Can AI deliver empathetic medical consultations?

Specific intervention: Participants watching a medical consultation delivered by a Digital Human (DH). DHs are embodied conversational agents (ECAs) whose interactions are shaped by artificial intelligence (Loveys et al., 2022). DHs possess dialogue systems, allowing them to communicate with users through text and speech-based channels (Loveys, 2021). DHs may take on various forms with animated faces or bodies, ranging from simple, cartoon-like depictions (Hudlicka, 2015) to more complex, realistic, and humanlike embodiments (Loveys, 2021). Participants will be randomly assigned to the intervention condition (medical consultations delivered by a DH), or the control condition (medical consultation delivered by a human physician). Information for non-drug trials: Materials used: Through qualtrics (survey platform): Participant information sheet (PIS) and consent forms for participants, videos of brief medical consultations and a questionnaire for participants to complete after watching the video they have been randomly assigned to. Digital human training/instructional and design materials provided by Soul Machines and IBM Watson for the researchers to create and program the digital human BMJ Clinical Review article used as an instructional material to devise the script for the medical consultations. The script was used for the consultation delivered by the human physician, which also informs the content for the consultation delivered by the digital human Filming and electronic equipment used to record videos of the human physician and digital human interacting with an actor patient Procedures: Firstly, through the qualtrics survey link, participants will be asked to read the PIS, sign a consent form and complete a demographic information questionnaire. Participants will be randomly assigned to one of the four conditions using the qualtrics software, and they will watch a brief video of a medical consultation (empathetic or non-empathetic) delivered by a human physician or a digital human. Next, participants will be asked to fill out a measure of empathy, rating their perceptions of the interaction. Participants will also complete measures of trustworthiness, competence, and warmth, given that these are outcomes closely related to perceptions of physician empathy and can enhance knowledge of the acceptability of digital humans compared to human physicians on multiple fronts. Participants will also be asked about their prior experiences with technology, as those with previous exposure may differ from those without prior knowledge or encounters with digital humans/virtual agents. The administering of these measures will be counterbalanced, given the possibility for the first measure to prime responses to the subsequent measures. However, counterbalancing will depend on the parameters of qualtrics. Who will deliver the intervention and their expertise: The intervention will be delivered by a Digital Human. Participants will have access to pre-recorded videos embedded into the qualtrics survey, where they will watch the consultation they have been randomly assigned to before answering the survey questions. The digital human delivering the medical consultation will be designed and programmed using Soul Machines and IBM Watson, mirroring the content and behaviours of Dr. Anna (the human physicians) in her consultations for comparibility. The human physician delivering the medical consultation is Dr. Anna Perera, a Medical Registrar at Auckland DHB. Mode of delivery (individual/or group): Individual, as each participant will be randomly assigned to watch either a digital human consultation or a human physician consultation based on the study protocol. Number of times the intervention will be delievered and over what period of time: For participants assigned to the intervention conditions (Digital human videos: empathetic or non-empathetic), the intervention will be delivered only once per session. This means participants will watch the one pre-recorded video they have been randomly assigned to, lasting approximately 4.5 minutes, and then compltete the attached survey questions. Overall, it is expected that participation will take about 15-20 minutes, of which the intervention will be 4.5 minutes, delivered once per participant. The location where the intervention occurs: the intervention is delivered through the University of Auckland (UoA) qualtrics platform. The survey link will allow participants access to one of the pre-recorded videos (filmed on the UoA campus), and the study measures. Digital human intervention: The content of the digital human intervention will be based on the human physician videos. Dr. Anna Perera (the human physician) was consulted in designing an empathetic and non-empathetic consultation, along with a literature review. The digital human will be scripted to say the same things as Dr. Anna in her videos, to keep the content as similar as possible with minor changes to account for the modality. In sum, the digital human will vary in empathy based on the empathy or non-empathy condition. In each condition, it will ask the student actor (Tim) about his symptoms and suggest that he has a common cold based on his symptoms and provide recommendations such as taking paracetamol and decongestants. In the empathy condition, the digital human will show verbal/non-verbal behaviours indicative of empathy, such as eye contact, smiling, nodding and reflecting the actor's emotions, as well as making empathic statements. In the non-empathetic condition, the digital human will be programmed to not display these behaviours.