A “multimodal” approach to research brings together multiple, often disparate, methods or
modalities of measurement to solve problems and accomplish scientific objectives. For
example, it can enable us to integrate measurements of both verbal and nonverbal behavior,
thus providing a more complete and nuanced representation of patient-therapist
communication. Multimodal approaches can also enable us to connect traditional measures,
such as self-reports and clinician ratings, with novel measures (e.g., neural and physiological
activity, and technology usage). Machine learning is especially well-suited to sifting through
multi-modal variables and modeling their complex and non-linear relationships.
The use of multimodal measurement approaches should yield the most accurate prediction and
offer new opportunities for implementation in clinical practice.

Our research program focuses on two aspects of psychotherapy research: the therapeutic
relationship and professional development. I research different (un) conscious aspects of the
therapeutic relationship, including the working alliance, (counter) transference and the real
relationship. The research on therapists’ professional development includes facilitative
interpersonal skills training, as well as routine outcome monitoring, and deliberate practice. In
recent years, I have expanded our research to the teletherapy context, including research on
the therapeutic relationship, therapeutic agency, therapeutic presence and teletherapy
training.

Most of our previous research has been based on self-report and observer-based
measures, including computerized text-analysis of transcribed sessions. Most recently, I am
focusing on multi-modal measurements of visual (e.g., facial expression and head motion),
voice (e.g., pitch), and speech (text) markers of the patient and therapist obtained from video
recordings of teletherapy sessions. Machine learning is especially well-suited to sifting through
multi-modal variables and modeling these complex and non-linear dyadic relationships. Our
current focus is on the development of automatic inference models for the working alliance,
and therapists’ interpersonal skills. This is an exciting direction of research, especially because it
allows for examinations of the paraverbal and non-verbal aspects the therapeutic relationship,
which are crucial in any communication between individuals, but especially in psychotherapy.
Most importantly, these computational models will also allow for direct feedback to therapists
in the moment of the session, therefore providing a direct usability to the therapist in their
clinical work and professional development.