Facilitating Verbal Participation in Quiet PBL Groups with a Voice-Based AI: A design Science Study of LIFT
Translated title
Facilitering af mundtlig deltagelse i stille PBL-grupper med en stemmebaseret AI: Et Design Science Studie af LIFT
Authors
Ahmed, Yusuf Issa ; Reiding, Lucas Lindholm
Term
4. Term
Publication year
2026
Submitted on
2026-06-11
Pages
54
Abstract
This thesis investigates how a voice-based AI facilitator can support verbal participation in quiet Problem-Based Learning (PBL) groups, guided by the question: To what extent can a voice-based AI facilitator shape participation during meetings in quiet PBL groups? Using a Design Science Research approach, we developed LIFT, a prototype that listens to group discussions, transcribes speech, delivers spoken facilitation, and visualizes participation patterns. LIFT was tested with three PBL groups involving seven master’s students and evaluated through thematic analysis of meeting transcripts, debrief interviews, and participation summaries. Participants accepted LIFT as a facilitator or helper rather than an equal group member, valued its summaries and clarifications, and generally did not find speaking-time visualizations overly exposing; however, verbal invitations intended for quieter members were not fully tested because the main feature did not trigger during sessions. The study indicates promise for reflection, summarization, and participation awareness, and contributes design principles for future voice-based AI facilitation tools. Limitations suggest improving response timing, speaker identification, and contextual understanding, and conducting longer-term evaluations with larger groups.
Dette speciale undersøger, hvordan en stemmebaseret AI-facilitator kan understøtte mundtlig deltagelse i stille Problem-Based Learning (PBL)-grupper, guidet af forskningsspørgsmålet: I hvilket omfang kan en verbal AI-baseret facilitator forme deltagelse under møder i stille PBL-grupper? Med en Design Science Research-tilgang udviklede vi LIFT, en prototype der lytter til gruppediskussioner, transskriberer tale, giver talte faciliteringsresponser og visualiserer deltagelsesmønstre. LIFT blev afprøvet i tre PBL-grupper med i alt syv kandidatstuderende og evalueret gennem tematisk analyse af mødeudskrifter, debrief-interviews og deltagelsesresuméer. Deltagerne accepterede LIFT som facilitator/hjælper snarere end som ligeværdigt gruppemedlem, værdsatte opsummeringer og afklaringer og fandt generelt ikke visualisering af taletid for udstillende; dog blev verbale invitationer til mere stille medlemmer ikke fuldt afprøvet, fordi hovedfunktionen ikke blev udløst under sessionerne. Studiet peger på potentiale for refleksion, opsummering og bevidsthed om deltagelse og bidrager med designprincipper til fremtidige stemmebaserede AI-faciliteringsværktøjer. Begrænsningerne peger på behov for bedre svartiming, taleridentifikation, kontekstforståelse og længerevarende afprøvninger i større grupper.
[This apstract has been generated with the help of AI directly from the project full text]