AI sparks more fear than it treats. But could it help us cope with anxiety disorders?
In 1987, a woman by the name of Carol White collapsed at her luxurious California home. Diagnosed with multiple chemical sensitivity — a syndrome linked to a wide variety of physical symptoms including itching, sneezing, diarrhea, bloating, skin rash, memory problems, and chest pain — she cautiously continued to go about her housewife chores. But the panic attacks wouldn’t go away. Something was making Carol sick. Yes, there was something in the fans, the cars, the dust, the water — her environment was to blame for her symptoms. Overcome by the fear that her surroundings would lead to a slow death, she left to join a desert cult.
This is the plot of the 1995 movie “Safe” by Todd Haynes which explores Carol’s unraveling and increasing unease toward invisible pollutants in her environment. It portrays some of the symptoms of generalized anxiety disorder, whereby people “apply negative interpretations to apparently innocuous things.”
An estimated 284 million people worldwide suffer from anxiety disorders. These include social anxiety, obsessive-compulsive disorder, phobias, and generalized anxiety disorders. Anxiety disorders are among the most prevalent mental illnesses in the US affecting over 40 million adults, or 18.1% of the population. In addition, 25% of children suffer from anxiety disorders.
“Anxiety is mysterious. It can feel like an invisible cage that keeps you prisoner on your sofa, unable to move for fear of something that you can’t quite identify.” (Anonymous)
Although we all feel fear at some point in our lives, not everyone experiences the overwhelming loss of control when anxiety takes hold of their life. Take social anxiety for example. “For some, the (social) anxiety becomes so unbearable that their life (becomes) organized around avoiding social interactions,” explains Karly Hoffman King, a mental health counselor. “They stop hanging out with friends or trying to meet people. Things like dating or going to a job interview may feel impossible. Even smaller interactions such as checking out at the grocery store or ordering food at a restaurant can feel like unbearable tasks for someone with social anxiety. It truly affects so many areas of their life.”
Treatments for anxiety disorders typically include psychotherapy, medication, and alternative therapies. But research by the Anxiety and Depression Association of America found that just 37% of adults in the US received treatment. Cost, availability, and location are major hurdles in getting help. Some people may be unable to leave the house to visit a therapist. Students and low-income patients may find it hard to afford talk therapy, which starts at $100 for a 50-minute session.
There are, of course, alternatives to conventional therapeutic approaches. Apps and software tools have been around for a few years to aid in the diagnosis and treatment of anxiety. But could artificial intelligence (AI) really replace your therapist?
Fast and accurate diagnosis using AI
A quick search online reveals that 2019 has seen a bit of a surge in AI-based technologies aimed at diagnosing anxiety disorders. Earlier this year, scientists at the University of Vermont (UOV) proposed a wearable sensor that monitors the presence of internalizing disorders — a group of disorders that includes anxiety.
“If left untreated, childhood internalizing disorders predict later health problems including substance abuse, development of comorbid psychopathology, increased risk for suicide, and substantial functional impairment,” the authors write. Because long-term outcomes can be so severe, early diagnosis is important, so children get the help they need faster.
Traditionally, researchers monitor children’s behaviors and responses to certain tasks designed to diagnose anxiety and depression. Not only is this process slow, but it’s also less accurate. When UOV researchers tested the wearable sensor with 63 children, they noticed that diagnosis was more accurate (81%) compared to parents reporting their child’s symptoms (68%-75%). But what’s even more impressive is the speed at which disorders can be identified thanks to AI. The algorithm predicted anxiety and depression within 20 seconds.
“Something that we usually do with weeks of training and months of coding can be done in a few minutes of processing with these instruments,” Ellen McGinnis, a clinical psychologist at the UOV said. The team is still finalizing the tracker, but the initial findings are encouraging. Soon, children could be routinely screened for internalizing disorders.
Over in New York, Charles R. Marmar, a psychiatrist and researcher at the Department of Psychiatry, New York University School of Medicine, has been studying how AI can be used to detect post-traumatic stress disorder (PTSD) in war veterans. Around 70% of adults in the US have experienced PTSD at some point in their lives. Traditional diagnosis is based on people filling out questionnaires. The problem with self-reported diagnosis is that it’s full bias. Symptoms are often under- or over-reported.
To develop a more objective assessment, Marmar and his team had an idea: analyze speech to evaluate the speaker’s emotional state. They fed voice software created by the team that invented Siri (Apple’s voice assistant) a range of speech-based features. Then they used a machine learning algorithm called random forest to classify speech tone. Linked to changes in brain function, speech has been shown to be a valuable predictor of PTSD.
“The software analyzes words — in combination with frequency, rhythm, tone, and articulatory characteristics of speech — to infer the state of the speaker, including emotion, sentiment, cognition, health, mental health, and communication quality,” explains Dimitra Vergyri, director of SRI International’s Speech Technology and Research (STAR) Laboratory. Classification of PTSD candidates was impressively accurate at 89%. Not only does the discovery cement the idea that speech could help diagnose PTSD, it also paves the way for AI to assist in the diagnosis of anxiety disorders in the future.
Therapeutic AI approaches for anxiety disorders
Once patients receive a diagnosis, they may be referred to a therapist or receive medication. But getting treatment can be a challenge. “This is a hidden disorder and many patients find it difficult to communicate their struggle to their healthcare providers,” says Torkil Berge, a psychologist at Diakonhjemmet Hospital in Oslo, Norway. “Thousands upon thousands of individuals end up not receiving adequate treatment. Of those who do get treatment, most are probably offered drug therapy.”
According to the British Medical Association, it can take up to two years to be referred to a specialist for mental health issues in the UK. For some patients, long wait times are the difference between feeling anxious and feeling suicidal. Outside of the UK, access to therapy is stifled by costs and accessibility. Computer and app-assisted therapeutic tools have been around for a few years, but until recently it was unclear if they had made any difference. Could speaking to a bot help us heal?
It’s something researchers at Northwestern University, US, have been asking themselves. So, they put an AI system to the test. Called Tess, the integrative psychological AI designed by X2AI Inc., communicates with users who feel in distress. Whether you’re having an anxiety attack or would like someone to talk to, Tess is available for immediate support.
Combining emotion algorithms with machine learning to analyze language, the AI can be accessed through Facebook Messenger and other third-party technology integrations. To ensure that help is adequate, it is maintained by mental health experts. And Tess seems to be making a difference. When students were given unlimited access to Tess for two to four weeks, they said they felt less anxious compared to students who were not using Tess.
Admittedly, talking to a bot may seem weird when we’re used to expressing our emotions face to face. We associate icy indifference with automated chatbots. But most people appear to be accepting of cognitive-behavioral therapy (CBT) delivered by a computer. What’s more, we seem to find it easier to open up to an app than to a psychologist.
Tess isn’t the only bot available for immediate emotional support. There’s the SU Bot developed by Us AI, which is aimed at people undergoing significant life changes and those with suicidal thoughts. Then there’s Woebot which lets users track the way they feel on a daily basis. It can send videos and suggestions based on CBT.
Intrigued and easily accessible on the App Store, I tried Woebot for a week; my new digital friend. The interface is sleek and simple enough to navigate. After a brief introduction to itself, Woebot asks me to write down three negative emotions that I am currently experiencing. I choose “I am worried about a loved one” as one of them.
The bot and I then work through my negative thought by focusing on the distortions associated with it. Distortions are irrational or exaggerated thoughts. For example, the bot asks whether I am putting too much blame on myself in relation to my worry for a loved one, or whether the word is connected to black-and-white thinking (e.g. Do I ‘always’ worry about a loved one?).
Having resolved some of these distortions, Woebot prompts me to rewrite my negative thought. After I attempt the rewrite, the session ends rather abruptly and Woebot tells me that it will check in with me again tomorrow.
Although I find the bot’s explanations helpful, reading them takes a little getting used to. There’s just something about sitting in a room with a therapist or a pub with a friend to discuss one’s problems. I’m on week two with Woebot now. It’s the initial trial period set by the app and I must admit I feel a little less concerned for my friend already. Woebot minimized my woes; not bad for a free app.
None of these technologies claim to replace psychologists. It’s about breaking down barriers to entry and motivating those who find it hard to seek help, to get help. Although technology can provide relief in a moment, it cannot replace face-to-face or medical therapy.
Yet the real hurdle may be sticking with it. “Computerized CBT often does well in a clinical trial where people volunteer, where they are motivated, and where they may be paid to complete the study,” explains John Torous, director of the digital psychiatry division in the Department of Psychiatry at Beth Israel Deaconess Medical Center in Boston. “When you put it in the real world and you don’t provide that extra support, it becomes really hard for people to get through it.”
Although CBT and other talk or activity-based interventions are the norm, sometimes medication can be a lifesaver. If your anxiety is so paralyzing that you cannot focus on work or relationships suffer, SSRIs can provide a way to cope initially.
“I have no doubt that getting on the right anti-anxiety medication has changed — and maybe even saved — my life,” writes Locke Hughes for the Huffington Post. “It’s helped me get a grip on my anxious worries and thoughts, and with my symptoms managed, I’ve been able to dig deeper into the source and triggers of my anxiety.”
But finding the right medication can be another challenge altogether. Tel-Aviv based start-up Taliaz has been busy working on a solution. It’s called Predictix — an AI-based tool that lets doctors match the right medication to their patients. Predictix combines research data from genetics, neurological, and clinical trials to identify condition markers and predicts a medication that provides the best outcome.
“With a simple, non-invasive test, we collect a patient’s DNA sample alongside his or her demographic and clinical history via an online questionnaire. Predictix then provides predictive information about antidepressant drugs and recommends the ones best suited to each patient’s genetic makeup and health record,” explains the company’s CEO Dekel Taliaz. And Predictix is showing promise. One patient, who had been on four different anti-depressants, reported feeling much better after starting medication recommended by the AI.
It’s early days for AI in the diagnosis and treatment of anxiety disorders. These are difficult tasks because each patient is different; personalized therapies are a must. But bots and predictive diagnostic tools appear promising. They may not work for everyone, and they won’t replace conventional therapies just yet, but they will play an increasingly important role in a digitizing world.
Images and illustrations by Anne Freier.