An Exploration of Mental Models of AI in Human–AI Co-Creativity: A Framework and Insights

This study aims to obtain a profile of the level of understanding of concepts and mental models of students in grades 10, 11, and 12 regarding the graphical concept of straight motion kinematics in high school. This research is descriptive research to provide an overview or description of a situation objectively. The research subjects were 25 students for each grade 10, 11, and 12 with a total of 75 with 23 male and 52 female taken by purposive sampling. The research data collection instrument used a test of understanding the concept of straight motion kinematics graphs in the form of open descriptions. The data analysis was carried out in two stages, the first stage was the analysis of the level of understanding of the concept and the second stage was the analysis of students’ mental models. Analysis of concept understanding data obtained by students’ mental models at the scientific level for students in total got a proportion of 20%. In the mental model, students at the Synthetic level obtained a proportion of 16.7%. Meanwhile, the mental level model dominates by reaching a proportion of 63.3%. From these different proportions of mental models, it can be used as a consideration for educators in choosing learning strategies in the classroom.

In recent years, generative AI has attracted significant public attention, and its use has been rapidly expanding across a wide range of domains. From creative tasks such as text summarization, idea generation, and source code generation, to the streamlining of medical support tasks like diagnostic report generation and summarization, AI is now deeply involved in many areas. Today’s breadth of AI applications is clearly distinct from what was seen before generative AI gained widespread recognition. Representative generative AI services include DALL·E 3 (OpenAI, California, USA) and Stable Diffusion (Stability AI, London, England, UK) for image generation, ChatGPT (OpenAI, California, USA), and Gemini (Google, California, USA) for text generation. The rise of generative AI has been influenced by advances in deep learning models and the scaling up of data, models, and computational resources based on the Scaling Laws. Moreover, the emergence of foundation models, which are trained on large-scale datasets and possess general-purpose knowledge applicable to various downstream tasks, is creating a new paradigm in AI development. These shifts brought about by generative AI and foundation models also profoundly impact medical image processing, fundamentally changing the framework for AI development in healthcare. This paper provides an overview of diffusion models used in image generation AI and large language models (LLMs) used in text generation AI, and introduces their applications in medical support. This paper also discusses foundation models, which are gaining attention alongside generative AI, including their construction methods and applications in the medical field. Finally, the paper explores how to develop foundation models and high-performance AI for medical support by fully utilizing national data and computational resources.

Mental models have an important role in the learning process because learning in general can be seen as mental modeling. The purpose of this study is to identify students' mental models and identify the relationship between mental models and prediction ability on the topic of convection heat transfer. The sample of this study was 15 on 12th-grade high school students, 8 male students and 7 female students. The students came from three different schools, namely from Tasikmalaya district, Ciamis district, and Banjar City. Sampling was done by purposive sampling with the characteristics of students who have studied heat and students who have high cognitive process abilities in their respective schools. Data collection was done by semi-structured interviews with the type of questions in the form of content and prediction. Data analysis is done by constant comparative method. The results of this study show that there are no students who have a scientific mental model. Five types of mental models were found, including unclear model, convection is a continuation of conduction, convection that does not change density, convection for evaporation, and model 3. In addition, the relationship between prediction and mental model was classified as complex. This is due to students who predict without using their mental models. Knowing the diverse mental models of students, educators become more knowledgeable about the level of representation of each student. So that educators can prepare appropriate learning strategies in order to construct students' mental models.

This study aims to explore the mental models of students about suspending objects in liquid fluid. The study used a descriptive qualitative method and implemented crosssectional approach. It involved 57 students from grade 5 of elementary school to fourthyear prospective physics teachers. The data collection used a test consisting of twenty-six essay and four multiple-choice items, which covered several contexts and factors. The data were analyzed by adapted phenomenographic procedures and integrated with some stages of thematic analysis. The types of mental models that were successfully explored include the density-, mass-, weight-, volume-, and gravity-based model, leaked boat model, air as a floater model, etc. The predominant students’ mental model was the initial one, followed by the synthetic and scientific level, respectively, and the adopted mental models tended to form a hierarchy based on the grade of students. The results showed that the suspending models tended to be adopted and was influenced by the mental model in the floating and sinking contexts. This result confirms the findings of previous studies, which stated that mental models depend on the context of the phenomenon being presented. The existence of variations in the students' mental models for reasoning about density and depiction of suspending objects revealed gaps in the consolidation of their mental models. In learning activities, it is not enough to teach the concepts of floating and sinking. Hence, an adequate portion for the suspending concept must be provided by depicting various object positions and emphasizing a more conceptual of density-based model..

The integration of Explainable AI (XAI) in mental health monitoring is poised to revolutionize the way clinicians diagnose, treat, and manage mental health conditions. While AI models have demonstrated significant potential in enhancing diagnostic accuracy and providing personalized care, the lack of transparency in decision-making processes remains a critical barrier to their adoption in clinical practice. This chapter explores the development and application of XAI models in mental health, emphasizing the need for transparency and interpretability to foster trust among clinicians and patients. By leveraging explainability techniques, XAI models not only provide insights into AI-driven decisions but also empower mental health professionals with actionable, comprehensible explanations that align with clinical judgment. The chapter further discusses the ethical, regulatory, and practical considerations for deploying XAI systems in mental health care, with a focus on ensuring fairness, privacy, and accountability. Key challenges, such as model complexity, data heterogeneity, and the need for continuous evaluation, are also examined. The role of AI transparency in strengthening clinician-patient relationships and enhancing decision-making is highlighted, positioning XAI as a crucial tool for improving the quality and accessibility of mental health care.

ObjectiveArtificial intelligence (AI) has impacted healthcare at urban and academic medical centers in the US. There are concerns, however, that the promise of AI may not be realized in rural communities. This scoping review aims to determine the extent of AI research in the rural US.Materials and MethodsWe conducted a scoping review following the PRISMA guidelines. We included peer-reviewed, original research studies indexed in PubMed, Embase, and WebOfScience after January 1, 2010 and through April 29, 2025. Studies were required to discuss the development, implementation, or evaluation of AI tools in rural US healthcare, including frameworks that help facilitate AI development (eg, data warehouses).ResultsOur search strategy found 26 studies meeting inclusion criteria after full text screening with 14 papers discussing predictive AI models and 12 papers discussing data or research infrastructure. AI models most often targeted resource allocation and distribution. Few studies explored model deployment and impact. Half noted the lack of data and analytic resources as a limitation. None of the studies discussed examples of generative AI being trained, evaluated, or deployed in a rural setting.DiscussionPractical limitations may be influencing and limiting the types of AI models evaluated in the rural US. Validation of tools in the rural US was underwhelming.ConclusionWith few studies moving beyond AI model design and development stages, there are clear gaps in our understanding of how to reliably validate, deploy, and sustain AI models in rural settings to advance health in all communities.

This paper investigates the problem of discrepancies between the capabilities declared by AI model developers and the actual quality of the corresponding tasks performed by these models. AI developers, especially those working on large language models (such as Gemini), identify several opportunities for creating educational content: generating content for different purposes, adapting to different audiences, personalizing learning, rapid content creation, and continuous improvement. However, a significant problem arises from the lack of publicly available data on the error rate of AI in these tasks. The issue is particularly critical in the context of creating educational content in higher mathematics. Errors in such materials can have critical consequences for students in higher education who use widely available AI models. This paper analyzes the possibilities of using Gemini AI to develop educational materials for higher mathematics courses. Examples of practical application of AI included its use to generate a solution to a given problem, generate task variants based on a template, and create multiple-choice test tasks. Most of the educational materials developed with the help of AI were related to testing students’ knowledge on the topic of the Theory of Functions of a Complex Variable. Real dialogues with AI are shown, which concern the correction of errors made by the model. The paper notes that all the comments made during the study were acknowledged and addressed by the AI model. Thus, at the current moment in time, the effective use of general-purpose AI models for generating educational content in higher mathematics can be carried out exclusively by users with a high level of mathematical training who can critically evaluate the results of the AI model. It is also shown that the lack of the ability to use formulas in Word Equation format when creating tasks for AI and generating its reports significantly reduces the effectiveness of AI in the professional activities of higher mathematics teachers.

Background.Artificial intelligence (AI) has impacted healthcare at urban and academic medical centers globally. The current focus on AI deployments in urban areas and the history of US urban-rural digital divides raises concerns that the promise of AI may not be realized in rural communities. This may exacerbate well-documented health disparities. Without the benefits of AI-driven improvements in patient outcomes and increased efficiency, rural healthcare facilities may fall farther behind their urban counterparts and rural hospital closure rates may continue to rise.Methods.We conducted a scoping review following the PRISMA guidelines. We included peer-reviewed, original research studies indexed in PubMed, Embase, and WebOfScience after January 1, 2010 and through April 29, 2025. Studies were required to discuss the development, implementation, or evaluation of AI tools in rural US healthcare, including frameworks that help facilitate AI development (e.g., data warehouses).Findings.Our search strategy found 26 studies meeting inclusion criteria after full text screening with 14 papers discussing predictive AI models and 12 papers discussing data or research infrastructure. AI models most commonly targeted resource allocation and distribution. Few studies explored model deployment and impact. Half noted the lack of data and analytic resources as a limitation to both development and validation. None of the studies discussed examples of generative AI being trained, evaluated, or deployed in a rural setting.Interpretation.Practical limitations may be influencing and limiting the types of AI models evaluated in the rural US. We noted validation of tools in the rural US was underwhelming, and ultimately, neglected. With few studies moving beyond AI model design and development stages, there is a clear gap in our understanding of how to reliably validate, deploy, and sustain AI models in rural settings to advance health in all communities.Funding.National Library of Medicine

This paper looks at Federated Learning (FL) as a privacy-preserving method of deployment of AI within the domain of mental health. Although FL in healthcare, in general, is currently being viewed with some interest, extending FL to mental health-driven AI models is still a vast research lacuna. Mental health data is sensitive in nature and necessitates robust protection of privacy, which FL seemingly provides as a cost-efficient alternative to conventional centralized AI models. Our paper bridges this gap because we compare FL-based methods against centralized models along the dimensions of privacy, efficiency, and feasibility. In addition to privacy benefit assessment, we carry out a comparative analysis to identify the advantages and disadvantages of FL in mental health AI. We further suggest a more advanced FL framework that integrates block chain-based security protocols with adaptive aggregation techniques with the objectives of bridging weaknesses and enhancing robustness. The figures, tables, and experimental comparisons presented are empirical proof of our assertions, demonstrating improvements in privacy protection and model performance. We appreciate constructive criticism of our work and invite proposals for further improvement, further experiments, or other approaches. Through our contribution to the development of privacy-preserving AI in mental health treatment, we envision the ethical, safe, and effective implementation of AI-based mental health treatments. Our results hope to close the current research gap and act as a stepping stone for future research in this field.

This study aims to analyze the mental model of students in learning the periodic system of elements using 3D representations. This research was conducted in the odd semester of the 2022/2023 academic year. The research subjects were 36 students of class XB SMAN 54 Jakarta. This study used qualitative research methods. Data collection techniques through mental model tests, interviews, reflective journals, and classroom observations. The students' mental models were analyzed in three stages, namely engage, explore, and explain. The students' mental models are categorized into three, namely scientific, synthetic, and initial mental models which are viewed from the three levels of macroscopic, submicroscopic, and symbolic representations. The results showed that class XB had three categories of mental models, namely scientific mental models with an overall average of 70.86%, synthetic mental models with an overall average of 17.54%, and initial mental models with an overall average of 11.60 %. Mental models are formed from the thoughts and experiences of students who previously had, teacher explanations, 3D representation media, and handbooks or other sources.

As more and more forms of AI become prevalent, it becomes increasingly important to understand how people develop mental models of these systems. In this work we study people's mental models of AI in a cooperative word guessing game. We run think-aloud studies in which people play the game with an AI agent; through thematic analysis we identify features of the mental models developed by participants. In a large-scale study we have participants play the game with the AI agent online and use a post-game survey to probe their mental model. We find that those who win more often have better estimates of the AI agent's abilities. We present three components for modeling AI systems, propose that understanding the underlying technology is insufficient for developing appropriate conceptual models (analysis of behavior is also necessary), and suggest future work for studying the revision of mental models over time.

Social intelligence (SI) is of great importance in the success of the counseling and psychotherapy, whether for the psychologist or for the artificial intelligence systems that help the psychologist, as it is the ability to understand the feelings, emotions, and needs of people during the counseling process. Therefore, this study aims to identify the Social Intelligence (SI) of artificial intelligence represented by its large linguistic models, "ChatGPT; Google Bard; and Bing" compared to psychologists. A stratified random manner sample of 180 students of counseling psychology from the bachelor's and doctoral stages at King Khalid University was selected, while the large linguistic models included ChatGPT-4, Google Bard, and Bing. They (the psychologists and the AI models) responded to the social intelligence scale. There were significant differences in SI between psychologists and AI's ChatGPT-4 and Bing. ChatGPT-4 exceeded 100% of all the psychologists, and Bing outperformed 50% of PhD holders and 90% of bachelor's holders. The differences in SI between Google Bard and bachelor students were not significant, whereas the differences with PhDs were significant; Where 90% of PhD holders excel on Google Bird. We explored the possibility of using human measures on AI entities, especially language models, and the results indicate that the development of AI in understanding emotions and social behavior related to social intelligence is very rapid. AI will help the psychotherapist a great deal in new ways. The psychotherapist needs to be aware of possible areas of further development of AI given their benefits in counseling and psychotherapy. Studies using humanistic and non-humanistic criteria with large linguistic models are needed.

Abstract. A research on the application of blended learning in the material solubility and solubility product. This study aims to determine (1) the formation of students' mental models after using blended learning, (2) the difference between the results of student learning through the implementation of blended learning and conventional learning material is applied to the solubility and solubility product of high school students of class XI Science PGRI 4 Banjarmasin, (3) students' response to the application of blended learning and conventional learning material that is applied to the solubility and solubility product of high school students of class XI Science PGRI 4 Banjarmasin. The research method used was a quasi experiment. Research sample are high school students of class XI Science PGRI 4 Banjarmasin which consists of class XI IPA 1 as the control class (n = 27) and class 2 as class XI Science experiments (n = 25). Data collection techniques and the value of using student response test and questionnaire techniques. Statistical analysis using ANOVA test path. The results showed that (1) the formation of students' mental models after using blended learning on material solubility and solubility product for the better, (2) learning outcomes between students who learn the material solubility and solubility product using high Lebing blended learning compared with students who learn the material solubility and solubility product using conventional learning, (3) blended learning received a positive response from the students of class XI Science 2 SMA PGRI 4 Banjarmasin on material solubility and solubility product. Keywords: blended learning, mental model, learning outcomes, solubility and solubility product.

An increasing amount of citizen science projects involve citizens on levels of participation that go beyond data collection and entail the co-creation of research questions and methods, as well as the project pathway. The success of such projects depends on the establishment of shared knowledge, a task that can be especially challenging in citizen science that focuses on complex societal issues and the so-called wicked problems. We suggest that this challenge can be addressed through a deeper engagement with research on mental models -- cognitive representations of external reality that largely define human thinking, decision-making and behaviour. Moreover, particular emphasis should be put on the effective co-creation of shared mental models, whereby design thinking could provide valuable methodologies and tools. The objective of the workshop “Mental Models in Citizen Science” was to dive into mental model theory and design thinking toolbox, and to explore their potential for citizen science. This paper provides an overview of the workshop activities and insights, and proposes a research agenda shaped around mental models and their role in citizen science.

Programming is important for development of skills and thinking for the learners. but the problem of programming is abstract content. Learners can’t imagine a result from their program during the programming. so, it makes programming is boring and difficult for learning. Mental model is understanding of learner to create mental representation in media and symbol. Development of student’s mental model can help students to construct their knowledge in programming. Moreover, motivation is the one of important factor to success in learning. The purpose of this research was study leaners’ mental model and motivation. The participants target 10 students at rural school. Experimental research was employed in this study. The instruments used in the experiment were Constructivist online learning environment. Data collection used the mental model interview recording form and motivation survey form. The results found that learner’s mental model consists with 2 characteristics as follow: (1) Represent story or event by explaining in model (2) Change rule and procedure to solve problem. The learners’ motivation was very motivated (mean = 3.93, S.D. = 0.44). It was comprised of 2 components as follow: 1) Internal motivation was (mean = 4.1, S.D. = 0.35) 2) External motivation was very motivate (mean = 3.75, S.D. = 0.53). In conclusion, the Constructivism Online Learning Environment can improve programming and self-learning performance in rural school.