Javascript must be enabled for the correct page display

How Socially Aware Are Large Language Models? A Scoping Review With Implications for Education

Shea, Benjamin Michael (2026) How Socially Aware Are Large Language Models? A Scoping Review With Implications for Education. Master thesis, Educational Sciences.

[img] Text
22052026SheaBMSThesis.pdf
Restricted to Repository staff only

Download (947kB)

A thesis is an aptitude test for students. The approval of the thesis is proof that the student has sufficient research and reporting skills to graduate but does not guarantee the quality of the research and the results of the research as such, and the thesis is therefore not necessarily suitable to be used as an academic source to refer to. If you would like to know more about the research discussed in this thesis and any publications based on it, to which you could refer, please contact the supervisor mentioned.


Abstract

Abstract This scoping review examines social reasoning and interaction in relation to Yang et al.’s (2025) framework of social awareness in Large Language Models (LLMs) and Natural Language Processing (NLP) research. Education is framed as a language-mediated system of evaluation through which LLMs expose institutional reliance on discursive outputs rather than access to cognition. 1,312 records (2021–2025) were screened across four databases (ERIC, Scopus, IEEE Xplore, and ACM Digital Library), yielding 18 studies involving multi-turn or multi-agent interaction. Using Yang et al.’s (2025) framework of social factors, interaction, and implication, the analysis suggests that social awareness is not a unified construct but varies across roles, environments, and mechanisms. Theory of Mind appears as one pathway for operationalizing social reasoning and is inconsistently applied. Within simulations, socially relevant behavior is primarily evaluated through task-oriented metrics, with limited attention to interactional quality. While social factors and interaction are commonly represented across studies, downstream societal and institutional consequences receive comparatively limited attention. Overall, the field lacks a unified framework, and evaluation practices raise questions about how interactional claims are assessed as LLMs increasingly shape educational evaluation, institutional legitimacy, and pedagogical authority. Keywords: Large Language Models, social awareness, multi-agent systems, education

Item Type: Thesis (Master)
Supervisor name: Strijbos, J.W. and Brouwer, J.
Degree programme: Educational Sciences
Differentiation route/Track: Onderwijskunde [Master Educational Sciences]
Date Deposited: 22 Jun 2026 13:52
Last Modified: 22 Jun 2026 13:52
URI: http://gmwpublic.studenttheses.ub.rug.nl/id/eprint/6709

Actions (login required)

View Item View Item