[論文レビュー] SoulSeek: Exploring the Use of Social Cues in LLM-based Information Seeking

Yubo Shu, Peng Zhang+9 more|arXiv (Cornell University)|Jan 3, 2026

Information Retrieval and Search Behavior被引用数 0

ひとこと要約

This paper develops SoulSeek to integrate social cues into LLM-based search on a social content platform and evaluates how cues affect user perceptions, experience, and information-seeking behavior.

ABSTRACT

Social cues, which convey others' presence, behaviors, or identities, play a crucial role in human information seeking by helping individuals judge relevance and trustworthiness. However, existing LLM-based search systems primarily rely on semantic features, creating a misalignment with the socialized cognition underlying natural information seeking. To address this gap, we explore how the integration of social cues into LLM-based search influences users' perceptions, experiences, and behaviors. Focusing on social media platforms that are beginning to adopt LLM-based search, we integrate design workshops, the implementation of the prototype system (SoulSeek), a between-subjects study, and mixed-method analyses to examine both outcome- and process-level findings. The workshop informs the prototype's cue-integrated design. The study shows that social cues improve perceived outcomes and experiences, promote reflective information behaviors, and reveal limits of current LLM-based search. We propose design implications emphasizing better social-knowledge understanding, personalized cue settings, and controllable interactions.

研究の動機と目的

• Identify which social cues users expect to use in LLM-based information seeking on social content platforms.
• Design and implement a cue-aware LLM-based search prototype (SoulSeek) based on user workshop findings.
• Empirically evaluate how social cues influence outcome metrics and user processes during information seeking.

提案手法

• Conduct design workshops to extract social cue types and integration approaches from users of a social platform (RedNote/Xiaohongshu).
• Implement SoulSeek with social cue extraction, cue-aware query refinement, and cue-based matching and generation using an open-source LLM (Qwen-Max) within the Coze framework.
• Perform a between-subjects empirical study comparing a cue-enabled prototype versus a cue-naïve system.
• Collect mixed qualitative and quantitative data (questionnaires, think-aloud protocols, interviews) and analyze with thematic analysis and statistics.

実験結果

リサーチクエスチョン

• RQ1 RQ1: What social cues do users expect to use, and how should these cues be utilized in LLM-based search on social platforms?
• RQ2 RQ2: How does integrating social cues affect users’ information seeking outcomes and processes?

主な発見

Category	Dimension	C Mean	C SD	E Mean	E SD	Delta	p-value
Perception	trustworthiness	3.50	0.83	4.55	0.60	+1.05	0.0002
Perception	usefulness	2.75	0.85	3.85	1.09	+1.10	0.0011
Experience	user control	2.60	0.88	4.20	0.52	+1.60	0.0001
Experience	sense of direction	3.05	0.94	4.50	0.61	+1.45	0.0001
Experience	serendipity	2.20	0.83	4.00	0.86	+1.80	0.0001
Experience	willingness to use	2.55	0.83	4.20	0.62	+1.65	0.0001

• Social cues improved perceived usefulness and trustworthiness of results.
• Cue-enabled systems increased user control, sense of direction, serendipity, and willingness to use.
• Users reported enhanced perception, experience, and reflective information behaviors, while also revealing gaps in model understanding of social knowledge.