"It's hard to argue with a computer"

Abstract

We present CORE-MI, an automated evaluation and assessment system that provides feedback to mental health counselors on the quality of their care. CORE-MI is the first system of its kind for psychotherapy, and an early example of applied machine-learning in a human service context. In this paper, we describe the CORE-MI system and report on a qualitative evaluation with 21 counselors and trainees. We discuss the applicability of CORE-MI to clinical practice and explore user perceptions of surveillance, workplace misuse, and notions of objectivity, and system reliability that may apply to automated evaluation systems generally.

Machine Learning in Mental Health

“…Further, such data was reported to be easy-to-access and retrieve; to offer a route to timely information for timely interventions; and to allow for data collection to be realized at scale [ 33 , 165 , 168 , 222 ]. The analysis of data that is generated as part of peoples' everyday technology interactions and digital content creation was also reported to help identify objective markers [ 23 , 168 , 201 ] and systematic tools for capturing [ 61 , 135 , 184 , 192 ] mental health behaviors, or assessing the skills of health professionals [ 78 ]. This argument was mostly justified through descriptions of the disadvantages of traditional questionnaires, interviews, self-report and survey tools with regards to: sampling biases, subjective reporting biases, risks of incomplete information, or underrepresentation [ 64 , 78 , 128 , 153 , 155 , 211 ].…”

“…The remaining records primarily captured data from individuals who were described as "normal users," "healthy subjects," "students", or "older adults" [ 44 , 67 , 139 , 140 , 153 , 168 , 179 , 193 , 201 , 217 , 218 , 222 ], or for whom data was sampled from public social media (n = 8, plus 1 record that also includes a diagnostic sample [ 53 ]). One record further collected audio data from mental health professionals (MHPs) [ 78 ]. Table 2 provides an overview of the numbers of people, including those included as "control" groups, that were studied in each data experiment.…”

“…Finally, Hirsch et al [ 78 ] described automatically extracting insights about the MI skills of counsellors from audio signals. Initially, session recordings were labelled using an established MI Skills Code (MISC) [ 121 ].…”

“…Finally, Hirsch et al [ 78 ] reported the findings of a study with 21 counsellors evaluating an interactive user interface that visualizes the output of a system that automatically assesses their MI skills from audio. They evaluated how counsellors responded to the concept of automated skills assessment, how the system may fit within or disrupt their clinical practice, and what concerns they may have.…”

“…This challenge was particularly pronounced where information to help (clinically) validate assessed phenomena was missing [ 31 , 53 , 133 , 165 , 168 ], and also for research aimed at identifying data models that are transferable to other (real-world) data contexts [ 23 , 45 , 67 , 218 ]. In terms of ambiguity in data, Rabbi et al [ 153 ] and Hirsch et al [ 78 ] describe difficulties to isolate, in audio signals, the speaker of interest from the environment (e.g . , speech emitted from a television).…”