Breaking the Silence: Encouraging Conversations with Digital Humans
Authors
Thøgersen, Sofie Kristine ; Givskud, Jeppe Firring ; Hansen, Rebecca Damgaard
Term
4. term
Education
Publication year
2025
Submitted on
2025-06-04
Pages
137
Abstract
Digital humans (AI-powered virtual receptionists or avatars) are increasingly used in service settings. This thesis examines how user engagement and experience are affected when a conversation is started by the user versus by the AI. Fieldwork at Aarhus Jobcenter and Krifa revealed barriers to interaction: unclear social cues and uncertainty about how to begin and conduct a conversation. To explore this, a digital receptionist named Rosie was placed in semi-public spaces at Aalborg University under two conditions: Rosie either greeted people proactively or waited passively for them to start. In total, 175 trials were completed with 29 participants. We assessed interaction through success rates, live observations, exit interviews, and questionnaires about two constructs: the Technology Acceptance Model (TAM), which captures perceived usefulness and ease of use, and the Temple Presence Inventory (TPI), which measures the sense of presence or realism. There were no statistically significant differences between conditions in success, TAM, or TPI. However, participants were less likely to ask a facilitator for help when Rosie initiated, suggesting that a proactive greeting provides a clear cue to speak to the system. Thematic analysis also indicated that when users initiated the conversation, it elicited more reflective and cognitively engaged responses. Overall, who starts the conversation shapes how people interpret the encounter with a digital agent. Design implication: Digital humans that greet first can reduce ambiguity and social friction, which may enhance engagement and the perceived intelligence of the system, especially in social contexts. The study highlights the importance of framing the interaction and calls for further research into the contextual and social dynamics of AI-mediated encounters.
Digitale mennesker (AI-drevne virtuelle receptionister eller avatarer) bruges i stigende grad i serviceprægede miljøer. Denne kandidatafhandling undersøger, hvordan det påvirker brugerens engagement og oplevelse, om samtalen startes af brugeren eller af den digitale medarbejder. Indsigter fra feltarbejde i Aarhus Jobcenter og Krifa pegede på barrierer: uklare sociale signaler og usikkerhed om, hvordan man starter og gennemfører en samtale. For at undersøge dette blev en digital receptionist, Rosie, placeret i semi-offentlige områder på Aalborg Universitet i to betingelser: Enten hilste Rosie proaktivt på folk, eller også ventede hun passivt på, at nogen tog initiativ. I alt blev der gennemført 175 forsøg med 29 deltagere. Vi målte interaktionen via succesrate, løbende observationer, exitinterviews samt spørgeskemaer om to konstrukter: Technology Acceptance Model (TAM), der handler om oplevet nytte og brugervenlighed, og Temple Presence Inventory (TPI), der måler fornemmelsen af tilstedeværelse eller realisme. Der var ingen statistisk signifikante forskelle mellem betingelserne i succes, TAM eller TPI. Til gengæld bad deltagerne sjældnere om hjælp fra en facilitator, når Rosie tog initiativ, hvilket tyder på, at en proaktiv hilsen fungerer som et tydeligt signal om, at man kan tale til systemet. Den tematiske analyse viste også, at når brugeren startede samtalen, gav det mere reflekterede og kognitivt engagerede svar. Samlet peger resultaterne på, at måden, en samtale initieres på, former, hvordan mennesker fortolker mødet med en digital medarbejder. Implikation: Digitale mennesker, der hilser først, kan reducere tvetydighed og social friktion og dermed øge engagement og måske den oplevede intelligens, særligt i sociale sammenhænge. Afhandlingen understreger betydningen af at indramme interaktionen tydeligt og behovet for mere forskning i de kontekstuelle og sociale dynamikker i AI-formidlede møder.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords