Artificial Intelligence at Ieso - Part 3: The Future - Valentin Tablan
On March 1st 2017 I gave a presentation at the summit dedicated to Deep Learning in Healthcare, organised by ReWork in London. I used that opportunity to talk about Ieso’s work on using artificial intelligence to improve mental health provision in the UK. This blog post is a companion for the content of my presentation.
Part 3: The Future
Last time we spoke about an experimental statistical model we built, to see if we can use our historic data to help computers make useful predictions in the area of psychotherapy.
Based on the success of that experiment, we are now launching a research and development programme aimed at building a variety of intelligent tools to support our team of therapists in their work. For example, the therapists’ user interface can pre-populate itself with information relevant to the expected condition of every new patient, based on the hypothesis produced by the system using the patient’s answers on a questionnaire. It has been established that faithful adherence to the treatment protocol is a strong indicator for the success of therapy – the closer the therapist stays to the protocol, the more likely the patient is to recover, and the quicker recovery will be reached. Our unique data-set can be used to build computer models endowed with sufficient understanding of language to be able to follow a therapy session as it happens. Because the patient’s condition and treatment protocol are known, the system can provide hints and prompts to the therapists, helping prevent drift, and optimising the delivery of the therapy active ingredient rather than plain conversation.
 A. Gyani, R. Shafran, R. Layard and D. M. Clark, “Enhancing recovery rates: lessons from year one of IAPT,” Behaviour Research and Therapy, vol. 51, no. 9, pp. 597--606, 2013.
Even within the constraints of a given treatment protocol, there are multiple ways a certain intervention can be provided. There can be variations in the style of language being used, in the choice of motivational device, the examples given, or the precise timing in the treatment episode when a particular intervention is made. Our historic data can tell us which of these variants is likely to work better for each individual patient, based on a variety of indicators specific to the patient, and we can use this information to provide hints to our therapists. With the support of these types of tools, our therapists will become more effective and more efficient than they would otherwise be. Our goal is to help each and every one of them be the best therapist that they can be. Over time, this means that more of our patients will get better, that they will recover faster, and that our therapists will be able to treat more patients in the same amount of time.
We are excited to start working on these tools, and are looking forward to being able to report in the future what the effects of these interventions were measured to be.