VICA, a visual counseling agent for emotional distress

Abstract

We present VICA, a Visual Counseling Agent designed to create an engaging multimedia face-to-face interaction. VICA 1 is a human-friendly agent equipped with high-performance voice conversation designed to help psychologically stressed users, to offload 2 their emotional burden. Such users specifically include non-computer-savvy elderly persons or clients. Our agent builds replies 3 exploiting interlocutor's utterances expressing such as wishes, obstacles, emotions, etc. Statements asking for confirmation, details, 4 emotional summary, or relations among such expressions are added to the utterances. We claim that VICA is suitable for positive 5 counseling scenarios where multimedia specifically high-performance voice communication is instrumental for even the old or digital 6 divided users to continue dialogue towards their self-awareness. To prove this claim, VICA's effect is evaluated with respect to a 7 previous text-based counseling agent CRECA and ELIZA including its successors. An experiment involving 14 subjects shows VICA 8 effects as follows: i) the dialogue continuation (CPS: Conversation-turns Per Session) of VICA for the older half (age >40) substantially 9 improved 53% to CRECA and 71% to ELIZA. ii) VICA's capability to foster peace of mind and other positive feelings was assessed 10 with a very high score of 5 or 6 mostly, out of 7 stages of the Likert scale, again by the older. Compared on average, such capability of 11 VICA for the older is 5.14 while CRECA (all subjects are young students, age<25) is 4.50, ELIZA is 3.50, and the best of ELIZA's 12 successors for the older (>25) is 4.41.13

An Evaluation of Generative Pre-Training Model-based Therapy Chatbot for Caregivers

“…AIML, firstly adopted by ALICE (Artificial Linguistic Internet Computer Entity), is an XML-compliant language that allows for efficient pattern matches in a tree structure for retrieving responses. Seminal therapy chatbots reported in the literature-VICA, a virtual agent equipped with voice-communication [35], a conversational agent for alcohol misuse intervention [36], and a counseling agent in the IT industry [37]-all applied AIML to build the chatbot. Vivibot [38], Woebot [39], and a virtual agent for post-traumatic stress disorder [40] also applied decision tree structures.…”

“…User experience. Research in therapy chatbots applied user research to evaluate user experience , including measuring users' trust and comfortability [35], emotional states [39,12,38], overall satisfaction [36], and acceptability and usability outcomes [39,4,40]. Several researchers used the Positive and Negative Affect Schedule (PANAS) to test emotional states [39,12,54].…”

“…The responses are auto-generated by the chatbots whose quality can not be guaranteed, unlike human moderated responses in the retrieval-based approach. Previous therapy chatbot research used similar evaluations to measure speech performance, such as the average number of interactions in a single dialogue session called Conversation-turns Per Session (CPS) [35,37,56].…”