Frontiers in Education: Digital Learning Innovations
1 day 8 hours ago
Danny Glick
2 days 10 hours ago
IntroductionDespite the growing use of generative artificial intelligence (AI) in English as a Foreign Language (EFL) education, how interactive AI features and personalized learning feedback shape learners' acceptance remains insufficiently understood. This study examines their influence on Egyptian university students' behavioral intention to use AI tools through an extended Technology Acceptance Model (TAM), focusing on the mediating roles of perceived usefulness and perceived ease of use.MethodsAn explanatory sequential mixed-methods design was employed. Survey data from 596 Egyptian university students were analyzed using structural equation modeling (SEM), followed by thematic analysis of semi-structured interviews with eight purposively selected participants.ResultsPerceived usefulness and perceived ease of use significantly predicted behavioral intention. Interactive AI features influenced intention indirectly through perceived usefulness, whereas personalized learning feedback influenced intention indirectly through perceived ease of use. Interviews revealed that interactivity supported engagement, clarification, and learner control, while personalized feedback facilitated error correction, reduced ambiguity, and alleviated anxiety. Participants viewed AI tools as flexible learning support rather than replacements for human instruction.DiscussionThe findings highlight the importance of pedagogically meaningful interaction and personalized feedback in shaping EFL learners' acceptance of AI tools. Responsible integration should emphasize learner agency, critical evaluation of AI-generated responses, and continued human pedagogical guidance.
Sultan Hammad Alshammari
2 days 11 hours ago
IntroductionGenerative artificial intelligence (AI) offers new forms of learning support, but assistance with a task should be distinguished from independent learning. This exploratory study examined questionnaire changes during a 16-week graduate Training and Development Seminar at a national university in Taiwan and considered students' written accounts of AI use.MethodsThe reported course sample comprised 10 master's and three doctoral students. Thirteen responses were available at each measurement occasion; eight participants could be matched by email address and formed the paired analysis sample. A 20-item SDLI-based self-directed learning (SDL) questionnaire and a 24-item ARCS course motivation questionnaire were administered in Chinese.ResultsAcross all responses, SDL means were 80.62 (SD = 10.24) at pretest and 87.38 (SD = 7.12) at posttest; motivation means were 98.08 (SD = 11.84) and 106.54 (SD = 9.61), respectively. In the eight verified pairs, mean changes were 7.25 and 9.88 points, respectively. Two-sided exhaustive signed-rank p-values were 0.015625 and 0.03125, with seven nonzero differences for each outcome. Cronbach's alpha was 0.895/0.864 for SDL and 0.941/0.900 for motivation at pretest/posttest. Written reflections described perceived learning support, professional relevance, and concerns about dependence on AI.DiscussionThe small matched subset, self-report measures, and absence of a comparison group prevent attributing the changes specifically to AI or generalizing them to the full cohort. The findings raise questions about learner judgment, output verification, and assessment of unaided understanding.
Yin-Che Chen
5 days ago
As generative artificial intelligence (AI) increasingly automates technical execution, the labour-market value of narrowly defined technical skills is widely argued to be shifting toward human capacities that remain difficult to replicate, including perspective-taking, critical judgment, ethical discernment, and meaning-making. This conceptual analysis, informed by the author's reflective teaching practice, challenges the marginal positioning of literary study within innovation education and argues that literary thinking can contribute a substantive foundation for human-centred innovation. Its primary purpose is theory-building: it develops and specifies a pedagogical framework rather than reporting an empirical evaluation. It is guided by the question: how can literary and philosophical practices be integrated into innovation education as a structured pedagogical sequence that complements design thinking and responds to the risks of uncritical reliance on generative AI? Drawing on the contested experimental and meta-analytic literature on fiction reading and social cognition, on Aristotle's and Heidegger's accounts of technē (craft), poiesis (making or bringing-forth), and phronēsis (practical wisdom), and on research on design thinking, design fiction, responsible innovation, and AI literacy, the article develops the Narrative-Humanistic Innovation Pedagogy (NHIP). NHIP integrates four interrelated pillars: Empathic Immersion, Critical Interrogation, Ethical Grey-Zone Reasoning, and Narrative Prototyping, each with a specified form of engagement with generative AI. Rather than displacing established design-thinking, design-fiction, and responsible-innovation approaches, the framework complements them by sequencing literary reading, Socratic deliberation, and narrative construction as practices proposed to support cognitive, ethical, and imaginative capacities relevant to innovation. NHIP was developed through the author's teaching of innovation classes at the University of Kalba, United Arab Emirates, in which its activities have been implemented as part of routine teaching but have not been formally evaluated. A ten-week illustrative module, with a generative-AI strand running through every pillar, shows how the framework could be operationalised in classroom practice. The article advances, as a hypothesis for empirical testing, the proposition that humanistic pedagogy may help safeguard against cognitive offloading arising from uncritical technology adoption. Testable propositions and implications for curriculum design, faculty development, and empirical validation are discussed.
A. Nasir Yousuf Abdelkareem
5 days 9 hours ago
IntroductionContemporary technological interventions in the pedagogical system have created substantial learning opportunities across educational communities. Learning theories provide a framework for effective learning practices. This study focuses on pedagogical synthesis frameworks that integrate generative artificial intelligence technologies and align with learning theories to enhance learning.MethodsThis study employed bibliometric and content analyses to examine research articles relating to the integration of generative AI technologies, guided by learning theories, within the educational system. A total of 614 articles were collected from the Scopus and Web of Science databases using the keywords “generative artificial intelligence” AND “education” AND “digital learning”. The data were visualised and interpreted through bibliometric analysis using the Biblioshiny application integrated with RStudio. For the content analysis, 163 articles were screened through a systematic review process.ResultsThe findings reveal influential countries, authors, and core sources in the field, as well as students' learning with artificial intelligence as an emerging theme, highlighting the prominence of generative AI in education. The penetration of generative artificial intelligence technology was higher at the higher education level (57.05%), followed by the professional level (36.80%) and the school level (6.13%).DiscussionThematic categorisation of generative AI research into adoption, academic learning, English language learning, ethical usage, and professional contexts reveals patterns in the adoption of generative AI and its determinants in the education domain. The findings also indicate a strong intrinsic link between generative AI-integrated learning approaches and the alignment of Constructivist, Connectivist, and Cognitive learning theories, particularly at the higher education and professional levels.
Saranya P
6 days 10 hours ago
IntroductionThe integration of artificial intelligence (AI) into workplaces is changing the capabilities expected of graduates, while higher-education curricula often adapt more slowly to labour-market developments. This article addresses the need for a structured, accountable process through which academic programmes can respond to AI-integrated work.MethodsThe study uses a purposive conceptual synthesis of literature on digital capability, labour-market information, constructive alignment, human capital, stakeholder participation, socio-technical systems, and responsible AI governance. Kenya is used as an illustrative context rather than as a site of empirical validation.ResultsThe synthesis produces a programme-level process framework comprising six stages: sensing, validation, diagnosis, redesign, implementation, and outcome review. These stages are supported by cross-cutting conditions involving data infrastructure, stakeholder participation, faculty capability, ethical governance, and academic authority. AI is positioned as a decision-support resource for interpreting labour-market signals and identifying potential competency gaps. However, stakeholder validation, academic prioritisation, formal authorisation, curriculum decisions, and outcome review remain subject to human and institutional accountability.ConclusionThe framework integrates previously separate strands of research into an evidence-to-model process with explicit decision gates and traceability requirements. It offers propositions for empirical testing and provides a basis for programme-level curriculum review and renewal. The framework does not establish that AI-enabled curriculum processes improve graduate employability; such effects must be evaluated through programme-level pilots and longitudinal studies.
Duncan Nyale
1 week ago
IntroductionArtificial intelligence in the form of Large Language Models is increasingly being incorporated into music education, yet much of the existing literature concentrates on composition, performance, assessment, creativity, and teacher adoption rather than music humanities education. This article examines student responses to Mystic Melodies, a gamified adaptive-difficulty AI quiz used in a compulsory first-year musicology unit at the Sydney Conservatorium of Music.MethodsStudents attended a one-hour lecture and then completed the quiz based on the lecture content during a one-hour tutorial. The quiz used short answer questions, adaptive difficulty, hints, immediate feedback, explanatory teaching points, and a quest narrative in which correct responses generated gifts while incorrect responses withheld them.ResultsThirty-four post-activity surveys comprising five-point Likert items and open-ended questions were received. The quantitative data show that students felt positive about how well they understood the topic, but were divided about all other aspects of their app experience. The qualitative data suggests an explanation for this pattern. Several critical comments point to concerns about humanities learning, expert teaching, independent thinking, meaningful feedback, and the limits of correct-incorrect quiz logic.DiscussionThis intervention highlighted a design problem: in music humanities education, AI-supported activities may be accepted as useful review or feedback tools, but they risk resistance when students perceive them as replacing expert-led, interpretive, or dialogic forms of learning.
Gavin S. K. Lee
1 week 5 days ago
BackgroundThe educational metaverse presents emerging opportunities for transforming university learning, yet students' willingness to adopt such environments may depend on psychological, technological and contextual factors. This study examined the predictors of university students' behavioural intentions to use the educational metaverse in Ghana, with particular attention to mediating and moderating mechanisms and gender differences.MethodsAn analytical survey design was employed, and data were collected from 917 university students selected through purposive sampling. Partial Least Squares Structural Equation Modelling (PLS-SEM) was used to examine the direct, mediating and moderating relationships among the study variables.ResultsPerceived observability, compatibility, cyber risk, personal innovativeness, trust and perceived usefulness were significantly associated with behavioural intention, whereas perceived complexity was negatively associated with behavioural intention. Perceived ease of use and enjoyment were not directly associated with behavioural intention. Trust significantly mediated the association between perceived usefulness and behavioural intention, while observability mediated the association between personal innovativeness and behavioural intention. However, enjoyment did not significantly mediate the association between perceived ease of use and behavioural intention. Gender significantly moderated the associations of compatibility, personal innovativeness, trust and perceived usefulness with behavioural intention, indicating differences in adoption patterns between male and female students. Cyber risk also moderated the association between trust and behavioural intention, weakening the positive relationship between the two constructs.ConclusionsThe findings suggest that educational metaverse adoption among university students is shaped by a combination of technological perceptions, personal innovativeness, trust and cyber risk, with several relationships operating through mediating and moderating mechanisms. The study extends educational metaverse adoption research by demonstrating the importance of examining these mechanisms rather than relying solely on direct predictors of behavioural intention. Universities and educational technology developers should therefore consider trust, cybersecurity and gender-related differences when designing and implementing metaverse-based learning environments.
Monica Ofosu-Koranteng
2 weeks ago
IntroductionThis study evaluates interactive digital practice modules for grammar acquisition in German as a foreign language, focusing on verb-complement patterns involving verb meaning, complement selection, case marking, and prepositional choice. Drawing on construction-based pedagogy, it examines digitally supported retrieval practice, self-paced learning, response checking, and structured comparison.MethodsThe study analyzed an archived dataset of 61 Arabic-speaking university learners who participated in a six-week quasi-experimental pretest-posttest design. The experimental group used interactive digital modules, while the control group received conventional instruction. Because the recovered test forms differed in length, scores were converted into percentages, although this adjustment could not ensure measurement equivalence.ResultsBoth groups demonstrated statistically significant improvements in percentage scores from pretest to posttest. However, the difference in learning gains between the experimental and control groups was not statistically significant.DiscussionThe findings suggest that grammar learning improved under both instructional conditions without establishing a significant advantage for the digital modules. Because digital delivery and pedagogical organization varied simultaneously, their independent contributions could not be determined. Furthermore, percentage-based score conversion did not address potential differences in test difficulty, content coverage, or measurement precision. The study highlights the importance of comparable assessment instruments, transparent outcome measures, and designs capable of distinguishing technological effects from pedagogical effects in digitally mediated second language grammar instruction.
Nadia Naqrash
2 weeks ago
IntroductionArtificial intelligence (AI) is transforming quality assurance in higher education by enabling adaptive, data-driven assessments of teaching performance and student learning. This integrative review synthesizes the purposes, methodologies, and outcomes of AI-enabled assessments, examines emerging technological trends, and identifies opportunities for improving teaching quality and learner support.MethodsAn integrative review was conducted using the Whittemore and Knafl framework, with study identification and selection reported in accordance with PRISMA guidelines. Of 367 records identified, 21 studies were included. Methodological quality was appraised using the Mixed Methods Appraisal Tool (MMAT). Descriptive and thematic analyses were supplemented by keyword network mapping using VOSviewer version 1.6.20.ResultsMost studies were multi-authored (71.43%) and published in interdisciplinary education and technology journals (57.14%), with publication activity concentrated in 2024 (28.57%) and 2025 (33.33%). The Western Pacific Region accounted for the largest proportion of studies (38.10%), while no studies originated from the African Region. Teacher performance was the predominant assessment focus (n = 11; 52.38%), followed by learner behavior (n = 4; 19.05%) and learner performance (n = 3; 14.29%). Teacher-focused assessments primarily analyzed learner-generated data using machine learning and natural language processing, whereas some learner-focused assessments employed video-based data and computer vision. Machine learning and natural language processing were the most frequently identified technological approaches (n = 6), followed by computer vision and multimodal or speech recognition approaches (n = 3). Most implementations followed a data-input, AI-processing, and output-generation structure, with varying levels of complexity. Thematic analysis identified four domains, while keyword network analysis revealed eight clusters.DiscussionThe findings highlight a fragmented assessment landscape in which teacher performance and learner behavior are predominantly examined independently. Although learner behavior may provide complementary evidence of teaching effectiveness, reciprocal relationships between teaching and learning remain insufficiently explored. Geographic disparities and methodological heterogeneity further constrain the generalizability and institutional applicability of existing evidence.ConclusionAI offers promising opportunities to strengthen teaching evaluation and learner support in higher education. However, broader geographic representation, methodological rigor, validated assessment frameworks, and equitable, human-supervised implementation are essential to ensure that institutional adoption is guided by robust evidence.Systematic Review Registrationhttps://osf.io/ugt4w.
Michael Joseph Dino
2 weeks ago
IntroductionArtificial intelligence (AI) is becoming increasingly relevant to forensic practice and has potential applications in forensic education, including evidence interpretation, simulation-based learning, and decision-making. Then, evidence regarding its educational implementation remains fragmented, with limited consensus on pedagogical frameworks and competency development.MethodologyA systematic review was conducted following PRISMA guidelines. Studies published between 2020 and 2025 were identified through Scopus, PubMed, and IEEE Xplore. Eligible studies for the primary synthesis were required to include an identifiable forensic educational or training context, an AI-supported educational intervention, or a measured educational/training outcome. Technical studies of AI applications in forensic practice were analyzed separately to identify emerging competency and curriculum requirements.ResultsOnly a limited subset of the retrieved literature directly examined AI within an educational or training context. The available educational evidence addressed areas such as learner skill development, professional training needs, and AI-supported learning activities, but remained heterogeneous and methodologically limited. In contrast, a much larger body of technical forensic AI research demonstrated rapid advances in image analysis, classification, digital forensics, and decision support. These developments identify emerging educational requirements in algorithmic literacy, validation, explainability, bias assessment, data governance, and human oversight.DiscussionCurrent evidence indicates that AI should complement rather than replace human expertise. Effective implementation requires competency-based curricula integrating computational literacy, ethical reasoning, interdisciplinary collaboration, and experiential learning supported by explainable AI.ConclusionDirect empirical evidence supporting AI-enhanced forensic education remains limited. This scarcity represents an important evidence gap. Rapid technical advances nevertheless provide a strong rationale for developing and prospectively evaluating competency-based AI education in forensic curricula.
María Belén Morales-Cevallos
2 weeks 5 days ago
Ahmed Lachheb
2 weeks 6 days ago
AI-mediated tools have expanded from specialised systems to generative and multimodal platforms, but research in primary English language education remains fragmented by technology, educational stage, and evidence type. This scoping review followed JBI methodology and PRISMA-ScR. Scopus, Web of Science Core Collection, and ERIC were searched with a cut-off of 21 September 2026 without publication-year limits. The synthesis included 119 reports, comprising 72 core empirical reports on primary English education, 31 adjacent empirical reports from related settings and 16 contextual sources. The synthesis organised the evidence into five recurring pedagogical functions. These covered oral interaction and speaking rehearsal, pronunciation and oral-reading feedback, and writing, vocabulary, and feedback scaffolding. They also included reading and multimodal meaning-making and teacher-facing assessment and materials design. AI-mediated tools supported structured language practice and feedback-guided activity. Reported performance differences varied across tasks, comparison conditions and outcome measures. Delayed outcomes varied, and transfer findings were tied to the tested material or task. Cross-study analysis developed an evidence-informed conceptual framework linking pre-task design and in-task regulation with post-task interpretation and subsequent instructional adjustment. The educational value of AI appears to vary according to task design, support during use, and professional interpretation. The review offers a function-based framework for selecting tools and an orchestration framework for connecting AI-mediated activity with classroom teaching.
Mengying Li
3 weeks ago
Generative AI has complicated how higher education distinguishes assistance from replacement in academic writing, and neither blanket prohibition nor permission resolves it, because the same GenAI-assisted practice may scaffold a student's reasoning or perform the work for students. This scoping review maps how empirical and systematically documented practice-based literature categorises responsible student GenAI use in academic writing, how these boundaries are translated into educational approaches, and what evidence exists for their evaluation and implementation. Following PRISMA-ScR, five databases were searched for English- and Spanish-language higher-education studies published between 1 January 2022 and 6 June 2026, supplemented by citation searching. Included studies reported original empirical data or first-hand documentation of an enacted practice, and contributed either to responsible-use boundaries or to a specified pedagogical assessment or guidance approach. In total, 31 studies were included. Charting yielded 192 practice-level records from 29 studies and 122 approach-level records from 29 studies. 27 studies contributed to both datasets. An interpretive synthesis of practice-level conditions, restrictions and explanatory notes identified six recurring principles: targeting learning objectives, preserving human agency, verifying GenAI output, transparency and accountability, authorship and human voice, and privacy, consent and fairness. These judgements are made at the micro level of the writing task but are sustained or eroded at the meso level of course design and the macro level of institutional guidance. Educational approaches were mainly pedagogical, but evaluation typically focused on short-term outcomes an relied on self-reporting. The included evidence suggests that responsible use depends on task purpose, expected intellectual contribution and the conditions surrounding AI assistance. The six principles offer a descriptive basis for developing and testing decision support. Further research could connect explicit boundaries with interventions and assess writing processes, textual outcomes and longer-term transfer.
Jiayu Tie
3 weeks ago
IntroductionTeachers' continuous professional development (CPD) is critical for effective digital technology integration in teaching and learning. This study explored how professional development initiatives support Grade 12 Business Studies teachers in integrating digital technologies in resource-constrained rural schools in South Africa.MethodsThe study adopted a qualitative case study design within the interpretivist paradigm. Data were generated through semi-structured interviews with six purposively sampled Grade 12 Business Studies teachers and analysed thematically.ResultsThe study findings revealed that teachers participated in formal continuous professional development initiatives, including departmental workshops and cluster meetings, as well as informal opportunities such as peer collaboration and self-directed online learning. However, most formal initiatives focused predominantly on technical skills rather than the pedagogical application of digital technologies. Consequently, teachers relied heavily on peer collaboration, professional learning communities and informal digital networks to address gaps in formal training.DiscussionThe study further highlights the need for sustained and contextually responsive professional development that aligns digital technologies with pedagogical practice. The study recommends evidence-based district support systems that integrate technical assistance with pedagogical mentorship to strengthen teachers' capacity to integrate digital technologies effectively in resource-constrained rural schools.
Siyabonga Andrias Magoso
3 weeks ago
The arrival of generative AI in every classroom has produced a crisis that is at once technological, epistemological, and pedagogical. We argue that the appropriate response is neither prohibition nor naive adoption, but a principled reconstruction of constructionism for a computational environment that is generative, context-sensitive, and dialogically accessible. Drawing on Peircean semiotics, Papertian constructionism, Polanyi's epistemology, and contemporary neural-network research, we develop a programmatic framework we call the New Constructionism. We make four connected contributions. First, we argue that generative AI introduces an indexical mode of computation: its relation to the empirical world may be direct, mediated, or incomplete, but in a situated human–AI configuration meaning is operationally conditioned by conversational history, artifacts, tools, observations, and action. Second, we introduce a three-dimensional analytical space — semiotic, psycho-epistemological, and pedagogical — in which learning paradigms can be located and compared. Third, we reinterpret the history of programming education as a movement from a predominantly symbolic environment toward a hybrid ecology in which symbolic execution, distributional generation, and situated reference coexist. This hybridity qualifies but does not cancel the indexical turn at the level of interaction and educational practice. Fourth, we articulate three transformations: from “object to think with” to “agent to think with”; from rule-first instruction to abductively initiated inquiry coordinating abduction, induction, and deduction; and from bricolage to artistic navigation. Navigational thinking has logical priority while the relevant problem space remains unstabilized, although it interleaves recursively with computational thinking as inquiry develops. The paper is conceptual rather than empirical: it advances a theoretical account intended to orient subsequent developmental research and pedagogical realization.
Ilya Levin
3 weeks 2 days ago
The urgency for sustainable consumption underscores the need to disseminate sustainability knowledge and promote environmentally responsible habits across societies. Due to its broad reach, social media has the potential to promote sustainable consumption by influencing habits and increasing environmental awareness. In this study, we investigate the role of social media through the design and strategic use of an Instagram account, WeSustainQtr, which aimed to disseminate sustainability knowledge and encourage sustainable consumption practices in Qatar and similar settings, through culturally relevant and solution-oriented content. Emphasizing a multi-theoretical framework through systematic and empirical qualitative research approach, we triangulated between secondary data from a scoping review and primary data of 420 initial survey and 415 post-course interaction survey responses. Baseline sustainability awareness was assessed through an initial survey, while the subsequent survey explored participants’ perceptions of sustainability and self-reported behaviors following their engagement with the account contents. Because responses could not be matched at the individual level, observed differences are interpreted as group-level patterns rather than evidence of individual behavioral change. Participants reported adopting a range of sustainable practices— from reducing single-use plastics and household energy consumption, to recycling — into the tapestry of their daily lives, with some reported influencing sustainability practices within their social circles. Findings suggest that social media can be used as a potential platform for sustainability education and user engagement by supporting knowledge sharing and collective participation. Reported engagement suggest that tailored implementations that introduced options for private participation and recognized culturally valuable themes may support followers’ engagement while also aligning with local norms towards privacy in the culturally sensitive context of Qatar and similar contexts. The findings also provide practical insights for integrating similar approaches into higher education curricula through structured learning activities that can encourage meaningful student engagement and behavioral change.
Maryam Al-Hail
3 weeks 2 days ago
The advent of generative artificial intelligence (AI) is transforming academic production and confronting the social sciences with the challenge of preserving critical thinking in the face of mechanically generated content. This research examines the perceptions and practices of social science lecturers and students at Ecuadorian universities regarding the use of AI, together with their motivations, constraints and the institutional policies currently in force. A mixed-methods design was adopted, combining qualitative and quantitative approaches. The instruments comprised semi-structured interviews with lecturers and students at national and Ibero-American universities; a non-probabilistic survey of 200 students; and a content analysis of seven sets of regulations from universities in Ecuador, all aligned with the framework of competencies identified by UNESCO-IESALC. The results show a moderate-to-high level of AI use (mean = 3.68), explained by perceived usefulness and extrinsic motivation rather than ease of use, and a gap between the demand for regulation (4.00) and perceived regulatory clarity (3.21); only one of the seven universities has a policy formally designated as such, and ethics predominates over professional and disciplinary development. The study concludes that the governance of AI in Ecuadorian higher education requires an evaluative redesign and discipline-differentiated competencies.
Abel Suing
3 weeks 2 days ago
BackgroundEffective communication is fundamental to high-quality healthcare. Communication skills training is a core component of medical curricula, yet opportunities to practice with standardized patients (SPs) are limited, costly, and can elicit anxiety in novice learners. Generative AI (GenAI) virtual patients may offer a scalable and accessible supplement. This quasi-experimental study examined the effect of practicing with a GenAI-patient on medical students’ self-reported nervousness, preparedness, self-efficacy, and self-perceived competence during their first SP encounter.Materials and methodsFirst-year medical students (n = 383) at the University of Amsterdam were allocated on the basis of the class schedule to a control group receiving standard e-learning preparation (n = 182) or an intervention group offered voluntary additional access to a GenAI-patient built on a Retrieval Augmented Generation (RAG) framework using GPT-4o (n = 201). Students completed a 31-item questionnaire after their first SP class. Primary outcomes were nervousness, preparedness, self-efficacy, and self-perceived competence. Group differences were assessed using Mann–Whitney U tests with Benjamini-Hochberg false discovery rate (FDR) correction. The primary analysis was by intention-to-treat, comparing all students allocated to the intervention with all control students; the comparison restricted to actual users of the tool was secondary.ResultsOf 180 students that completed the questionnaire, 84 were controls and 96 students were allocated to the intervention, of whom 49 had used the GenAI-patient. No statistically significant differences were found between groups on the outcomes after FDR correction. In the intention-to-treat analysis; all composite effect sizes were near zero (Cliff's delta −0.05 to 0.04, all 95% confidence intervals including zero), as were those in the per-protocol comparison (−0.12 to 0.07). Among tool users, experience was generally positive: 63%–71% gave positive ratings. The majority considered the GenAI-patient a valuable contribution to the standard class preparation (71%).ConclusionsVoluntary, single-session text-based practice with a GenAI-patient did not significantly reduce nervousness or improve self-assessed communication competence before a first SP encounter. Positive user experience suggests potential for the format, provided future iterations incorporate mandatory minimum engagement, multiple patient cases, automated feedback, and higher-fidelity multimodal interaction.
Remco Jongkind
3 weeks 5 days ago
IntroductionLarge language models (LLMs) are increasingly used to support evaluative tasks, raising questions about whether their judgements remain independent when users provide prior evaluative information. This study examined the context sensitivity of GPT-5.5 Instant-generated academic peer-review evaluations under positive and negative user-provided evaluative contexts.MethodSixteen academics evaluated two published manuscripts, classified for the study as relatively higher and lower quality, across four structured conditions. Each participant completed all conditions in separate ChatGPT sessions, yielding 64 peer-review outputs. Descriptive summaries of overall scores and publishability judgements complemented qualitative content analysis.ResultsThree themes emerged: alignment of evaluative emphasis with user-provided evaluative context, differences in the relative prominence of evaluation criteria, and explicit alignment with the user-provided evaluation. Mean ChatGPT scores were 8.74 and 8.84 under positive evaluative context, compared with 3.37 and 3.64 under negative evaluative context for the lower- and higher-quality manuscripts, respectively.DiscussionBecause all evaluative inputs included both a verbal evaluation and a numerical rating, the findings reflect sensitivity to the broader evaluative context and do not distinguish the contribution of verbal framing from numerical anchoring. The results highlight a potential limitation in treating ChatGPT-generated evaluations under comparable interaction conditions as independent second opinions and underscore the importance of interaction design in human–AI evaluation workflows.
Gülşah Subaşı