June « 2012 « The AI-Therapy Blog

The six decades of excitement over Artificial Intelligence

Fresh air blew over the field of artificial intelligence (AI) in the 1950s. The initial excitement and optimism of its founders is well described by the famous quote of Herbert Simon: “I believe that, in our time, computers will be able to do anything a man can do. I believe that computers already can read, think, learn, create” (Simon, 1965 p. xiii). Yet the task turned out to be much harder than early researchers anticipated. For example, it took another three decades (1997) for the first computer to outsmart the top human chess player (IBM, 2010). This was unexpected, as chess, with its rigid and well-defined rules, seemed like an easy target for computers to master. Even in this case, the victory in 1997 was primarily due to the use of massive supercomputers and the availability of raw computational power, rather than the successful mimicking of human strategy, game play, and intelligence. Despite initial optimism, understanding and reproducing true intelligence (in a human sense) continues to be well beyond the reach of modern AI.

Eliza was a computer program that was released in 1966, still in the early days of the AI movement. The program responded to its users questions and statements that they typed on a keyboard. In many cases the users were convinced that a real person was behind the scenes (Weizenbaum, 1966). However, Eliza was fully automated, and based its output on algorithms that parsed the user’s input and formulated responses based on a programmed model. The model was designed to imitate the style of an empathic therapist using the Rogerian approach (Rogers, 1951).

Artificial intelligence and clinical psychology

Eliza was the first use of automation to create the illusion of human-human clinical interaction through a human-computer interface. In general, little progress has been made towards algorithmic techniques that are useful for treating mental disorders. The original goal of Eliza was to demonstrate and advance AI technologies such as natural language processing and pattern matching. In contrast, the goal within the clinical psychology community should be the application of these techniques using evidence-based treatment strategies to tackle real world problems. The development of fully automated therapists that are indistinguishable from human therapists remains an unsolved problem, and will continue to be for the foreseeable future. However, there are techniques currently available in the psychology literature that are suitable for automation, when treating specific, well-defined conditions.

A “computer psychologist” has been developed for AI-Therapy that can identify specific problem areas that patients report, and design individualized formulations and tailored treatment components with corrective feedback. Importantly, the computer psychologists has variety of strategies in place to direct the user in such a way that errors made on behalf of the users are kept to a minimum. As an example, participants are not asked to tell the program what unhelpful thoughts they experience, but rather the program offers suggestions based on file audit data and the clinical experience of developers. In this way, the computer psychologist is able to propose tailored cognitive behavioural therapy (CBT) techniques, such as cognitive restructuring exercises, behavioural experiments, etc. Furthermore, sample answers were written for over 1000 of the possible situations that a client might encounter, allowing the “computer psychologist” to give automatic corrective feedback. The advantage of this is that this knowledge and experience can be made available around the globe and accessed by an unlimited number of people at the same time.

For more information about AI in clinical psychology:

Helgadottir, F. D., Menzies, R., Onslow, M., Packman, A. & O’Brian, S. (2009a). Online CBT I: Bridging the gap between Eliza and modern online CBT treatment packages. Behaviour Change, 26 (4), 245-253. Cambridge Journal More …

Fjola Helgadottir, PhD, MClinPsych, is a clinical psychologist, a senior research clinician at the University of Oxford, and is a co-creator of AI-Therapy.com, an online CBT treatment program for overcoming social anxiety

Nobody sat down and decided to create the Internet

In March 1989 the blueprint for the Internet were sent by Tim Bernes-Lee to his boss at CERN. It had the humble title of: Information Management: A Proposal. The comment he got back was “Interesting, but vague”, but he was allowed to continue to pursue this Information Management System. It’s creation was motivated by Tim wanting to communicate with other researchers around the world. For this he created what we now know as the Internet. It certainly is not exclusive to nerdy researchers anymore, more accurately it takes up most of our waking lives!

The community needs better access to evidence based techniques

AI-Therapy.com came about in a bit similar manner. I was finishing my clinical psychology Internships in Sydney, Australia. However, before my last year had finished I was awarded a prestigious PhD scholarship at the University of Sydney. There was a major dilemma, do I miss out of this exciting scholarship award which was funded to respond to the need for evidence based treatments using the internet, or do I not finish my clinical psychology training/internships?

From this AI-Therapy was born.

My solution to this problem was to generate a fully automated computer psychologist which could be treating people online simultaneously with me working at the various hospital settings for my internships. It turns out, that to solve my problem of doing two things at once, I created a product which helped bridge the gap between the need for access to evidence based psychology techniques in the general community. After many years of hard work, I am pleased to announce that this service is now available to the general public under the name of AI-Therapy overcome social anxiety. You can see how it works here.

If you are interested in the original research for this computer psychology service, please refer to the following research and the associated publications