Conversational Agents at Work: How Can HCI Help Facilitate Better Social Interaction in the Work place?
Authors
Hakonsen, Alexander Rath ; Ildstrøm, Thomas
Term
4. Term
Publication year
2022
Submitted on
2022-05-31
Pages
25
Abstract
This thesis examines how conversational agents (AI‑driven digital assistants that interact in natural language) can support social interaction in workplace meetings, focusing on adoption barriers and user needs. It addresses a gap in HCI research, where AI support systems have been studied mainly for intellectual tasks but rarely for social ones. We pose two questions: (1) What are the barriers to adopting human–AI support systems for social interaction in group meetings? and (2) What are end‑users’ needs and expectations for using a conversational agent in meetings? Methodologically, we conducted qualitative initial interviews with master’s students at Aalborg University to inform scenario design, followed by scenario‑based evaluations with relevant employees from a wind turbine manufacturing company. In total, five interviews and four scenario‑based evaluations were analyzed using Reflexive Thematic Analysis. Findings indicate that pervasive device use disrupts meeting communication; offloading routine tasks such as transcription and meeting recaps to an agent can improve efficiency; and features for time management and agenda support are highly valued. Features aimed at reducing miscommunication and enhancing inclusion received mixed reactions; inclusion is best facilitated by the meeting leader unless there is mutual agreement on the agent’s role, and augmenting the facilitator rather than directly acting on participants increases acceptance. We identified 14 challenges in total (eight in meeting facilitation and six in adopting conversational agents) and propose seven recommendations for HCI practitioners and researchers, emphasizing participant engagement, individual adaptability, and reducing menial work. These insights were shared with eight companies at a design exhibition and resonated with their experiences. The thesis offers practical guidance and helps address the research gap on AI support for social tasks in the workplace.
Dette speciale undersøger, hvordan samtaleagenter (AI‑drevne digitale assistenter, der interagerer i naturligt sprog) kan støtte sociale interaktioner i arbejdspladsmøder, med særligt fokus på barrierer for adoption og brugeres behov. Udgangspunktet er en identificeret forskningskløft i HCI-litteraturen, hvor AI-støttesystemer primært er studeret i intellektuelle opgaver, men sjældent i sociale. Vi formulerer to forskningsspørgsmål: (1) Hvilke barrierer findes for at adoptere Human‑AI-støttesystemer til social interaktion i gruppemøder? og (2) Hvilke behov og forventninger har slutbrugere til at anvende en samtaleagent i møder? Metodisk gennemførte vi kvalitative indledende interviews med kandidatstuderende fra Aalborg Universitet for at skabe scenarier, efterfulgt af scenariebaserede evalueringer med relevante ansatte fra en vindmølleproduktionsvirksomhed. I alt blev fem interviews og fire scenariebaserede evalueringer analyseret med Reflexive Thematic Analysis. Resultaterne peger på, at udbredt brug af personlige enheder under møder forstyrrer kommunikationen; at overdragelse af rutineprægede opgaver som transskription og opsummeringer til en agent kan øge effektiviteten; og at funktioner til tidsstyring og dagsordener er særligt efterspurgte. Funktionen til at reducere misforståelser og fremme deltagelse blev mødt med blandede reaktioner; inklusion bør som udgangspunkt varetages af mødelederen, medmindre der er gensidig aftale om agentens rolle, og at styrke mødelederen med AI frem for at påvirke deltagerne direkte øger accepten. Vi identificerede i alt 14 udfordringer (otte ved mødefacilitering og seks ved CA-adoption) og formulerede syv anbefalinger til HCI-praktikere og forskere om bl.a. deltagerinddragelse, individuel tilpasning og reduktion af trivialt arbejde. Disse fund blev drøftet med otte virksomheder på en designeksponering og afspejlede deres egne erfaringer. Specialet bidrager med praktiske retningslinjer og belyser den undersøgte forskningskløft om AI-støtte til sociale opgaver i arbejdslivet.
[This apstract has been generated with the help of AI directly from the project full text]