No open-access location is known for this record yet.
No open-access location is known for this record yet.
Motivational Interviewing (MI) is a conversational technique that empowers change. MI has been increasingly applied in schools. Teacher coaching embedding MI is an implementation strategy for shifting teacher practices and student outcomes; importantly, many school consultation models incorporate communication skills emphasized by MI. Though MI principles and assumptions are simple to convey, skilled use takes time and is best supported by specific, ongoing feedback about one’s practice. The capacity for such training and supervising educators to engage in MI is limited. This paper explores the extent to which machine learning (ML) can accurately code MI data; such automated MI coding could ultimately be used for training and supervision. We utilized 74 archival Double Check coaching session transcriptions from a randomized efficacy trial; the transcriptions were previously coded manually using an adapted version of the MI-SCOPE. After segmenting each conversation into individual utterances, we conducted a text classification analysis using three ML methods to assess the performance of ML for such coding. Results demonstrate the potential of ML when utterances were pre-defined and for language codes that occurred frequently (i.e. “other” language); unfortunately, those of greatest interest in MI (i.e. MI-consistent and inconsistent language; teacher change and sustain talk) occurred with relatively low frequency and thus were coded with poor accuracy. These results replicate and extend the prior limited research using ML to code MI within a therapeutic context. We identify future directions and lessons learned in leveraging ML to process such data.
Loading…
Stored only when both papers are in the corpus — 1 citations exist upstream.
Papers that cite the same references — the relational fallback while embeddings are still backfilling.
Imported from OpenAlex (CC0) — nothing here is generated.