Explanations are often promoted as tools for transparency, but they can also foster confirmation bias; users may assume reasoning is correct whenever outputs appear acceptable. We study this double-edged role of Chain-of-Thought (CoT) explanations in multimodal moral scenarios by systematically perturbing reasoning chains and manipulating delivery tones. Specifically, we analyze reasoning errors in vision language models (VLMs) and how they impact user trust and the ability to detect errors. Our findings reveal two key effects: (1) users often equate trust with outcome agreement, sustaining reliance even when reasoning is flawed, and (2) the confident tone suppresses error detection while maintaining reliance, showing that delivery styles can override correctness. These results highlight how CoT explanations can simultaneously clarify and mislead, underscoring the need for NLP systems to provide explanations that encourage scrutiny and critical thinking rather than blind trust. All code will be released publicly.

In the context of AI-based decision support systems, explanations can help users to judge when to trust the AI's suggestion, and when to question it. In this way, human oversight can prevent AI errors and biased decision-making. However, this rests on the assumption that users will consider explanations in enough detail to be able to catch such errors. We conducted an online study on trust in explainable DSS, and were surprised to find that in many cases, participants spent little time on the explanation and did not always consider it in detail. We present an exploratory analysis of this data, investigating what factors impact how carefully study participants consider AI explanations, and how this in turn impacts whether they are open to changing their mind based on what the AI suggests.

Evaluating the efficiency of human-AI interactions is challenging, including subjective and objective quality aspects. With the focus on the human experience of the explanations, evaluations of explanation methods have become mostly subjective, making comparative evaluations almost impossible and highly linked to the individual user. However, it is commonly agreed that one aspect of explanation quality is how effectively the user can detect if the predictions are trustworthy and correct, i.e., if the explanations can increase the user's appropriate trust in the model. This paper starts with the definitions of appropriate trust from the literature. It compares the definitions with model performance evaluation, showing the strong similarities between appropriate trust and model performance evaluation. The paper's main contribution is a novel approach to evaluating appropriate trust by taking advantage of the likenesses between definitions. The paper offers several straightforward evaluation methods for different aspects of user performance, including suggesting a method for measuring uncertainty and appropriate trust in regression.

Intelligent systems, especially those with an embodied construct, are becoming pervasive in our society. From chatbots to rehabilitation robotics, from shopping agents to robot tutors, people are adopting these systems into their daily life activities. Alas, associated with this increased acceptance is a concern with the ethical ramifications as we start becoming more dependent on these devices [1]. Studies, including our own, suggest that people tend to trust, in some cases overtrusting, the decision-making capabilities of these systems [2]. For high-risk activities, such as in healthcare, when human judgment should still have priority at times, this propensity to overtrust becomes troubling [3].

As robots complement or replace human efforts with more regularity, people may assume that the technology can be trusted to perform its function effectively and safely. Yet designers, users, and others must evaluate this assumption in a systematic and ongoing manner. Overtrust of robots describes a situation in which a person misunderstands the risk associated with an action because the person either underestimates the loss associated with a trust violation; underestimates the chance the robot will make such a mistake; or both. We deliberately use the term "trust" to convey the notion that when interacting with robots, people tend to exhibit similar behaviors and attitudes found in scenarios involving human-human interactions. Placing one's trust in an "intelligent" technology is a growing phenomenon.