Observing Dialogue in Therapy: Categorizing and Forecasting Behavioral Codes

Abstract

Automatically analyzing dialogue can help understand and guide behavior in domains such as counseling, where interactions are largely mediated by conversation. In this paper, we study modeling behavioral codes used to asses a psychotherapy treatment style called Motivational Interviewing (MI), which is effective for addressing substance abuse and related problems. Specifically, we address the problem of providing real-time guidance to therapists with a dialogue observer that (1) categorizes therapist and client MI behavioral codes and, (2) forecasts codes for upcoming utterances to help guide the conversation and potentially alert the therapist. For both tasks, we define neural network models that build upon recent successes in dialogue modeling. Our experiments demonstrate that our models can outperform several baselines for both tasks. We also report the results of a careful analysis that reveals the impact of the various network design tradeoffs for modeling therapy dialogue.

Balancing Objectives in Counseling Conversations: Advancing Forwards or Looking Backwards

“…Orientation also complements prior work on dialogue acts, which consider various roles that utterances play in discourse (Mann and Thompson, 1988;Core and Allen, 1997;Ritter et al, 2010;Bracewell et al, 2012;Rosenthal and McKeown, 2015;Prabhakaran et al, 2018;Wang et al, 2019). In counseling settings, such approaches have highlighted strategies like reflection and questionasking (Houck, 2008;Gaume et al, 2010;Atkins et al, 2014;Can et al, 2015;Tanana et al, 2016;Pérez-Rosas et al, 2017Park et al, 2019;Lee et al, 2019;Cao et al, 2019). Instead of modeling a particular taxonomy of actions, we model how counselors balance among the underlying objectives; we later relate orientation to these strategies (Section 5).…”

Towards Detecting Need for Empathetic Response in Motivational Interviewing

“…It is worth mentioning that there is more recent work on deeplearning-powered MI analysis, including text-based [1,7,32,36], speech-based [30], and multimodal [2] coding of therapist actions according to MI behaviour codes, but we do not consider this sphere of research since it is still predominantly for offline assessment of therapy quality and does not involve explicit empathy modelling.…”

A Multi-Persona Chatbot for Hotline Counselor Training

“…Recent interest in hotline conversations has analyzed the structure of real conversation transcripts with the hope of better understanding what makes hotline conversations perceived to be helpful (Althoff et al, 2016;Zhang and Danescu-Niculescu-Mizil, 2020). Similar work has considered talk therapy conversations and tried to understand conversations and predict counseling acts (Pérez-Rosas et al, 2017 or forecast therapists' next actions (Cao et al, 2019).…”

“…To understand input counselor messages and enable the model to look up similar contexts, we develop a counselor strategy annotation scheme and annotate the strategies of counselor sub-utterances in 160 transcripts. tion strategies for persuasion , negotiation (He et al, 2018), and motivational interviewing (Cao et al, 2019;Pérez-Rosas et al, 2017), but these annotations are specialized for a hotline counseling context. However, they could be generalized to related applications, and used to extract related conversational exchanges to generate diverse response components.…”