-intentioned. Training people alone changes none of it.
Leverage points: how to tip this system
Donella Meadows produced a hierarchy of places where you can intervene in a system[4]. At the bottom: adjusting parameters — raising a budget, expanding headcount. Limited effect. In the middle: changing the rules — who has which authority, who carries which burden? Powerful effect. Higher: changing the goals — what is the system steering towards? At the top: changing the paradigm — which worldview underlies the system?
For partner violence, three leverage points stand out.
At rule level the burden of proof and action currently sits almost entirely with the victim. She has to file the report. She has to give a statement. She has to move out. She has to take the children. The perpetrator, in many cases, stays in the home. A rule change that flips that relationship — when violence is reported, the perpetrator leaves the home by default until the situation is clarified — operates at a different level than more staff or more money. Austria has had this rule, the Wegweisung, since 1997, and the effects are documented[5]. It changes who the system burdens.
At goal level the most painful question is: what are services actually steering on? On case throughput, processing time and closed files. Then the system becomes a production line for completed cases. Shift the goal to recovery of autonomy by the victim and demonstrable behaviour control by the perpetrator, and without moving a single wall the entire system tips. It is the same logic as a hospital being acquired by an investment fund: the building is identical, but the steering organ now steers on something different and every incentive shifts with it. Here it is the inverse move: not steering on returns, but steering on outcomes for people.
At paradigm level sits the hardest shift. As long as we frame perpetrators as "monsters" or "sick men", we keep them out of reach and confirm their own self-image that they have to keep a secret. A workable paradigm is different: this is a man on the verge of losing everything he is desperately trying to hold on to, and we are the only place he can prevent that from happening. Not by sympathising with his behaviour, but by precisely framing his own interest.
If I were minister of health
The question every conversation about systems change ends with: what would you actually steer on? Which measure would the minister of health introduce tomorrow to make visible whether anything is working?
Four candidates do not survive scrutiny.
Number of repeat incidents sounds reasonable, but under-reporting destroys it. Successful service can drive reports up because trust grows. Failing service can drive reports down because the victim never calls again. The measure steers the system towards silence.
Number of femicides per year carries moral urgency, but is statistically too rare to steer on. Useful as a background indicator. Useless as an operational KPI.
Victim satisfaction sounds modern. It is dangerous. A victim in a coercive control situation is conditioned to respond positively to any external interference, because she has learned that complaining has consequences. Satisfaction measures a mixture of comfort, habituation and fear, not safety.
"Stopping the violence" itself, reported by the victim, remains dependent on the perpetrator and on the victim's willingness to speak. Too much noise, too much distortion.
The strongest candidate is recovery of autonomy by the victim, measured through a small number of concrete indicators. Does she have her own income or access to her own money? Does she decide her own daily schedule? Does she maintain self-chosen social contacts? Does she have her own home with her own key? Does she have her own GP record he cannot access? Measured at six and eighteen months from first contact.
This measure is interesting for three precise reasons.
It measures the inverse of what the perpetrator produces. The perpetrator produces isolation, financial dependence, control over the day, restriction of social contacts. Autonomy is the direct inverse. If autonomy rises, the violence system is structurally retreating, even if individual incidents continue.
It is measurable without depending on the perpetrator. Unlike "has the violence stopped", this measure does not have to wait for his behaviour change. It can rise while he has not yet changed, simply because she has gained ground.
It does not place the burden on the victim. Autonomy is something professionals can facilitate, not something the victim has to "achieve". Services are held to account for how much room they have helped create, not for how many steps the victim has taken.
Anyone introducing this measure shifts six things at once.
Professionals start asking questions they currently do not ask. Not "how is the situation at home", but "who has the PIN to your bank account" and "who have you spoken to this week other than him".
The system reallocates resources. Financial independence and housing become core business rather than aftercare.
A large share of cases that "succeed" by the old metric will "fail" by the new one. That is painful and will trigger resistance, but it is information the system currently lacks.
The pressure on the victim drops. She is no longer asked to choose between "staying" or "leaving" as a service performance metric. Every step towards autonomy counts, including inside the relationship. That is psychologically more realistic and strategically smarter: building autonomy is often the precondition for ever being able to leave.
Responsibility for stopping the violence returns to the perpetrator and to enforcement, where it belongs. Services are no longer judged on something over which they have no power, but on something over which they do have influence: her room to manoeuvre.
A feedback loop appears that does not currently exist. Professionals rarely hear today what their work has actually delivered; the case closes and the outcome disappears. Measuring at six and eighteen months creates a learning loop that reconnects professionals with the meaning of their work. That is a second-order gain that touches the entire field.
A complementary candidate belongs alongside it: demonstrable behaviour control of the perpetrator, independently monitored. Not via the victim, but via probation, treatment observation, contact with the children, behaviour at work. In most countries this is currently not available for the majority of perpetrators. That is itself a finding: the outcome cannot be measured because the perpetrator is out of sight.
Three leverage points, one problem
Three blind spots are three points of leverage. Anyone working on only one layer gets the other two as opponents. Anyone changing the language while leaving the system in place produces better sentences without better outcomes. Anyone reforming the system without shifting the framing inside people gets new rules that are quietly bypassed in practice. Anyone treating perpetrators without understanding their Job To Be Done breeds cynicism in the professionals who watch it fail.
The minister of health who has to do one thing tomorrow does this: stop holding services to account for the number of files closed, and start holding them to account for the degree to which victims visibly recover their autonomy. With that, the system is steering on something it can influence, rather than on something over which it has no power. With that, the centre of gravity shifts away from looking away and privatising, and towards protection and recovery.
The words we use to name violence decide whether we fight it or sustain it. The incentives we hold professionals to decide what behaviour the field actually produces. The role we give perpetrators in our frames decides whether they move or shut down. Three leverage points, one problem. Anyone willing to move all three at once changes what until now has looked unchangeable.
Frequently asked questions
What is the difference between domestic violence and partner violence?
The terms describe largely the same phenomenon, but "domestic violence" activates a tame, neutral association in System 1 and hides who is the perpetrator and who is the victim. "Partner violence" names more sharply that this is violence inside an intimate relationship, with a perpetrator and a victim. For anyone who wants to intervene effectively, more precise language matters because words activate the scripts in our heads that determine what professionals, bystanders and perpetrators do next.
What does outside-in thinking mean when working with perpetrators of violence?
Outside-in thinking is the discipline of suspending your own moral judgement long enough to ask which underlying need the perpetrator is trying to meet through his behaviour. Almost always it is an attempt at control or the avoidance of loss. Effective interventions offer a better solution to the same need: being a better example to the children, lowering the chronic stress that is making him ill, or preventing him from losing everything he is desperately trying to hold on to. This is not excusing the behaviour. It is the precise behavioural analysis that decides whether someone shuts down or starts moving.
What is the Wegweisung and does it work outside Austria?
The Wegweisung is an Austrian measure introduced in 1997: when violence is reported, the perpetrator is removed from the home, not the victim. This shifts the burden of proof and action from victim to perpetrator and enforcement. Research shows a substantial drop in repeat incidents. Several countries have adopted variations, including the Netherlands' temporary restraining order from 2009, but enforcement and follow-through vary widely. Structural use requires institutional support around the perpetrator, the children and enforcement.
Why would autonomy be a better outcome measure than "stopping the violence"?
Stopping the violence depends on the perpetrator and is measured through victim reporting, with severe under-reporting. Autonomy measures the inverse of what the perpetrator produces: financial independence, self-chosen social contacts, an own home, an own daily schedule. That measure can rise even before he changes, does not have to wait on his behaviour, and does not place the burden on the victim. Services held to that measure structurally start asking different questions and deploying different resources.
How can bystanders intervene more effectively?
Bystanders are trapped in a feedback loop of pluralistic ignorance: because nobody acts, everyone infers that nothing is wrong. The intervention is not a moral appeal but the disruption of that loop. Effective action requires pre-designed routines: one specific question to the possible victim in a safe context, a report to a domestic violence service, or a calibrated framing conversation with the perpetrator about what he stands to lose. Pre-designed routines are necessary because in the moment itself System 1 almost always chooses to look away.
1.5 minutes of influence
Join 10,000+ readers · Free · Unsubscribe anytime