Conversational search interfaces, like ChatGPT, offer an interactive, personalized, and engaging user experience compared to traditional search. On the downside, they are prone to cause overtrust issues where users rely on their responses even when they are incorrect. What aspects of the conversational interaction paradigm drive people to adopt it, and how it creates personalized experiences that lead to overtrust, is not clear. To understand the factors influencing the adoption of conversational interfaces, we conducted a survey with 173 participants. We examined user perceptions regarding trust, human-likeness (anthropomorphism), and design preferences between ChatGPT and Google. To better understand the overtrust phenomenon, we asked users about their willingness to trade off factuality for constructs like ease of use or human-likeness. Our analysis identified two distinct user groups: those who use both ChatGPT and Google daily (DUB), and those who primarily rely on Google (DUG). The DUB group exhibited higher trust in ChatGPT, perceiving it as more human-like, and expressed greater willingness to trade factual accuracy for enhanced personalization and conversational flow. Conversely, the DUG group showed lower trust toward ChatGPT but still appreciated aspects like ad-free experiences and responsive interactions. Demographic analysis further revealed nuanced patterns, with middle-aged adults using ChatGPT less frequently yet trusting it more, suggesting potential vulnerability to misinformation. Our findings contribute to understanding user segmentation, emphasizing the critical roles of personalization and human-likeness in conversational IR systems, and reveal important implications regarding users' willingness to compromise factual accuracy for more engaging interactions.

The capacity of Large Language Models (LLMs) to reason is fundamental to their application in complex, knowledge-intensive domains. In real-world scenarios, LLMs are often augmented with external information that can be helpful, irrelevant, or even misleading. This paper investigates the causal impact of such auxiliary information on the reasoning process of LLMs with explicit step-by-step thinking capabilities. We introduce SciAux, a new dataset derived from ScienceQA, to systematically test the robustness of the model against these types of information. Our findings reveal a critical vulnerability: the model's deliberative "thinking mode" is a double-edged sword. While helpful context improves accuracy, misleading information causes a catastrophic drop in performance, which is amplified by the thinking process. Instead of conferring robustness, thinking reinforces the degree of error when provided with misinformation. This highlights that the challenge is not merely to make models "think", but to endow them with the critical faculty to evaluate the information upon which their reasoning is based. The SciAux dataset is available at https://huggingface.co/datasets/billhdzhao/SciAux.

Large language models (LLMs) have demonstrated exceptional proficiency in language understanding. However, when LLMs align their outputs with deceptive and / or misleading prompts, the generated responses could deviate from the de facto information. Such observations are known as fawning hallucinations, where the model prioritizes alignment with the input's implied perspective over accuracy and truthfulness. In this work, we analyze fawning hallucinations in various natural language processing tasks and tailor the so-termed contrastive decoding method for fawning-hallucination mitigation. Specifically, we design two paradigms to generate corresponding deceptive and / or misleading inputs for the consistent fawning hallucinations induction. Then, we propose the collaborative contrastive decoding (CCD) to handle the fawning hallucinations across di ff erent tasks in LLMs. By contrasting the deviation in output distribution between induced and transformed neutral inputs, the proposed CCD can reduce reliance on deceptive and / or misleading information without requiring additional training. Extensive experiments demonstrate that the proposed CCD can e ff ectively mitigate fawning hallucinations and improve the factuality of the generated responses over various tasks. Introduction Large language models (LLMs), exemplified by the Chat-GPT series [1], have demonstrated their remarkable capabilities across a wide range of natural language processing (NLP) tasks. These tasks include text translation [2, 3], summarization [4, 5], and a ffective computing [6, 7, 8, 9], showcasing the versatility and impact of artificial intelligence (AI). Despite the impressive performance, LLMs are criticized for the potential to generate fabricated, inaccurate, or incorrect information. This phenomenon, known as "hallucination", hinders the further practical application of LLMs.

Mobile GUI agents are designed to autonomously execute diverse device-control tasks by interpreting and interacting with mobile screens. Despite notable advancements, their resilience in real-world scenarios where screen content may be partially manipulated by untrustworthy third parties remains largely unexplored. Owing to their black-box and autonomous nature, these agents are vulnerable to manipulations that could compromise user devices. In this work, we present the first systematic investigation into the vulnerabilities of mobile GUI agents. We introduce a scalable attack simulation framework AgentHazard, which enables flexible and targeted modifications of screen content within existing applications. Leveraging this framework, we develop a comprehensive benchmark suite comprising both a dynamic task execution environment and a static dataset of vision-language-action tuples, totaling over 3,000 attack scenarios. The dynamic environment encompasses 58 reproducible tasks in an emulator with various types of hazardous UI content, while the static dataset is constructed from 210 screenshots collected from 14 popular commercial apps. Importantly, our content modifications are designed to be feasible for unprivileged third parties. We evaluate 7 widely-used mobile GUI agents and 5 common backbone models using our benchmark. Our findings reveal that all examined agents are significantly influenced by misleading third-party content (with an average misleading rate of 28.8% in human-crafted attack scenarios) and that their vulnerabilities are closely linked to the employed perception modalities and backbone LLMs. Furthermore, we assess training-based mitigation strategies, highlighting both the challenges and opportunities for enhancing the robustness of mobile GUI agents. Our code and data will be released at https://agenthazard.github.io.

Large language models (LLMs) exhibit excellent performance in natural language processing (NLP), but remain highly sensitive to the quality of input queries, especially when these queries contain misleading or inaccurate information. Existing methods focus on correcting the output, but they often overlook the potential of improving the ability of LLMs to detect and correct misleading content in the input itself. In this paper, we propose a novel three-stage fine-tuning method that enhances the ability of LLMs to detect and correct misleading information in the input, further improving response accuracy and reducing hallucinations. Specifically, the three stages include (1) training LLMs to identify misleading information, (2) training LLMs to correct the misleading information using built-in or external knowledge, and (3) training LLMs to generate accurate answers based on the corrected queries. To evaluate our method, we conducted experiments on three datasets for the hallucination detection task and the question answering~(QA) task, as well as two datasets containing misleading information that we constructed. The experimental results demonstrate that our method significantly improves the accuracy and factuality of LLM responses, while also enhancing the ability to detect hallucinations and reducing the generation of hallucinations in the output, particularly when the query contains misleading information.

AI, especially Large Language Models (LLMs) like ChatGPT, have rapidly developed and gained widespread adoption in the past five years, shifting user preference from traditional search engines. However, the generative nature of LLMs raises concerns about presenting misinformation as fact. To address this, we developed a web-based application that helps K-12 students enhance critical thinking by identifying misleading information in LLM responses about major historical figures. In this paper, we describe the implementation and design details of the DoYouTrustAI tool, which can be used to provide an interactive lesson which teaches students about the dangers of misinformation and how believable generative AI can make it seem. The DoYouTrustAI tool utilizes prompt engineering to present the user with AI generated summaries about the life of a historical figure. These summaries can be either accurate accounts of that persons life, or an intentionally misleading alteration of their history. The user is tasked with determining the validity of the statement without external resources. Our research questions for this work were:(RQ1) How can we design a tool that teaches students about the dangers of misleading information and of how misinformation can present itself in LLM responses? (RQ2) Can we present prompt engineering as a topic that is easily understandable for students? Our findings highlight the need to correct misleading information before users retain it. Our tool lets users select familiar individuals for testing to reduce random guessing and presents misinformation alongside known facts to maintain believability. It also provides pre-configured prompt instructions to show how different prompts affect AI responses. Together, these features create a controlled environment where users learn the importance of verifying AI responses and understanding prompt engineering.

Ensuring that Multimodal Large Language Models (MLLMs) maintain consistency in their responses is essential for developing trustworthy multimodal intelligence. However, existing benchmarks include many samples where all MLLMs exhibit high response uncertainty when encountering misleading information, requiring even 5-15 response attempts per sample to effectively assess uncertainty. Therefore, we propose a two-stage pipeline: first, we collect MLLMs' responses without misleading information, and then gather misleading ones via specific misleading instructions. Eventually, we establish a Multimodal Uncertainty Benchmark (MUB) that employs both explicit and implicit misleading instructions to comprehensively assess the vulnerability of MLLMs across diverse domains. Our experiments reveal that all opensource and close-source MLLMs are highly susceptible to misleading instructions, with an average misleading rate exceeding 86%. To enhance the robustness of MLLMs, we further fine-tune all ...

The societal and economic significance of machine learning (ML) cannot be overstated, with many remarkable advances made in recent years. However, the operation of complex ML models is most often inscrutable, with the consequence that decisions taken by ML models cannot be fathomed by human decision makers. It is therefore of importance to devise automated approaches to explain the predictions made by complex ML models. This is the main motivation for eXplainable AI (XAI). Explanations thus serve to build trust, but also to debug complex systems of AI.

During 2023, the shape of politics to come appeared in a video. In it, Hillary Clinton – the former Democratic party presidential candidate and secretary of state – says: "You know, people might be surprised to hear me saying this, but I actually like Ron DeSantis a lot. I'd say he's just the kind of guy this country needs." It seems odd that Clinton would warmly endorse a Republican presidential hopeful. Further investigations found the video was produced using generative artificial intelligence (AI).

Large Language Models possess skills such as answering questions, writing essays or solving programming exercises. Since these models are easily accessible, researchers have investigated their capabilities and risks for programming education. This work explores how LLMs can contribute to programming education by supporting students with automated next-step hints. We investigate prompt practices that lead to effective next-step hints and use these insights to build our StAP-tutor. We evaluate this tutor by conducting an experiment with students, and performing expert assessments. Our findings show that most LLM-generated feedback messages describe one specific next step and are personalised to the student's code and approach. However, the hints may contain misleading information and lack sufficient detail when students approach the end of the assignment. This work demonstrates the potential for LLM-generated feedback, but further research is required to explore its practical implementation.