Skills used Data cleaning, feature selection, machine learning, Python (Jupyter notebooks and PyCharm)
Abstract Virtual humans are frequently used to help medical students practice communication skills. Here, we show that communication skills features drawn from the literature on best practices for doctor-patient communication can be used to predict student interviewers' success in a given domain skill. We also demonstrate the viability of Bayesian Rule Lists, an interpretable machine learning model, for this use case. Bayesian Rule Lists' predictive performance is comparable to that of other other commonly used algorithms, including decision trees. This suggests that Bayesian Rule Lists, which produce simple, human-readable trained binary classifiers, may be suitable for providing feedback for educational purposes.
Skills used Statistical analysis (SPSS), Study and survey design, Qualtrics
Abstract Because human-produced language is often still considered the golden standard in the natural language generation field, many virtual humans (VHs) used in communication-skills training simulations continue to be accompanied by human-authored dialogue instead of computer-generated dialogue. The human author used for writing such dialogue, however, has yet to be investigated. Dialogue authors can have a variety of identities - gender, culture, profession, etc. - which may or may not have an effect on the VH dialogue they produce. An identity incongruency between an author and the VH itself could have unexpected effects on VH interviewers' interactions. Therefore, in this paper, we examine the effect of an author's identity on VH dialogue produced in the context of a virtual patient (VP) interview. We evaluated writing samples from speech-language pathology (SLP) students and Chinese and Chinese Americans when creating a Chinese VP suffering from trouble swallowing. We also conducted a user study in which SLP students interviewed VPs created with dialogue from these different author identities. Our results indicate that there are some content differences in the dialogue produced by different author identities and that VH interviewers are able to recognize these differences.
Adapting Virtual Patient Interviews for Interviewing Skills Training of Novice Healthcare Students
Abstract The purpose of this paper is to explore the potential of using a selection-based interaction method to adapt virtual patient interviews for training novice healthcare students in interviewing skills. We outline a method for identifying topics about which novices are likely to ask by reviewing previous transcripts of novice interactions, and results indicate that the method was successful in such identification. Additionally, we examine the possibility of using a selection-based virtual patient as a modeling tool for learning interviewing skills. Our initial results support the possibility of such question modeling and reveal that healthcare students already view selection-based virtual patient interactions as modeling opportunities.