“To stop violence, we must try to understand it”

In 2024, more than 265,000 victims of domestic violence—most of them women—were officially recorded in Germany. The real number is likely to be many times higher. In the context of preventing violence, however, the AI-powered web app “Lizzy” may soon become an important tool. Why? Because Lizzy makes it possible to predict the risk of imminent violence with a high degree of accuracy.

“Risk assessments play a crucial role in determining how much support affected individuals receive—and when,” explains 29-year-old doctoral researcher Ba Linh Le. A place at a women’s shelter? An electronic ankle tag for the perpetrator? Many such protective and support measures for women and children at risk are based on violence risk assessments. “The greater the danger, the more resources are approved.”

On the front line

Le is working on a Ph.D. in public health interventions at LMU’s Chair of Public Health and Health Services Research. Together with a fellow student, she founded the Berlin-based start-up “Frontline”, which developed Lizzy.

The people on the real-world “front line” are professionals who confront domestic violence directly: police officers, as well as staff in women’s shelters, safe houses and counseling centers. Working under intense time pressure, they are required to make assessments that can have far-reaching consequences. And one central problem is the considerable variation in the quality of these judgments. “Even experienced forensic psychologists struggle to predict the risk of violence,” says Le, citing studies showing that expert judgments are prone to bias. Standardized tools, by contrast, significantly improve accuracy.

Such instruments are also used in Germany, but they are not particularly reliable. Most were developed decades ago in the United States and Canada. “They work well there,” Le explains, “but the relevance of individual risk factors varies depending on the country and legal system. That’s why these tools cannot simply be mapped onto the German context.”

Typically, violence surveys are conducted with people who have already experienced abuse. In contrast, the Lizzy web app was developed on the basis of a representative German longitudinal study involving 7,400 participants. It predicts the risk of renewed violence with an accuracy rate of around 80 percent—significantly more precise than the instruments currently used in Germany.

From Research to Practice: Entrepreneurship at LMU

In spin-offs like “Frontline,” innovative ideas come to life. This is an important example of transferring academic knowledge into practice, a process that LMU actively promotes. To strengthen entrepreneurial thinking across the university community, LMU is establishing a new central unit for entrepreneurship and knowledge transfer. Various programs have been set up to support students and researchers in developing viable business ideas.

“Our goal is to catalyze innovation,” explains Dr. Philipp Baaske, LMU Vice President for Entrepreneurship. “We focus on the people behind the innovations. We challenge and support them, advise and connect them, remove obstacles and accelerate the transfer of ideas into the economy and society.”

Six questions, powerful predictions

The app is intentionally low-threshold and easy to use. A short version consists of six questions and takes about one minute to complete. A longer version generates a more comprehensive violence profile.

Trained professionals ask structured questions, and the app then analyzes the responses. Lizzy integrates physical, sexual, emotional, digital and financial forms of violence in a single overall assessment. Because direct questions about physical or sexualized violence are often not answered honestly due to fear or shame, the app relies on what are known as proxy indicators, which research has shown to be highly informative—for example, whether a partner has thrown objects at the victim, exerted financial pressure or monitored emails and text messages.

“Violence is a web”

Enlarge

Lizzy is based on the understanding that violence rarely occurs in one form only. Physical assaults are often accompanied by attempts at control, social isolation, financial dependence or digital surveillance. “Severe violence is almost always a web of different forms of abuse,” says Le.

Once responses across various dimensions of violence have been entered, Lizzy calculates the risk for the next three months. No personal data are collected, cases are recorded anonymously and artificial intelligence is used strictly as a statistical tool.

Medical screening as prevention

Lizzy is now used by counseling centers and protective facilities in eleven German states. Le hopes to expand its reach further. So far, the team has not earned any money from the app.

At LMU, the doctoral candidate continues her research into risk assessment and violence prediction. She would like to see domestic violence identified earlier in medical settings—for example through a small number of carefully formulated screening questions during case history interviews in emergency rooms or at obstetrics and gynecology departments. “That could be done without any additional costs,” she says.

Help for victims of domestic violence

At LMU, victims of domestic violence can find support in places such as the Munich Emergency Department at the Institute of Forensic Medicine. Here, they receive counseling and prompt medical examinations. Proper documentation of injuries and evidence is crucial if victims later decide to report a crime to the police.

No crystal ball

At the same time, Le emphasizes that “Risk assessments are not crystal balls. There will never be a tool that predicts danger with one hundred percent certainty. Human behavior is too variable.” Risk assessment, she adds, is only the first step in combating domestic violence. And yet striving for the greatest possible accuracy is still worthwhile, because every percentage point of reliability can determine whether the right support arrives in time.

Doing the “hard yards”: Providing support

Enlarge

“The hardest work—managing support—comes afterward,” Le says. Survivors of violence often need extensive assistance. Therapy may be necessary after years of humiliation. A new job can help someone escape financial dependence on a partner. A place in a women’s shelter could even save a life.

“Everyone on our team has personal experience with domestic violence,” Le explains. She reflects: “If Lizzy had existed earlier, I might not have been spared violence. But I would have received help sooner.”

The orange bench

Last November, LMU made a visible statement against violence: The orange bench, which stands at the main entrance under the UN motto “UNITE to End Violence against Women”. And for several years now, LMU has also been flying the campaign flags at its main building during the UN Action Week to Stop Violence Against Women. "As a university, we want to take a stand, send a message, raise awareness, and call for action. We need more people like Ba Linh Le," says Dr. Margit Weber, Vice President for Equal Opportunities, Talent Development, and Diversity.
In December 2023, LMU adopted a policy to protect against discrimination, harassment, and sexual violence at the university. The goal is to foster a “culture of awareness,” and LMU explicitly encourages those affected to file complaints. Close cooperation partners include the criminal investigation department and the women's emergency hotline.