Ronen Grinberg, Dan Waisman, Rinat B. Rosenberg-Kima E69 Table 1. Main themes concerning the credibility and appropriateness of the Virtual Participant (VP) Theme Example Quotes 1. Verbal and Content Credibility • "The character's responses are too general." • "It's also like an autistic trait that isn't always true, that people think they only respond to their areas of interest." • "During the conversation, the character always stays on topic, doesn't lose focus and scatter to other topics like in reality." 2. Emotional and Behavioral Credibility • "The really extreme reactions stood out. The direct gaze, sort of piercing and all, like you imagine it, but it doesn't really happen that way." • "I felt it was too extreme, like very, very extreme reactions." • "That binary that was there, like, between softening and rigid, softening and rigid, there was something too dichotomous about it." 3. Credibility in representing resistance and initiative • "I don't need any incentive to go to the Lego store. I go there when I want to" – marked as a non-credible sentence. 4. Inconsistent verbal utterance and described behavior • "The character says: 'Just. It doesn't interest me' and it's described as: [Shifts gaze back to the phone, ignores the question], the interpretation that the character ignoring the question is incorrect." • "The resistance here is very linguistic." 5. Absence of physical presence • "You can, really, without listening, just by looking at him, you know he's anxious right now, you know he's tense, you know he's getting annoyed with you, but you know he's with you right now." • "It feels to me that the difficulty in the conversation is sitting with the person in the room and feeling him." 6. Absence of shared history • "You learn each person, you know that if he folds his fingers, he's stressed, and this is how he talks to you when he's calm." • "Is there no option to create a personal history in the prompt? Like an ongoing conversation... We're talking about the personal plan, and he remembers." • "Maybe [we need] to add another distinction between a participant who is new, just arrived, just got settled, and a participant who has been in the program for a year, a year and a half" Discussion The findings point to a mixed picture: on the one hand, participants experienced the system as a structured, engaging space for practicing empathic dialogue; on the other hand, they identified clear limitations regarding the credibility and appropriateness of the VP. From the perspective of the Media Equation and CASA frameworks, the VP was expected to function as a believable social actor that could support sociocultural learning processes (Nass & Moon, 2000; Reeves & Nass, 1996). However, the social workers’ critiques indicate only a partial realization of this promise. Rather than a specific, coherent individual, the VP often appeared as a statistical "average". This gap is consistent with concerns raised by Kirkeby-Hinrup & Stenseke (2025), who describe how LLM-based language can approximate human expression while still leaving users with a sense of unease and uncertainty about the "mind" behind the text.
RkJQdWJsaXNoZXIy Mjk0MjAwOQ==