Development of chatbots designed for low back pain triage
Author
Petersen, Jonas Asbjørn
Term
4. Semester
Publication year
2026
Submitted on
2026-06-01
Abstract
Chronic low back pain has substantial personal and societal impacts, and many individuals are unsure about the appropriate level of care to seek. Digital self-triage tools may offer guidance, but their use depends on user trust. This thesis examined whether the interaction style of an AI-based self-triage chatbot affects trust among people with chronic low back pain. In a two-condition, counterbalanced crossover experiment, 10 participants interacted with two chatbots that differed in style: deliberative versus paternalistic. The primary outcome was trust, measured with the Trust in AI-Generated Health Advice (TAIGHA) scale and analyzed using a linear mixed effects model adjusted for period, sequence, and baseline trust. The deliberative style increased trust significantly compared with the paternalistic style; however, the effect may be small, and a manipulation check indicated limited recognition of the independent variable. These findings suggest a deliberative interaction style may enhance trust more than a paternalistic style in digital self-triage chatbots for low back pain, though the practical importance remains uncertain and warrants further research.
Kroniske lænderygsmerter medfører store personlige og samfundsmæssige konsekvenser, og mange er usikre på, hvilket omsorgsniveau de bør søge. Digitale selv-triageværktøjer kan vejlede, men deres anvendelse afhænger af brugerens tillid. Denne afhandling undersøgte, om interaktionsstilen i en AI-baseret selv-triage-chatbot påvirker tilliden hos personer med kroniske lænderygsmerter. I et eksperiment med to betingelser og modbalanceret crossover-design interagerede 10 deltagere med to chatbots, der adskilte sig ved en deliberativ kontra en paternalistisk stil. Primært udfald var tillid, målt med Trust in AI-Generated Health Advice (TAIGHA), og analyseret med en lineær mixed effects-model justeret for periode, sekvens og baseline-tillid. Den deliberative stil øgede tilliden signifikant sammenlignet med den paternalistiske, men effekten kan være lille, og manipulationstesten viste begrænset genkendelse af den uafhængige variabel. Fundene antyder, at en deliberativ interaktionsstil kan styrke tilliden mere end en paternalistisk i digitale selv-triage-chatbots for lænderygsmerter, men den praktiske betydning er usikker og kræver yderligere forskning.
[This apstract has been generated with the help of AI directly from the project full text]