跳转至

文章背景与核心概要

本文探讨了在无人机（UAV）入侵检测场景中，对话式可解释AI（Conversational XAI）与传统基于仪表盘的可解释AI（Dashboard-based XAI）的实际效果。尽管基于机器学习的入侵检测系统（IDS）功能强大，但其“黑箱”属性给操作员带来了显著的可解释性挑战。

研究发现，虽然由大语言模型（LLM）驱动的对话式界面在易用性和信息综合方面表现更佳，但也引入了一个关键风险：不恰当的过度依赖（over-reliance）。自然语言的交互形式容易让操作员过于顺从地接受AI建议，从而在系统出现错误时降低了他们核查底层证据的倾向。作者最后指出，易用性与批判性监督之间存在一种权衡（trade-off），并建议未来的XAI设计中应引入“认知强制机制（cognitive forcing functions）”。

Conversational versus Dashboard Explainable AI for UAV Intrusion Detection: An Empirical Study of Operator Trust and Reliance

arXiv: 2608.10434 [cs.AI]
Submitted: 11 Aug 2026
Authors: Cong Chi Nguyen, Trang Mai Xuan, Vu-Duc Ngo, Kim-Ngan Thi Nguyen, Trong-Nghia Nguyen, Thien Van Luong

Summary

This study investigates the effectiveness of Conversational Explainable AI (XAI) versus traditional Dashboard-based XAI in the context of Unmanned Aerial Vehicle (UAV) intrusion detection. While machine learning-based Intrusion Detection Systems (IDS) are powerful, their "black-box" nature creates significant interpretability challenges for operators.

The research finds that while conversational interfaces—powered by Large Language Models (LLM)—are perceived as more useful and easier to synthesize, they introduce a critical risk: inappropriate over-reliance. The natural language format may lead operators to accept AI advice too readily, reducing their tendency to verify evidence when the system is incorrect. The authors conclude by highlighting a trade-off between usability and critical oversight, suggesting the need for "cognitive forcing functions" in future XAI designs.

Abstract

摘要

Machine learning-based Intrusion Detection Systems (IDS) have demonstrated superior performance in securing Unmanned Aerial Vehicle (UAV) networks. However, the 'black-box' nature of these models, combined with the high dimensionality of multimodal cyber-physical data, poses significant interpretability challenges. Static visualization dashboards may struggle to present complex relationships among multimodal cyber-physical features in a form that is easy for operators to inspect and interpret.

基于机器学习的入侵检测系统（IDS）在保障无人机（UAV）网络安全方面表现出了优异的性能。然而，这些模型的“黑箱”特性，加上多模态信息物理数据的高维性，带来了巨大的可解释性挑战。静态的可视化仪表盘很难将多模态信息物理特征之间复杂的相互关系以一种便于操作员检查和理解的形式呈现出来。

To address this, we propose a Conversational XAI interface powered by Large Language Models (LLM) to facilitate on-demand investigation. In a controlled experiment with participants, we systematically evaluated the impact of this conversational interface versus a traditional XAI Dashboard on operator understanding, trust, and reliance during post-incident auditing tasks. Our results suggest that the conversational interface was perceived as more useful than the dashboard, potentially because it helped participants access and synthesize relevant information more easily. However, this benefit was accompanied by a lower level of appropriate self-reliance, indicating a potential risk of over-reliance. One possible interpretation is that the natural-language responses made the AI advice easier to accept, which may have reduced participants' tendency to verify the underlying evidence when the IDS was incorrect. These findings point to a potential trade-off in human-AI collaboration for UAV intrusion auditing: interaction mechanisms that improve perceived usability may also increase the risk of inappropriate reliance. We conclude by discussing design implications for future XAI systems that balance seamless interaction with cognitive forcing functions to foster appropriate reliance.

为了解决这一问题，我们提出了一种由大语言模型（LLM）驱动的对话式XAI界面，以促进按需调查。通过针对参与者的对照实验，我们在事后审计任务中系统地评估了该对话式界面与传统XAI仪表盘相比，对操作员理解、信任和依赖性的影响。我们的结果表明，参与者认为对话式界面比仪表盘更有用，这可能是因为它帮助参与者更轻松地访问和综合相关信息。然而，这种优势伴随着较低的适当自我依赖水平，表明存在过度依赖的潜在风险。一种可能的解释是，自然语言回复使AI建议更容易被接受，当IDS不正确时，这可能降低了参与者核实底层证据的倾向。这些发现指出了无人机入侵审计中人机协作的一个潜在权衡：提高感知易用性的交互机制也可能增加不适当依赖的风险。最后，我们讨论了未来XAI系统的设计启示，即在平衡无缝交互与认知强制机制的同时，培养适当的依赖性。

Paper Metadata

论文元数据

Access & Resources

访问与资源

Citation Tools

引用工具