Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

En utvärdering av AI-chattboten Albas användarupplevelse och korrekthet i diagnostisering av ADHD/ADD

Almgren, Andreas LU ; Nilsson, Emelie LU and Kindbom, Erik LU (2025) PSYK12 20242
Department of Psychology
Abstract
This study explored the effectiveness and user experience of the AI
chatbot Alba as a diagnostic tool for mental health, focusing on ADHD/ADD. The
research aimed to assess Alba’s ability to gather relevant information, engage
users, and provide accurate diagnoses. A mixed-methods 2x2 factorial design
involved 160 participants recruited via Prolific, divided into experimental
(ADHD/ADD and mental health issue) and control (neurotypical, no mental
health issues) groups. Participants were randomly assigned to one of four
conditions: voice vs. text communication and RAG vs. no-RAG models. Key
results indicated no significant difference between RAG and no-RAG models
(34% vs. 40% diagnosis accuracy). Text-based communication was rated... (More)
This study explored the effectiveness and user experience of the AI
chatbot Alba as a diagnostic tool for mental health, focusing on ADHD/ADD. The
research aimed to assess Alba’s ability to gather relevant information, engage
users, and provide accurate diagnoses. A mixed-methods 2x2 factorial design
involved 160 participants recruited via Prolific, divided into experimental
(ADHD/ADD and mental health issue) and control (neurotypical, no mental
health issues) groups. Participants were randomly assigned to one of four
conditions: voice vs. text communication and RAG vs. no-RAG models. Key
results indicated no significant difference between RAG and no-RAG models
(34% vs. 40% diagnosis accuracy). Text-based communication was rated more
positively in terms of elegance and perceived humanity than voice-based
interactions, which were often seen as mechanical. The no-RAG model was
perceived as more understandable and more precise than the one with RAG.
Thematic analysis identified empathy, communication, and perceived humanity as
critical factors influencing user experience. While Alba’s accessibility and
non-judgmental nature were appreciated, participants highlighted its lack of
emotional depth compared to human clinicians. The results suggest that AI-driven
chatbots like Alba hold promise as supplementary diagnostic tools in psychiatry. (Less)
Abstract (Swedish)
Denna studie undersökte effektiviteten och användarupplevelsen av AI-chatboten Alba som ett diagnostiskt verktyg för mental hälsa, med fokus på ADHD/ADD. Forskningen syftade till att utvärdera Albas förmåga att samla relevant information, engagera användare och ge korrekta diagnoser. En mixed-methods 2x2 faktoriell design med 160 deltagare rekryterades via Prolific, indelade i experimentgrupp (ADHD/ADD och psykisk ohälsa) och kontrollgrupp (neurotypiska, utan psykisk ohälsa). Deltagarna tilldelades slumpmässigt en av fyra betingelser: röst- eller textkommunikation och RAG eller utan-RAG modell. Resultaten visade ingen signifikant skillnad mellan RAG- och utan-RAG modellerna (34 % respektive 40 % diagnossäkerhet). Textbaserad kommunikation... (More)
Denna studie undersökte effektiviteten och användarupplevelsen av AI-chatboten Alba som ett diagnostiskt verktyg för mental hälsa, med fokus på ADHD/ADD. Forskningen syftade till att utvärdera Albas förmåga att samla relevant information, engagera användare och ge korrekta diagnoser. En mixed-methods 2x2 faktoriell design med 160 deltagare rekryterades via Prolific, indelade i experimentgrupp (ADHD/ADD och psykisk ohälsa) och kontrollgrupp (neurotypiska, utan psykisk ohälsa). Deltagarna tilldelades slumpmässigt en av fyra betingelser: röst- eller textkommunikation och RAG eller utan-RAG modell. Resultaten visade ingen signifikant skillnad mellan RAG- och utan-RAG modellerna (34 % respektive 40 % diagnossäkerhet). Textbaserad kommunikation fick högre betyg för dess elegans och upplevd mänsklighet än röstbaserade interaktioner, som ofta upplevdes som mekaniska. Utan-RAG modellen upplevdes som mer förståelig och mer precis än RAG-modellen. Tematisk analys identifierade empati, kommunikation och upplevd mänsklighet som viktiga faktorer för användarupplevelsen. Även om Albas tillgänglighet och icke-dömande natur uppskattades, betonade deltagarna en brist på emotionellt djup jämfört med mänskliga kliniker. Resultaten tyder på att AI-drivna chatbotar som Alba har potential som kompletterande diagnostiska verktyg inom psykiatri. (Less)
Please use this url to cite or link to this publication:
author
Almgren, Andreas LU ; Nilsson, Emelie LU and Kindbom, Erik LU
supervisor
organization
course
PSYK12 20242
year
type
M2 - Bachelor Degree
subject
keywords
AI-chatbot, ADHD/ADD-diagnos, användarupplevelse, röst-/text- kommunikation, Retrieval-Augmented Generation (RAG), ADHD/ADD diagnosis, user experience, voice/text communication
language
Swedish
id
9181966
date added to LUP
2025-06-17 14:23:57
date last changed
2025-06-17 14:23:57
@misc{9181966,
  abstract     = {{This study explored the effectiveness and user experience of the AI
chatbot Alba as a diagnostic tool for mental health, focusing on ADHD/ADD. The
research aimed to assess Alba’s ability to gather relevant information, engage
users, and provide accurate diagnoses. A mixed-methods 2x2 factorial design
involved 160 participants recruited via Prolific, divided into experimental
(ADHD/ADD and mental health issue) and control (neurotypical, no mental
health issues) groups. Participants were randomly assigned to one of four
conditions: voice vs. text communication and RAG vs. no-RAG models. Key
results indicated no significant difference between RAG and no-RAG models
(34% vs. 40% diagnosis accuracy). Text-based communication was rated more
positively in terms of elegance and perceived humanity than voice-based
interactions, which were often seen as mechanical. The no-RAG model was
perceived as more understandable and more precise than the one with RAG.
Thematic analysis identified empathy, communication, and perceived humanity as
critical factors influencing user experience. While Alba’s accessibility and
non-judgmental nature were appreciated, participants highlighted its lack of
emotional depth compared to human clinicians. The results suggest that AI-driven
chatbots like Alba hold promise as supplementary diagnostic tools in psychiatry.}},
  author       = {{Almgren, Andreas and Nilsson, Emelie and Kindbom, Erik}},
  language     = {{swe}},
  note         = {{Student Paper}},
  title        = {{En utvärdering av AI-chattboten Albas användarupplevelse och korrekthet i diagnostisering av ADHD/ADD}},
  year         = {{2025}},
}