Back to search

Institutional Repository of Vilnius University

Socialinio darbo praktikai pritaikytų užklausų kūrimas didiesiems kalbos modeliams /

Abstract

dc:description

With the rapid development of artificial intelligence technologies, Large Language Models (LLMs) are increasingly integrated into the public sector and social services to optimize administrative processes and manage resource shortages. However, this technological shift poses fundamental challenges to the identity of the social work profession. Commercial general-purpose models (such as ChatGPT-4) often fail to meet professional accuracy standards, generate "hallucinations", present data confidentiality and privacy risks, and lack an understanding of the specific context of social work practice. Existing academic literature reveals a distinct gap in empirical research, showing a lack of analysis on how to apply AI within direct practice and how to formulate specialized prompts that respond to the actual professional needs and values of social workers across all stages of service delivery. To address this gap, the object of this research is the development of LLM prompts based on the professional needs of social workers. In order to examine the context of social work practice and construct prompt templates, an interdisciplinary approach was applied, combining qualitative ethnographic research methodology with the principles of prompt engineering. For the empirical part of the study, a qualitative method was chosen, consisting of 10 ethnographic interviews with social work professionals who possess at least 2 years of experience working with specific client groups. The research instrument was designed according to J. Spradley's ethnographic interview methodology, combining three types of questions: descriptive, structural, and contrast questions. The interviews were conducted within the premises of non-governmental organizations, strictly adhering to ethics and anonymity requirements, while the gathered audio recordings were transcribed and analyzed using the qualitative data analysis software MAXQDA. During the ethnographic analysis, core domains of cultural knowledge and semantic relationships were successfully extracted. The empirical findings revealed that despite the variations in specific client groups, social workers identify a universal five-stage structure of the helping process: problem identification, plan development, action, management of unforeseen situations, and evaluation of the results. The study documented detailed goals for each stage, actions performed by both workers and clients, encountered challenges, professional knowledge, skills, values, personal traits, and used working tools. Based on these comprehensive ethnographic findings and prompt engineering techniques, a specialized prompt template for social workers was constructed. The developed template incorporates five essential components of an effective LLM prompt: a precisely formulated instruction, a professional role, contextual background information, practical examples, and requirements for output formatting.

Degree

thesis:*
Grantor dc:publisher
Institutional Repository of Vilnius University
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Sas, Artūras,

Rights

dc:rights
Statement dc:rights
  • info:eu-repo/semantics/openAccess
Language dc:language
lit

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
Vilnius University
Base URL
epublications.vu.lt/oai
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Sas, Artūras,. Socialinio darbo praktikai pritaikytų užklausų kūrimas didiesiems kalbos modeliams /. Institutional Repository of Vilnius University, 2026. https://repository.vu.lt/VU:ELABAETD308329064&prefLang=en_US