HUMAN-CENTRIC INTERACTION IN IMMERSIVE AND GENERATIVE SYSTEMS: INVESTIGATING TRUST, EMPATHY, AND INTEGRATION IN MENTAL HEALTH CONTEXTS

THE RAPID INTEGRATION OF ARTIFICIAL INTELLIGENCE (AI) INTO DAILY LIFE HAS POSITIONED CONVERSATIONAL USER INTERFACES (CUIS) AS PRIMARY MEDIATORS OF HUMAN-DIGITAL INTERACTION. WHILE GENERATIVE AI (GENAI) AND VIRTUAL REALITY (VR) OFFER UNPRECEDENTED OPPORTUNITIES FOR ENGAGEMENT, THEY ALSO INTRODUCE CRITICAL CHALLENGES REGARDING TRUST AND EMPATHY. IN THE FRAMEWORK OF HUMAN-CENTERED ARTIFICIAL INTELLIGENCE (HCAI), THIS THESIS INVESTIGATES HOW THESE TWO FUNDAMENTAL HUMAN CONSTRUCTS ARE SHAPED, MAINTAINED, OR COMPROMISED WITHIN AI-MEDIATED INTERACTIONS. ALTHOUGH THE RESEARCH IS EMPIRICALLY SITUATED WITHIN THE HIGH-STAKES DOMAIN OF MENTAL HEALTH USED HERE AS A RIGOROUS "STRESS TEST" FOR SYSTEM RELIABILITY AND RELATIONAL ATTUNEMENT THE FINDINGS OFFER GENERALIZABLE INSIGHTS FOR THE DESIGN OF TRUSTWORTHY AND EMOTIONALLY RESONANT SYSTEMS. THE RESEARCH IS STRUCTURED AROUND THREE CONVERGING EMPIRICAL STRANDS. THE FIRST STRAND ADDRESSES THE FRAGILITY OF TRUST IN TEXT-BASED CUI (E.G., CHATGPT). THROUGH CONTROLLED USER STUDIES, IT CHARACTERIZES HOW USERS REACT TO "HALLUCINATIONS" AND FACTUAL ERRORS. THE RESULTS REVEAL A SIGNIFICANT "PRIMACY EFFECT": WHILE USERS SHOW RESILIENCE TO OCCASIONAL ERRORS, EARLY EXPOSURE TO UNRELIABLE INFORMATION CAUSES A SHARP AND LASTING DECLINE IN TRUST THAT IS DIFFICULT TO RECOVER. THIS HIGHLIGHTS THE DANGERS OF OPAQUE "BLACK-BOX" SYSTEMS THAT FOSTER OVERTRUST. THE SECOND STRAND MOVES TO EMBODIED INTERACTION, EXPLORING EMPATHY IN VIRTUAL REALITY. BY LEVERAGING FACIAL TRACKING TECHNOLOGY TO MEASURE MIMICRY IN INTERACTIONS WITH AVATARS, THIS RESEARCH DEMONSTRATES THAT EMPATHY IS NOT MERELY A USER TRAIT BUT A "SITUATED" AND EMBODIED PHENOMENON. THE FINDINGS SHOW THAT AN INDIVIDUAL’S CAPACITY FOR EMOTIONAL SIMULATION IS SIGNIFICANTLY MODULATED BY PERSONAL DISPOSITIONS (SUCH AS FANTASY OR PERSONAL DISTRESS) AND BY THE EMOTIONAL CONGRUENCE OF THE VIRTUAL ENVIRONMENT, CHALLENGING THE NOTION OF VR AS A UNIVERSAL "EMPATHY MACHINE." SYNTHESIZING THESE INSIGHTS, THE THIRD STRAND PROPOSES A METHODOLOGICAL SHIFT FROM PASSIVE CONSUMPTION TO GENERATIVE CO-CREATION. THE THESIS INTRODUCES AND VALIDATES A NOVEL FRAMEWORK WHERE GENAI AND VR SERVE AS TRANSPARENT TOOLS FOR TRUST BUILDING. INSTEAD OF ACTING AS AUTONOMOUS ORACLES, THESE TECHNOLOGIES ARE EMPLOYED TO ALLOW USERS AND PROFESSIONALS TO COLLABORATIVELY GENERATE AND MANIPULATE 3D THERAPEUTIC OBJECTS AND ENVIRONMENTS IN REAL-TIME. VALIDATION STUDIES OF THIS PROTOTYPE DEMONSTRATE HIGH SYSTEM USABILITY (SUS SCORE OF 79) AND ACCEPTANCE (NPS OF 40). CRUCIALLY, THE SYSTEM PROVES TO BE INCLUSIVE, AS PERCEIVED USABILITY REMAINS HIGH REGARDLESS OF THE USER’S LEVELS OF ANXIETY OR DEPRESSION. ULTIMATELY, THIS THESIS ARGUES THAT TO FOSTER GENUINE TRUST AND EMPATHY, AI SYSTEMS MUST BE REDESIGNED NOT TO REPLACE HUMAN JUDGMENT, BUT TO FUNCTION AS CONTROLLABLE, TRANSPARENT, AND EMOTIONALLY RESPONSIVE.