2026-10-09T19:04:14Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34021062 2026-09-29T06:33:44Z category_29122 category_29146 item_type_3 month_year_09_2026

Building a digital-footprint-based typology of social agents in social media: A mixed-methods study of behavioral capabilities and information ecosystem roles Xiwei WANG (22635581) Mali LI (22632275) Yuhang Wu (23675442) Jiale Chen (25135348) Human information behaviour Social and community informatics social agents ideal-type typology public digital footprints human-AI collaboration mixed methods research <p dir="ltr">Generative large language models (LLMs) are reshaping social agents in ways that challengetraditional human–bot classifications, while existing typologies often rely on difficult-to-observe intentions and provide limited guidance for analyzing public behavioral traces. This study adopts an exploratory sequential mixed-methods design to develop and empirically examine an ideal-type typology of social agents based on public digital footprints. Through a systematic literature review and coding analysis, Study 1 contextualizes the algorithmic conduit brokerage framework and derives two dimensions—content production capability and dissemination interaction capability—whose intersection yields four ideal-type prototypes. Study 2 examines their empirical correspondence using 157 community-verified social agent accounts from the Reddit BotRank dataset. A human–AI-assisted coding procedure achieved substantial agreement among three independent expert coders (Fleiss’ κ = 0.8484, based on initial unreconciled account-type classifications). The results reveal asymmetric capability configurations: Amplifier agents account for 60.5% of the sample, while 9.6% of accounts could not be confidently assigned to a single prototype based on the available public evidence. These findings show the analytical value of characterizing social agents through observable capability configurations rather than binary human–bot distinctions or inferred intentions. The study provides a theoretically grounded and empirically applicable framework for examining heterogeneous social agents while highlighting the need for longitudinal and cross-platform validation.</p> 2026-09-29T06:33:44Z Dataset Dataset 10.6084/m9.figshare.34021062.v1 https://figshare.com/articles/dataset/Building_a_digital-footprint-based_typology_of_social_agents_in_social_media_A_mixed-methods_study_of_behavioral_capabilities_and_information_ecosystem_roles/34021062 CC0