Motivational Interviewing (MI) is a form of counseling that is exceptionally suited for substance abuse therapy. By using Automatic Speech Recognition, counselors are able to review sessions, see patient reactions, and view vocal arousal during the session. This is especially useful for counselors in training so that they can see which metrics need the most improvement and better understand why others do not.
This project was completed during a Resident Assistantship at the University of Washington School of Medicine, Psychiatry and Behavioral Sciences.
We present an overview of CORE-MI, an automated coding and visualization system that provides reportcard like feedback on psychotherapists’ adherence with Motivational Interviewing (MI). MI is a widely studied and effective method of substance abuse treatment, and is widely practiced. However, there are currently no technologies to provide immediate, specific feedback to therapists on the efficacy of their counseling. CORE-MI combines speech and language processing with automated coding and interactive visualization to help therapists and trainers identify strengths and areas for improvement, to help them provide effective counseling to their clients.
Full publication available here.
David Atkins, Department of Psychiatry, University of Washington. Seattle, WA
Geoff Gray, Division of Design, University of Washington, Seattle, WA
Tad Hirsch, Division of Design, University of Washington, Seattle, WA
Zac E. Imel, Department of Educational Psychology, University of Utah, Salt Lake City, UT
Shrikanth Narayanan, School of Engineering, University of Southern California, Los Angeles, CA
Research for this project was mostly reviewing existing literature on CORE-MI. Understanding existing metrics was crucial for developing consistent, normalized graphs. There were also a few books on data vis that were heavily referenced:
(1) Show Me the Numbers: Designing Tables and Graphs to Enlighten by Stephen Few
(2) The Functional Art: An introduction to Information Graphics and Visualization by Alberto Cairo
(3) Information Dashboard Design: Displaying Data for At-a-glance Monitoring by Stephen Few
Luckily, I was able to provide a JSON template for this project. Having this level of control over the data was a luxury not typical to many data vis projects. The design process was much more linear thanks to this.
This project was a continuation of existing work done by Jon Cook who had already designed a fairly robust print-only version of this report card. Using this as a starting point, I jumped straight into digital wireframes to imagine new functionality that could be gained by making this interactive.
We knew that this tool was primarily going to be used on either a tablet or a desktop/laptop. As such, it was important to keep in mind that mouse hover states and screen tapping would need to have the same functionality. Luckily, mobile phones were not seen as a likely platform. Data visualizations benefit greatly from screen size. If phones had been a primary consideration then this project would have taken a much different path.
The project brief pointed to an interactive data dashboard so most of the time was spent looking at what types of interactions would be most useful. Additionally, I quickly sketched a few other interaction models in the hope that they would elicit overlooked features for the dashboard.
As this was only a 10-week project and I was the sole designer and developer, I knew I would need to get through the design phase quickly. Through weekly feedback sessions with key stakeholders, features were added, reworked or removed, all while refining visual design.
The following designs had been approved and were the starting point for the development process.
There were definitely benefits to being the sole designer and developer for the project. Once visual design was approved and development was taking shape, we unsurprisingly came across a number of issues. Such is the nature of moving from design to development.
As a designer, I can’t help but imagine almost-indefinitely new features and refinements. No project is ever done. As a developer, I can’t help but dread these additional features that were originally out of the project’s scope. It was strange, the developer me being a bit pissed at the designer me at times.
Bootstrap for most of the responsive behavior and custom D3.js for the visualizations. Unfortunately the project had to be wrapped up before I had time to refactor the code and to a cleanup pass or two. But if you’re curious, Github repo is here.
Still honing in on most useful features and functionality, we refined our use case scenario with low-fi mockups.
Health metrics and data visualizations are proving exceedingly insightful. Reviewing and reflecting on patient-provider conversations, especially, can help providers refine their approach to me one that is more patient-centric. CORE-MI could easily be extending the tool beyond Motivational Intertviewing to all forms of counseling and, potentially, other fields of medicine.
At present, this tool is being used to help MI counselors in training improve their tecniques and tactics to reach a baseline proficiency. CORE-MI also serves as a sort of reference manual as each metric is described in a modal state.