Searcharxiv⌕ Search

arXiv subjects

Xiaofang Cai

Publications and source records attributed to Xiaofang Cai.

4 recordsLinked to original sources

Security Authentication as Protective Friction: Understanding the Dual Effects of Authentication Measures in e-Government Platform

E-Government platforms, such as government credit platforms, handle highly sensitive personal data, requiring stringent security authentication (e.g., facial recognition, multi-factor verification) that protects users but impedes their experience. Drawing on Tomalin's (2023) e-friction typology and the emerging literature on beneficial design friction, we reconceptualize these authentication measures as protective friction, a theoretically distinct form of non-elective, overt, impeding e-friction that serves a security-enhancing function. We propose a dual-pathway model in which protective friction simultaneously reduces privacy-security risk perception (protection pathway) and increases effort cost (cost pathway), with self-efficacy moderating only the risk pathway. This differential pattern suggests that what matters is not users' ability to overcome friction, but their comprehension of its protective purpose, a construct we term protective mechanism comprehension. We outline a mixed-methods research design combining a survey of platform users and semi-structured interviews with platform staff at a Chinese municipal credit center to test this model. This research-in-progress contributes to IS theory by extending e-friction research from conceptual typology to empirical investigation, identifying when security friction becomes beneficial rather than burdensome.

cs.CR↗

EvoProbe: Evaluation-Driven Benchmark Evolution for LLM Agents

Static benchmarks for large language model (LLM) agents face saturation and data contamination risks, motivating continual updates to evaluation tasks. However, aggregate scores and success rates offer limited insight into specific problem-solving bottlenecks and provide limited guidance for subsequent task design. We propose EvoProbe, an evaluation-driven benchmark evolution framework that integrates task generation, diagnostic evaluation, and feedback-driven regeneration into a closed loop. It combines verifiable subtask outcome scoring with evidence-supported multi-judge trajectory evaluation and distills phase-level bottlenecks, success patterns, and failure patterns into diagnostic feedback for generating new executable tasks. Across two rounds of cybersecurity experiments, evaluations of eight models on 66 tasks reveal phase-level performance differences and common failure patterns beyond aggregate scores. In a cross-round comparison with the Examiner model set and solver configuration held fixed, all eight models achieve lower Outcome scores on the updated tasks, with relative decreases of 11.8\% in mean Outcome and 40.0\% in mean Pass\textasciicircum{}3. These results support using fine-grained diagnostics to guide iterative task updates and probe agents' current capabilities and limitations.

cs.AI↗

People, IT, and Structuration (PIS): An Integrative Theoretical Framework for Management Information Systems

The Management Information Systems (MIS) discipline has long grappled with how to theorize the complex, mutually constitutive relationships among people, information technology, and organizational structures. Decades of research have produced influential but fragmented theoretical streams from socio-technical systems theory to technology acceptance models, from adaptive structuration theory to sociomateriality, and each illuminating important facets while leaving integrative questions unresolved. This paper proposes the People - IT - Structuration (PIS) framework as a unifying theoretical lens that synthesizes these streams. Drawing on Giddens' structuration theory, we conceptualize People (P), Information Technology (I), and Structure (S) not as independent variables but as mutually constitutive elements engaged in ongoing structuration processes. We trace the intellectual history of MIS theorizing to demonstrate how PIS resolves persistent tensions in the field,e.g. between technological and social determinism, between variance and process approaches, and between micro-level interaction and macro-level institutional dynamics. We develop a set of formal propositions articulating the mechanisms through which P, I, and S co-evolve, and extend the framework to address contemporary phenomena including artificial intelligence, algorithmic management, and human-AI collaboration. The PIS framework offers both a retrospective lens for understanding the discipline's theoretical evolution and a prospective tool for guiding research in the AI era.

cs.HC↗

League of LLMs: A Benchmark-Free Paradigm for Mutual Evaluation of Large Language Models

Although large language models (LLMs) have shown exceptional capabilities across a wide range of tasks, reliable evaluation remains a critical challenge due to data contamination, opaque operation, and subjective preferences. To address these issues, we propose League of LLMs (LOL), a novel benchmark-free evaluation paradigm that organizes multiple LLMs into a self-governed league for multi-round mutual evaluation. LOL integrates four core criteria (dynamic, transparent, objective, and professional) to mitigate key limitations of existing paradigms. Experiments on eight mainstream LLMs in mathematics and programming demonstrate that LOL can effectively distinguish LLM capabilities while maintaining high internal ranking stability (Top-$k$ consistency $= 70.7\%$). Beyond ranking, LOL reveals empirical findings that are difficult for traditional paradigms to capture. For instance, ``memorization-based answering'' behaviors are observed in some models, and higher in-family scores are found in the OpenAI model family ($Δ= 9$, $p < 0.05$). Finally, we make our framework and code publicly available as a valuable complement to the current LLM evaluation ecosystem.

cs.AI↗