The research, scheduled for presentation on October 8 at the IEEE International Symposium on Mixed and Augmented Reality in Bari, Italy, is the first to systematically evaluate hallucination awareness cues for embodied conversational agents. These agents are AI-powered systems that go beyond plain text, presenting themselves as graphical characters capable of gestures, facial expressions, and natural speech. They are rapidly appearing in applications from customer service to education, and they inherit a well-known weakness from the generative AI models that power them: a tendency to produce untrue statements, commonly called hallucinations, delivered with the same fluent confidence as accurate information.

In conventional text-based chatbots, the problem of flagging uncertainty has a fairly mature toolkit. Verification systems can check an agent’s statements against trusted sources, and when something looks unreliable, the interface can highlight the relevant words or attach a source link for the user to examine later. Qiao Jin, corresponding author of the paper and an assistant professor of computer science at NC State, notes that these cues are relatively straightforward in text-based conversational interfaces. The challenge, he explains, is far thornier in immersive environments like VR, where conversational agents appear as human-like characters speaking directly to users, and where spoken responses pass by quickly, leaving no persistent trace to highlight or annotate.

To explore how uncertainty might be communicated in this new context, the researchers recruited 24 college students and had each of them interact with a VR system featuring an embodied conversational agent resembling a woman wearing a suit. Participants encountered multiple hallucinated statements under four different experimental conditions. In the first, the agent used embodied cues, making physical gestures associated with the unreliable statement, such as crossing its arms when it was uncertain. In the second, icon cues appeared next to the agent, notifying users that there was additional information they should check about a statement. In the third, text cues were used: every statement the agent made appeared as a subtitle on the screen, and those subtitles could be highlighted in different colors to signal potential uncertainty. The fourth condition served as a control, with no additional cues of any kind.

After each interaction, the team measured how well participants detected hallucinations and gathered detailed feedback through questionnaires and interviews, probing not only whether the warnings were understood but also how much each technique disrupted the participant’s sense of being present in the virtual world. That second dimension matters enormously, because immersion is the entire point of embodied agents. A warning system that constantly yanks users out of the experience defeats the purpose of building lifelike characters in the first place, as Xiaoran Yang, first author of the paper and a Ph.D. student at NC State, emphasizes: the evaluation looked both at which cues were most effective at alerting users to potential hallucinations and at which techniques best preserved the user’s sense of immersion in the VR experience.

The results revealed a genuine trade-off rather than a single winner. All three cue types outperformed the control condition at helping users identify hallucinations, confirming that some form of signaling is far better than silence. But each approach carried distinct strengths and weaknesses. Embodied cues, delivered through the agent’s own body language, were associated with the highest levels of immersion and the highest trust in the agent itself. Users remained inside the fiction of talking to a person, and the character’s gestures felt like a natural extension of human communication. The cost was interpretability: subtle physical signals were harder for participants to decode reliably, meaning some hallucinations slipped past even attentive users.

Text cues sat at the opposite end of the spectrum. By displaying subtitles that could be color-highlighted, the system gave users the clearest possible interpretation of which statements were questionable, making text cues the most effective technique for actually catching hallucinations. Yet this clarity came at a price. Participants reported that reading subtitles distracted them from everything else happening in the virtual environment, pulling their attention away from the agent, the setting, and the flow of conversation. The very mechanism that made the warnings legible also fragmented the immersive experience that embodied agents are designed to create.

Icon cues emerged as a pragmatic middle ground. A small icon appearing beside the agent conveyed more interpretable information than body language alone, while causing less disruption to immersion than a wall of highlighted subtitles. For developers who need a single, balanced solution, this middle path may prove attractive, offering meaningful gains in hallucination detection without the steep immersion penalty of on-screen text or the ambiguity of purely gestural signals.

The implications extend well beyond the laboratory. Educational programs, training simulations, and virtual assistants are among the most promising applications for embodied conversational agents, and these are precisely the contexts where a confidently stated falsehood can do real harm, teaching a student something incorrect or steering a user toward a bad decision. The study’s findings suggest that designers need not choose blindly among warning strategies. By selecting different cues, or combinations of cues, for different contexts, developers can tune the balance between communicative clarity and experiential continuity, improving a system’s ability to communicate clearly and effectively about potential hallucinations while limiting the impact on the user’s experience, as Jin observes.

There is also an important open problem that the researchers themselves flag: culture. Body language is not a universal language, and a gesture that reads as uncertainty in one cultural context may signal respect, deference, or something else entirely in another. Avoiding eye contact, for example, may indicate doubt in one setting but politeness in another. Before embodied cues can be deployed at scale, additional work will be needed to ensure that the body language of these signals can be understood by users from diverse cultural backgrounds. The paper, titled Signals of AI Hallucination: Designing Hallucination-Aware Cues for Embodied Conversational Agents in VR, was co-authored by Yang Zhan, Yuxuan Huang, Yichen Yu, and Noboru Matsuda of NC State, Xie He of Carnegie Mellon University, and Zhuo Wang of Xi’an Jiaotong-Liverpool University. As generative AI characters step out of the chat box and into our headsets, the question of how they should admit doubt may become one of the defining design challenges of immersive computing.

Subject of Research: Designing hallucination-aware warning cues for embodied conversational AI agents in virtual reality

Article Title: Researchers offer three ways to alert users to AI hallucinations with ‘embodied’ agents

Article References: Researchers offer three ways to alert users to AI hallucinations with ‘embodied’ agents. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: AI hallucinations, embodied conversational agents, virtual reality, human-computer interaction, generative AI, user experience, immersion, verification cues, NC State University, ISMAR, educational technology, uncertainty communication

The study, commissioned by the Impact Assessment Agency of Canada, is not a typical meta-analysis. Because almost no published studies rigorously test whether particular communication strategies actually work, the authors deliberately avoided the phrase best practices. Instead, they conducted a scoping narrative review, searching Google Scholar and the Omni academic database in January 2021 with fifteen distinct search strings designed to capture work on communicating scientific evidence, bridging ecology and decision-making, and translating environmental assessments into policy. After screening out articles that lacked relevance to environmental impact assessment or evidence communication, 78 peer-reviewed papers remained. From those, the team extracted 261 distinct approaches, which they then clustered, de-duplicated, and refined through collective author review and consultation with the agency that commissioned the work. The result is a curated inventory of approaches and concrete action items, organized around five recurring guiding principles that surfaced again and again across the literature.

The first guiding principle is collaboration, and the evidence behind it is striking. When stakeholders, rights holders, knowledge generators, and decision-makers are involved from the very planning phase of an assessment, rather than consulted after the fact, the resulting evidence is more relevant, more trusted, and more likely to be used. Co-production of knowledge, in which scientists and decision-makers jointly shape the questions being asked, builds trust in findings and ensures that the evidence generated actually maps onto the decisions at hand. The synthesis highlights practical mechanisms: using social media and internet platforms to broaden participation in planning, deploying virtual cloud environments to enable collaboration across all phases of an assessment, and deliberately weaving together local, Indigenous, strategic, and scientific knowledge domains. This transdisciplinary approach, the authors argue, enhances efficiency through adaptive feedback and supports better decision-making under uncertainty, a condition that defines nearly every environmental controversy.

The second principle, credibility, may be the most consequential, because any lingering doubt about the trustworthiness of an assessment makes uptake by decision-makers highly unlikely. The literature the team reviewed points to a battery of techniques for building and maintaining trust. Those conducting assessments should disclose their credentials and any conflicts of interest during the impact statement phase. Data collection should be standardized, and the ecological indicators used should be backed by expert consensus and scalable across contexts. Diverse research methods, drawing on ecology, sociology, anthropology, political science, and economics, provide a more complete picture than any single discipline can. Bayesian statistical methods, which formally incorporate prior knowledge and quantify uncertainty, can produce results that are more relevant and less ambiguous for decision-makers. Peer review of assessment documents, transparency about assumptions regarding data reliability, and open access to raw data consistent with open science frameworks all reinforce the perception, and the reality, that the evidence can be trusted.

Closely tied to credibility is the honest treatment of uncertainty, an area where environmental assessments have historically fallen short. The synthesis emphasizes that proponents should disclose uncertainty in predictions and statistical models using clearly defined processes, then communicate that uncertainty concisely and highlight the measures most relevant to policy. This is not a call for hedging that paralyzes decision-makers; rather, it reflects evidence from the science-policy literature that officials interpret and trust evidence better when its limits are stated plainly. The authors also flag the danger of prescriptive advocacy, in which scientific expertise is deployed to push for a particular policy outcome. Such advocacy narrows the choices available to decision-makers and erodes the neutrality that gives assessments their authority. Limiting bias, they suggest, can be strengthened by assembling diverse knowledge-generating teams and by using Bayesian network approaches that explicitly identify and bound uncertainty.

The third guiding principle addresses uptake directly: even credible, collaborative evidence is useless if decision-makers cannot access or understand it. Here the review reads almost like a communication manual for the modern assessment office. Decision-makers are more likely to consider evidence that is digestible, so the authors recommend participatory modelling, plain and contextually relevant language, visuals paired with text, and a portfolio of dissemination products ranging from executive summaries and briefing notes to infographics, maps, conceptual diagrams, and posters. Interactive formats and multi-platform delivery maximize reach. Trusted individuals matter enormously: knowledge brokers and boundary organizations can translate between communities, influential members within decision-making bodies can champion findings internally, and embedding subject-matter experts, including social scientists, directly into decision processes dramatically improves the odds that evidence shapes outcomes. Training matters too, both for decision-makers learning to interpret scientific uncertainty and for proponents learning how the decision-making process actually works.

Timing and capacity emerge as underrated variables in this third principle. Evidence that arrives after a decision window has closed is functionally worthless, so the synthesis urges that communication timelines be matched to decision-making timelines from the outset of planning. Research capacity, funding for timely and credible analysis, and analytical support at the science-policy interface all condition whether evidence can be produced and absorbed at the pace governance demands. The authors note that organizations should identify time and resource constraints early and plan their analytical approaches around them, rather than delivering technically perfect products that no one has the bandwidth to read. It is a pragmatic acknowledgment that the science-policy interface is staffed by humans working under deadlines, budgets, and political pressures.

The fourth and fifth principles extend the timeline of engagement beyond the moment of decision. Continued communication between knowledge generators and decision-makers throughout the assessment process builds the relationships on which trust depends, and ongoing dialogue helps shape assessments so their outputs remain relevant. Concrete mechanisms include recognizing and rewarding collaboration, co-locating scientists and officials in the same offices, providing technology that facilitates exchange, and convening workshops, round tables, and seminars throughout assessment development. After decisions are made, proper management of evidence becomes critical. Environmental conditions change, and findings that were current at the time of an assessment can quickly become obsolete. The authors recommend cataloguing assessments and communication materials in accessible, up-to-date online databases, maintaining strategic plans for keeping evidence current, treating projects as perpetually live rather than finished, and sharing results with all stakeholders and rights holders who contributed knowledge, using web-based networks to close the loop.

The synthesis is candid about its limits, and those limits are instructive. The literature base skews heavily toward North America and Europe, partly because the search was conducted in English, leaving large parts of South America, Africa, and Asia underrepresented, regions that often operate under different ecological, social, and political systems. The review also predates some recent literature, though the authors argue the recurring themes are robust. More fundamentally, the team acknowledges that context dictates everything: an approach that relies on internet technology to broaden participation makes little sense in remote locations with limited connectivity, and governance structures marked by instability or corruption will shape what is feasible. Integrating Indigenous knowledge systems into frameworks streamlined within a Western science paradigm remains a particular challenge, and the authors point to two-eyed seeing, the Mi’kmaq approach that gives equal value to Indigenous and Western knowledge systems, as a promising but still underused path forward. Perhaps the most striking finding is what the literature lacks: almost no studies have empirically tested whether any of these communication approaches actually work, a gap the authors identify as a major research agenda. What they offer, then, is a set of presumed rather than proven practices, a foundation that organizations worldwide can adapt to their own realities, guided by principles that transcend institutional boundaries and push environmental assessments closer to their original promise of putting the best available evidence at the center of environmental decisions.

Subject of Research: Communication of environmental evidence from impact assessments to decision-makers

Article Title: A narrative synthesis of approaches for communicating environmental evidence to decision-makers in the context of environmental impact assessments

Article References: Jeanson, A., Taylor, J., Rytwinski, T., Bennett, J., Young, N., Nguyen, V., Lin, H., Kelly, L., & Cooke, S. (2026). A narrative synthesis of approaches for communicating environmental evidence to decision-makers in the context of environmental impact assessments. Environmental and Sustainability Indicators, 32, Article 101537. https://doi.org/10.1016/j.indic.2026.101537

Image Credits: AI Generated

DOI: 10.1016/j.indic.2026.101537

Keywords: environmental impact assessment, science-policy interface, evidence-based decision-making, knowledge mobilization, environmental communication, Indigenous knowledge, credibility, uncertainty communication, stakeholder collaboration, environmental management, scoping review, sustainability indicators