There's a perfectly logical explanation of why your therapist pretends to be your offender

Deepfakes blur the line between reality and fiction, making us doubt everything we see online. In therapy, however, deepfakes show promise in helping victims to confront their perpetrators and get the necessary closure with the deceased ones.

Imagine facing your abuser or someone you have always struggled to confront. This time, however, that person listens, offers some sympathy, and maybe even tells you that the terrible pain they caused you is not your fault.

Except it isn’t actually that person, but your therapist is disguising themselves as your wrongdoer using the artificial intelligence (AI) powered deepfake technology at DeepTherapy.ai, a platform developed by Dutch researchers.

The technology replicates a person's likeness and voice based on a single photo or video, enabling real-time conversations between the patient and a therapist-controlled avatar.

I sit virtually with Dr. Theo Gevers, a professor of computer vision at the University of Amsterdam and co-founder of DeepTherapy.ai, who tells me about the potential benefits of still-experimental technology that is gaining traction in the Netherlands.

How does deepfake therapy work?

Gevers says the deepfake therapy is linked to regular interventions such as exposure therapy, which may involve talking to a picture of someone who abused you.

While considered controversial by some, exposure therapy is an evidence-based intervention that has been shown to be effective in treating a wide range of conditions, like phobias and post-traumatic stress disorder (PTSD).

The therapy involving deepfakes can be used in low-risk scenarios, such as helping patients deal with grief. The conversation can be fully scripted – a patient may write a letter for the person they want to talk to, and the therapist-controlled avatar will only react to it.

“A patient lost her mother and developed hatred towards her. Regular therapy didn’t work to deal with negative emotions. However, the conversation with the therapist disguised as her mother helped her to develop more positive feelings,” Gevers says.

Grief bots, or AI replicas of deceased ones, have been rising in popularity, worrying therapists that they may disrupt the natural progression of grief. The deepfake therapy, however, is fully controlled by a professional and doesn’t involve talking to chatbots.

Gevers says deepfakes can also be used in high-risk cases, such as to help and empower women who were sexually abused.

“Speaking to the perpetrator in real life could be very dangerous, but they can talk to the perpetrator through the control of the therapist, which is safer,” he tells Cybernews.

Speaking to the perpetrator in real life could be very dangerous, but they can talk to the perpetrator through the control of the therapist, which is safer.

Theo Gevers

According to a case study of two female sexual violence survivors who spoke to the deepfakes of their abusers, a single intervention led to reduced negative emotions, PTSD symptoms,​ and self-blame, while self-forgiveness increased.

Deepfake therapy can only be used after multiple interventions have not yielded the expected result. For instance, a patient is qualified only after experiencing persistent grief lasting at least a year after their loved one’s death.

The tool is currently available to carefully vetted psychotherapists in the Netherlands and is being implemented in trauma centers across the country.

Does deepfake therapy ensure data privacy?

Information disclosed in psychotherapy sessions is highly sensitive, and AI systems often pose significant privacy risks.

Gevers says the platform is secure and doesn’t use third-party servers. Nor does it require installing software, as it runs on the browser. Sessions are encrypted, they aren’t recorded or stored, and their content isn’t used to train models.

“We don't know any names or any information, except for psychologists, because we have to know their background to make sure they are legitimate,” he explains.

Aras Nazarovas, an information security researcher at Cybernews, says the platform still likely retains some metadata, including information on who participated in the meeting and deepfake metadata, which are sensitive data points that could be targeted by hackers.

The platform complies with the European General Data Protection Regulation (GDPR). While this is good news for the patients, it raises questions about the consent of individuals whose photos are uploaded to the platform.

The GDPR requires explicit consent for processing someone’s data, including photos and images, and offenders may not provide it. Nor their victims may feel safe asking for it.

However, the platform states that within the specific therapeutic context, the intervention is justified by the principle that the patient’s psychological health and recovery take precedence.

The risks of the deepfake therapy

Deepfakes are increasingly used for malicious purposes, such as highly advanced scams, nonconsensual pornography, and the dissemination of fake news, so the deepfake therapy holds hope that there’s a bright side to the powerful technology.

However, it doesn’t mean that deepfake therapy is entirely risk-free, even if it is tightly controlled by a healthcare professional.

According to a study published in the Journal of Medical Ethics, the deepfake therapy might impact the patient-therapist relationship, if the patient starts to associate or even identify the therapist with the perpetrator or deceased loved one.

“This could possibly lead to confusion, feelings of unsafety, or an unhealthy attachment to the therapist,” the study authors wrote.

When the therapy is used for grief counselling, a patient could become “addicted” to communicating with the generated image.

However, the study emphasizes that deepfake therapy should function in “a safe and controlled therapeutic setting where the therapist is in control of what is said by the deepfake.”

Unlock exclusive Cybernews content on YouTube