There are some problems that continue to defeat computer science researchers. One example is natural language understanding. It is currently not possible for a computer to participate in an unconstrained, fluid and meaningful conversation with a person. This result alone places limitations on computerized therapy. Even optimistic experts agree that solutions to this, and other “hard AI” problems, are at least several decades away.
On the other hand, there are some existing and emerging technologies that hold great potential for tomorrow’s automated therapy systems. Perhaps the most obvious, and important, is the Internet. This future of Internet-supported therapy has been covered elsewhere , so in this article I’ll give some speculative ideas about other advances we may see in the years to come.
Simulated Therapist Interaction
Perhaps the most common objection to automated therapy is that it lacks the personal connection between the patient and therapist. This is a valid concern, and this intimate relationship cannot be replicated with existing technology. However, steps can be taken in the right direction.
Computerized treatments can involve a simulated relationship with a therapist . For example, a program may feature a virtual therapist who walks the user through each stage of the program. There are two general approaches: a fictional therapist or a real therapist. A fictional therapist may be an actor, animation, or computer-generated avatar. A real therapist is usually one of the clinicians who helped create the treatment program. My preference is for a real therapist, as this is at least moving toward a relationship with a real person. The relationship can be enhanced with personalized audio and video messages, interactive exercises with automated feedback and event-driven emails.
The field of computer graphics is advancing rapidly, and within the next few years it will be possible for computer-generated “talking heads” to be almost indistinguishable from real people. This technology is likely to be incorporated into future computerized therapy systems. However, don’t be misled by visual realism – the technology to make a computer behave and interact intelligently is progressing more slowly.
One of the primary advantages of computerized therapy over traditional self-help is the ability to adapt the content to the user’s needs. In the past few years, successful clinical trials have been conducted for a number of such programs, including ones targeted toward depression, anxiety disorders, alcohol misuse, quitting smoking, and panic [3-9].
Personalization can take several forms. As a concrete example, I’ll give a brief overview of how AI-Therapy’s social anxiety treatment program works. At the beginning, the user fills out several standardized questionnaires. The system uses these data to formulate a personalized anxiety model for the user, and devises a treatment plan accordingly. For example, the program assigns exposure exercises based on the user’s specific symptoms. Underlying the system is a database of over 8,000 pieces of individualized content derived from a detailed audit of decades of real-world clinical data .
My hope is that computerized therapy systems of the future will take personalization much further, making today’s systems appear primitive by comparison.
Data Mining and Machine Learning
The statistical analysis of large data sets is changing the world we live in. It has led to novel and unexpected insights in diverse fields, such as medicine, finance, marketing, physics and politics. It is hard to understate the impact that data mining and machine learning technology has on our daily lives. However, I feel that mental health has been largely unexplored. Therapists have been recording clinical information about patients for decades, and it is exciting to consider what a large-scale analysis might reveal.
In 2011, IBM developed a computer called Watson that was able to beat the world’s top contestants at the game show “Jeopardy!.” Watson uses advanced natural language processing capabilities to parse a clue, and queries a massive database of general knowledge to formulate an answer.
IBM is adapting this technology for use in other areas. For example, when combined with a carefully curated database of evidence-based medical information, “Watson for Health care” can help doctors diagnose and treat patients. I think it is likely that in the future a similar system will be developed as an aid for human therapists, allowing more precise diagnoses and increasingly targeted and effective interventions.
Truly empathetic robot therapists still belong to the realms of science fiction. However, modern robots do have some therapeutic applications.
Personal robots are being developed for a range of general health care applications, such as monitoring vital signs and reminding people to take their medicine. These same robots also can provide an important source of companionship for the elderly, lonely, or socially isolated. The technology already is fairly advanced, with the robots capable of displaying a variety of different “personalities.” This has the potential to decrease suffering in several often-overlooked populations.
Computer vision is another example of rapid progress. Researchers still have a long way to go to match the ability of the human visual system, but some problems have been solved. This has interesting implications for computerized therapy, as computers no longer have to rely purely on self-reported information.
Automated emotion detection using video (e.g. from a webcam) is one active research area. Multispectral sensing has the potential to take this idea even further. For example, my company (AICBT Ltd.) has investigated the use of face detection technology combined with infrared detectors to gain insight into a user’s emotional state.
A computerized therapy system might be able to incorporate this information into the treatment. For example, if a computer senses user frustration, it might recommend taking a break. Alternatively, if high levels of sadness or depression are detected, the system could automatically contact a live therapist.
Virtual reality is already being used by therapists in a number of different contexts. For example, it is possible to treat phobias by exposing patients to their fears in a virtual environment. Augmented reality potentially can take this idea even further.
The most common form of augmented reality is taking a live video stream of the real world and enhancing it with computer-generated graphics. Imagine wearing a pair of glasses that overlays your field of view with information about what you are looking at. This technology already exists, with Google Glass being the most well-known example.
Augmented reality would allow patents to take behavioral experiments and exposure exercises out into the real world. An advanced system might detect that your anxiety has increased (perhaps due to an increase in your heart rate), and give you real-time instructions about how to confront the situation.
- Barak, A., Grohol, J. (2011). Current and future trends in Internet-supported mental health interventions. Journal of Technology in Human Services, 29(3), 155-196.
- Helgadottir, F. D., Menzies, R., Onslow, M., Packman, A. & O’Brian, S. (2009). Online CBT I: Bridging the gap between Eliza and modern online CBT treatment packages. Behaviour Change, 26(4), 245-253.
- Johansson, R., Sjöberg, E., Sjögren, M., Johnsson, E., Carlbring, P., Andersson, T., … Andersson, G. (2012). Tailored vs. standardized Internet-based cognitive behavior therapy for depression and comorbid symptoms: A randomized controlled trial. PLoS ONE 7(5): e36905. doi:10.1371/journal.pone.0036905
- Helgadottir, F. D., Menzies, R., Onslow, M., Packman, A. & O’Brian, S. (2009). Online CBT II: A Phase I trial of a standalone, online CBT treatment program for social anxiety in stuttering. Behaviour Change, 26(4), 254-270.
- Carlbring, P., Maurin, L., Torngren, C., Linna, E., Eriksson, T., Sparthan, E., … Andersson, G. (2011). Individually-tailored, Internet-based treatment for anxiety disorders: A randomized controlled trial. Behav Res Ther 49(1):18-24. doi: 10.1016/j.brat.2010.10.002
- Bewick, B.M., Trusler, K., Mulhern, B., Barkham, M., Hill, A.J. (2008). The feasibility and effectiveness of a web-based personalised feedback and social norms alcohol intervention in UK university students: a randomised control trial. Addict Behav 33(9):1192-8. doi: 10.1016/j.addbeh.2008.05.002
- Schueller, S.M., Leykin, Y., Pérez-Stable, E.J., Muñoz, R.F. (2013, January 30). Selection of intervention components in an internet stop smoking participant preference trial: beyond randomized controlled trials, Psychiatry Res 205(1-2):159-64. doi: 10.1016/j.psychres.2012.08.030
- Silfvernagel K, Carlbring P, Kabo J, Edström S, Eriksson J, Månson L, Andersson G. (2012). Individually tailored internet-based treatment of young adults and adults with panic symptoms: a randomized controlled trial. J Med Internet Res June 26;14(3):e65. doi: 10.2196/jmir.1853
- Andersson, G., Estling, F., Jakobsson, E., Cuijpers, P., Carlbring, P. (2011). Can the patient decide which modules to endorse? An open trial of tailored internet treatment of anxiety disorders. Cogn Behav Ther 40(1):57-64. doi: 10.1080/16506073.2010.529457