The scale of adoption surprised even the study team. Among thirty-nine surgeons and surgical trainees who completed a baseline survey before the workshop, 89.7 percent reported that they had already used AI tools such as ChatGPT or Copilot for scientific activities, with 35.9 percent using them occasionally, 35.9 percent frequently, and 17.9 percent systematically. Only 10.3 percent had never touched these tools. Language editing dominated at 74.4 percent, followed by manuscript structuring and literature summarization at 46.2 percent each, and data analysis or interpretation at 41 percent. More than 82 percent agreed or strongly agreed that AI improves scientific productivity, signaling that these tools are no longer experimental novelties but embedded instruments of daily academic work.

Yet the same survey revealed a striking asymmetry between enthusiasm and understanding. Before the workshop, 48.7 percent of respondents rated their knowledge of AI in scientific writing as basic, 35.9 percent as intermediate, 5.1 percent as advanced, and 10.3 percent reported none at all. Every single participant considered AI-generated errors or fabricated references at least possible, and 61.5 percent rated them as frequent or very frequent. Trust without verification was correspondingly low: 76.9 percent said they rarely or never trusted AI-generated scientific content without additional checking. Nearly 95 percent believed they needed more training in AI applied to scientific publishing, underscoring a widening educational gap between what surgeons do with these tools and what they know about them.

The workshop itself was deliberately compact. Over ninety minutes, two presentations addressed current applications of AI in scientific writing from the perspective of early-career surgeons, and the viewpoint of a scientific publisher and editorial process regarding AI use in manuscript preparation and peer review. Topics spanned technical limitations, hallucinations, fabricated references, bias, confidentiality, authorship, disclosure, peer-review implications, and the need for professional guidance. Clinical applications of AI in abdominal wall surgery were not the primary focus. Participants completed a ten-item baseline survey beforehand and a ten-item post-workshop survey immediately afterwards, using anonymous codes that allowed thirty-four responses to be matched across both time points.

The matched analysis produced the study’s most statistically robust finding. Using the Wilcoxon signed-rank test on the single ordinal knowledge item that was directly comparable between surveys, self-reported knowledge improved in 18 of 34 participants, or 52.9 percent, with a p-value of 0.0005. The median knowledge score rose from basic before the workshop to intermediate afterwards. Knowledge remained unchanged in 38.2 percent and was lower in 8.8 percent, a small subgroup the authors suggest may reflect differences in attendance, engagement, or a more critical reassessment of prior knowledge after the session. After the intervention, 68.3 percent of all post-workshop respondents rated their knowledge as intermediate or advanced, compared with roughly 41 percent at baseline.

Beyond raw knowledge scores, the workshop produced strong signals of intended responsible behavior. Ninety-two point seven percent of post-workshop respondents agreed or strongly agreed that they now better understood AI’s technical limitations, including hallucinations, bias, and fabricated references. Reference verification emerged as the single strongest post-workshop commitment: 97.6 percent said they would almost always or always verify references generated by AI, with 80.5 percent committing to always doing so. Support for disclosure of AI use in manuscripts reached 85.4 percent, with 70.7 percent specifically favoring clear regulation. And 87.8 percent considered specific guidelines from scientific societies to be recommended or essential, while 78 percent reported feeling moderately or highly prepared to use AI responsibly.

These attitudes align with a rapidly evolving policy landscape. The World Association of Medical Editors recommends that generative AI tools cannot be listed as authors because they cannot take responsibility for the work, and that their use should be disclosed when they assist with scholarly manuscripts. The Committee on Publication Ethics has issued similar guidance. The urgency is not theoretical: the JAMA Network has reported receiving manuscripts containing inaccurate or fabricated references and has updated its guidance to discourage the use of generative AI for generating or managing bibliographic references. Empirical studies have shown that large language models can produce references with low precision and high hallucination rates when prompted for systematic-review-type tasks, with error rates high enough to preclude unsupervised use for evidence synthesis.

Confidentiality emerged as another critical concern. At baseline, 89.7 percent of participants reported at least slight concern about uploading manuscripts or peer reviews to AI platforms, and 51.3 percent reported moderate or very high concern. This matters because peer-review materials are confidential and should not be uploaded to external AI systems unless explicitly permitted by journal policy. Reviews of AI in peer review describe potential efficiency gains in screening and quality control but emphasize risks to confidentiality, transparency, and the preservation of expert human judgment. The authors argue that these peer-review-specific issues deserve dedicated attention in future AI education for surgeons, many of whom serve as reviewers as well as authors.

The study also revealed the limits of a brief intervention. Only 41.5 percent of post-workshop respondents felt confident in identifying AI-related risks specific to abdominal wall surgery, with 51.2 percent answering only partially. The authors acknowledge that the workshop was designed to address responsible use in scientific publishing rather than clinical applications, and they suggest that case-based discussions covering hernia surgery, abdominal wall reconstruction, clinical prediction models, image analysis, and AI-assisted operative video assessment would be a natural next step. They also note several limitations: the sample was voluntary and modest, there was no control group, outcomes were self-reported and measured immediately, not all responses could be paired, and no demographic variables were collected, preventing subgroup analysis by career stage or editorial experience.

Despite these caveats, the study offers a practical model for integrating responsible AI training into surgical education at a moment when the technology is outpacing formal instruction. The authors conclude that scientific societies and journals are well positioned to lead this transition by producing specialty-specific guidance covering acceptable AI uses, required disclosure language, reference-verification workflows, confidentiality safeguards, authorship boundaries, peer-review restrictions, and minimum competencies for trainees. As surgeons become both users and producers of AI-related research, the gap between adoption and literacy will only widen unless deliberately addressed. Ninety minutes, this study suggests, is enough to start closing it.

Subject of Research: Impact of a brief AI education workshop on surgeons' knowledge and attitudes toward artificial intelligence in scientific publishing

Article Title: Impact of an AI workshop on knowledge and attitudes toward AI in scientific publishing among surgeons at an international abdominal wall surgery congress

Article References: Verdaguer-Tremolosa, M., Kotoreni, G., Hoggart, S., & López-Cano, M. (2026). Impact of an AI workshop on knowledge and attitudes toward AI in scientific publishing among surgeons at an international abdominal wall surgery congress. Journal of Abdominal Wall Surgery, 5, Article 17401. https://doi.org/10.3389/jaws.2026.17401

Image Credits: AI Generated

DOI: 10.3389/jaws.2026.17401

Keywords: artificial intelligence, generative AI, scientific publishing, surgical education, abdominal wall surgery, hernia surgery, peer review, research ethics, AI hallucinations, manuscript disclosure, medical education, European Hernia Society

The quantitative half of the study relied on a structured questionnaire administered online over fourteen weeks, covering demographics, patterns of AI use, and four psychological constructs: perceived learning outcomes, motivation and engagement, self-regulated learning, and digital literacy. Respondents rated their agreement with statements on five-point Likert scales, and the researchers computed composite scores for each construct before running descriptive statistics, Pearson correlations, independent-samples t-tests, and one-way analyses of variance. The sample had a mean age of 16.59 years, was 60.7 percent female, and 83.3 percent reported regular home internet access. Reliability was strong across most scales, with Cronbach’s alpha values exceeding 0.80 for the AI usage, learning outcomes, motivation, and self-regulation measures.

The usage data paint a picture of AI as an overwhelmingly mobile, chatbot-centered phenomenon. Some 74 percent of students accessed AI tools primarily through their smartphones, with tablets and laptops trailing far behind and not a single student reporting use of school computers. Chatbots dominated the tool landscape, with 80 percent of respondents reporting their use, followed by math problem-solving applications at 26 percent, writing assistants at 16.7 percent, and adaptive learning platforms at just 7.3 percent. The most common academic task was homework assistance, with a mean frequency rating of 3.31 on the five-point scale, while tools recommended by teachers scored lowest, suggesting that students are discovering and deploying AI largely on their own initiative rather than through institutional channels.

Perceptions of learning benefits were broadly positive but measured. A majority of students, 65.4 percent, agreed that AI improves their understanding of the subjects they study, roughly half said it provides clearer explanations than their textbooks and helps them correct mistakes, and about 48 percent credited AI with better grades or deeper skills such as critical thinking. Notably, only 32 percent admitted to depending on AI to the point of no longer attempting problems independently, a figure that hints at the tension running through the entire study. On the motivational side, 51.3 percent said AI increased their interest in certain subjects, 53.3 percent found homework less stressful with AI, and 42 percent reported greater classroom confidence after using the tools.

The correlation analysis revealed a tightly interwoven network of positive associations. Perceived learning outcomes correlated strongly with motivation and engagement, with a Pearson coefficient of 0.739, and moderately with self-regulation and digital literacy. Self-regulation and digital literacy were themselves strongly linked at 0.670. All associations were statistically significant at the 0.01 level. Group comparisons added nuance: no significant gender differences emerged on any scale, with effect sizes ranging from negligible to small, but educational level mattered for learning outcomes and motivation, where one-way ANOVAs produced significant effects across the three school levels studied, though post-hoc Tukey tests showed the pairwise differences were concentrated between specific grade cohorts.

The qualitative phase, involving semi-structured online interviews with twelve purposefully selected students, uncovered the study’s most striking finding: what the researchers call the laziness paradox. Half of the interviewed students reported that AI boosted their motivation by making difficult tasks feel manageable and building confidence before tests, while an equal half described the opposite effect, a slide into cognitive passivity when ready-made answers removed the need for independent effort. One respondent admitted, honestly, to feeling lazy when using AI because a complete answer arrives without struggle, while another warned that AI makes students lazy and erodes self-confidence. A third of the interviewees occupied an ambivalent middle ground, reporting that the motivational outcome depended entirely on their own self-regulatory discipline in any given moment.

Perhaps most surprising is how skeptical these teenagers are. Trust in AI outputs was overwhelmingly conditional: half of the interview cohort practiced what the researchers describe as triangulation, cross-checking AI responses against textbooks, official course notes, and teacher explanations before accepting them. Others compared outputs across competing AI platforms or consulted peers and instructors. This verification behavior suggests that traditional academic authorities retain epistemic primacy even in an AI-saturated study environment, and it represents an emergent form of AI literacy that goes beyond mere technical competence. Yet the skepticism had clear subject boundaries. Students trusted AI for language tasks such as grammar correction and rephrasing but expressed deep mistrust in mathematics and physics, where they reported hallucinated answers, methods that diverged from classroom instruction, and explanations pitched at a university level rather than their own developmental stage.

Ethical reasoning also featured prominently in the interviews. A majority of students, 58 percent, drew a firm internal line between using AI as a cognitive scaffold to understand lessons and using it to complete assignments dishonestly, and several expressed anxiety about being misjudged by teachers, wishing their instructors understood that they use AI to learn rather than to cheat. Teachers themselves emerged as powerful regulators of student behavior: 58 percent of respondents said their educators actively addressed AI in class, mostly steering students toward responsible, moderate use rather than absolute prohibition, though a minority of students perceived their teachers as simply hostile to the technology. The researchers argue that this pedagogical framing matters enormously, because students’ ethical frameworks appear to strengthen when teachers shift from prohibition toward critical, mediated guidance.

The study’s authors frame their results through constructivist learning theory and the Technology Acceptance Model, noting that perceived ease of use functions as a double-edged sword. When an AI tool is frictionless, learners can redirect cognitive bandwidth toward higher-order reflection and knowledge construction, but the same frictionlessness enables bypassing productive cognitive struggle altogether, shifting motivation from mastery-oriented learning toward mere task completion. Half of the interviewed students showed meta-cognitive awareness of this danger, with one observing that information retrieved easily is forgotten just as easily because it is never properly processed. The researchers also documented structural barriers beyond psychology: multimodal failures when students uploaded dense photographed documents, and a fundamental mismatch between the generic training corpora of mainstream models and the specific methods demanded by Moroccan school curricula.

The practical implications are pointed. Rather than restrictive bans, the authors advocate responsible integration: redesigning traditional homework into AI-supported critique activities in which students are evaluated on their ability to analyze, edit, and justify AI-generated content; teaching prompt engineering, source triangulation, and hallucination detection as explicit components of digital literacy curricula; and developing subject-specific prompting frameworks that acknowledge the different risk profiles of language versus STEM applications. Policymakers, they argue, should establish clear, context-sensitive guidelines and invest in teacher professional development rather than defaulting to punitive prohibition. The study’s limitations, a convenience sample from a single Moroccan region, self-reported perceptions rather than objective performance data, and a cross-sectional design that cannot establish causality, mean the findings are exploratory. But they deliver a clear message for educators worldwide: the educational value of generative AI is not inherent in the technology but contingent on pedagogical framing, learner agency, and the institutional norms that govern its daily use, and teenagers, left largely to their own devices, are already inventing the verification habits and ethical boundaries that schools have yet to formalize.

Subject of Research: Secondary school students' perceptions of AI-assisted learning, motivation, self-regulated learning, and digital literacy in Morocco

Article Title: Perceptions of AI-assisted learning and learner development among Moroccan secondary school students

Article References: Kandirou, R., El Youssfi, S., Elatlassi, M., Dahman, A., & El Hammoumi, M. M. (2026). Perceptions of AI-assisted learning and learner development among Moroccan secondary school students. Discover Education, 5(1), Article 1125. https://doi.org/10.1007/s44217-026-02109-1

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02109-1

Keywords: artificial intelligence in education, ChatGPT, secondary education, self-regulated learning, digital literacy, student motivation, generative AI, Morocco, mixed methods, academic integrity, AI hallucinations, perceived learning outcomes

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