Human Teacher Research Articles

Mixing Corrupted Preferences for Robust and Feedback-Efficient Preference-Based Reinforcement Learning

Preference-based reinforcement learning (RL) trains agents using non-expert feedback without the need for detailed reward design. Human teacher guides the agent by comparing two behavior trajectories and labeling the preference. Although recent studies improved feedback efficiency through methods like unsupervised exploration and self- or semi-supervised learning, they often assume flawless human annotation. In practice, human teachers might make mistakes or have conflicting opinions on trajectory preferences, posing a challenge in capturing user intent. To address this, we introduce mixing corrupted preferences (MCP) for robust and feedback-efficient preference-based RL. Inspired by the robustness of mixup against corrupted labels, MCP offers three key advantages. Firstly, by component-wise mixing of two labeled preferences, MCP mitigates the impact of corrupted feedback, enhancing robustness. Secondly, MCP improves feedback efficiency by generating unlimited new data even with limited labeled feedback. Lastly, MCP helps regulate overconfidence in preference predictor, moderating excessive reward divergence between two trajectories. We evaluate our method on three locomotion and six robotic manipulation tasks in B-Pref benchmark, in contexts with both perfectly rational and imperfect teachers (including actual human teachers). Our results show that MCP significantly outperforms PEBBLE, requiring fewer feedback instances and a shorter training period, highlighting its superior feedback efficiency. Our code is available at https://github.com/JongKook-Heo/MCP.

Ask Me

In this poster, we share ongoing research into the use of realistic AI-generated avatars for education in a large Australian business school. These highly realistic avatars have been used as educational presenters to provoke critical discussions on ethical AI by immersing students in text-to-video technology (Vallis et al., 2024). From this initial study, an emergent typology for avatars in educational media has been developed, to help educators design for teaching and learning based on learning purposes and modes of delivery Britton & Vallis, 2023; Vallis & Britton, 2024). Such practical, theoretical guidance in designing, developing, and deploying AI in education is much needed (Dawson et al., 2023). Drawing upon learnings to date, we further explore synthetic media designs for effective and ethical use in education. The poster presents three use cases of synthetic media with QR codes to demonstrate their application and to facilitate discussion at the conference. An additional fourth QR code will be added to gather participant feedback on which use case they found most compelling and educational, as well as their reasons for their preferences. The first use case explores short video presentations, where synthetic media offers students a choice of accents, voices, and faces to present educational content (Dao et al., 2021). This approach aims to free teacher time for more learner guidance and communication, potentially enhancing learning outcomes. Video-based learning can enhance student engagement and understanding by providing flexible, interactive, and visually stimulating experiences (Sabli? et al., 2021). The second use case is ‘Ask me’, a trial of a realistic avatar that answers frequently asked questions. Multimodal AI, like GPT-4o, with integrated voice, video, and audio capabilities, can simulate real-world scenarios for educational scenarios with engaging and realistic experiences (Mollick & Mollick, 2023). Future use cases may include specific learning activities such as role-play and rehearsal. Multimodal feedback on student progress is presented in the third use case. A lifelike digital representation of a teacher has been designed to closely mimic a real human teacher’s appearance and expressions. Simple video feedback, generated based on student interactions with university systems and combined with a teacher’s text, is sent via email to enhance engagement. This trial is designed to augment feedback, rather than replace interaction between students and teachers (Bozkurt & Sharma, 2023). By presenting these use cases, we aim to provide practical insights and stimulate discussions on the ethical and effective use of synthetic media in education. Our preliminary findings indicate benefits to using AI-generated avatars, if the ethical implications and the quality of human-AI interactions are carefully considered in the learning design (Vallis et al., 2024).

AI-Teacher Teaching Task Spectrum in Action

The AI-Teacher Teaching Tasks Spectrum (AITTTS) was conceived as a way to understand the relationship between human teachers and the ever-evolving AI technologies in education. This study demonstrates how the AITTTS framework can be operationalised into a tangible intervention, showcasing the design models and practical applications of the AITTTS in real-world educational settings. By categorising teaching tasks into a spectrum, the AITTTS delineates the roles that AI and human teachers can play, providing a structured and nuanced understanding of their collaboration. As a result of the practical application of the AITTTS, a design model was birthed in this study. It highlights various aspects of holistic student outcomes such as positive electronic nonverbal communication (eNVC) cues, adaptive learning paths, and interactive learning responses as elements by which AI should be designed. By providing a structured approach for educators to incorporate AI tools and interventions in their learning environments, this research lays the groundwork for further exploration of the synergistic relationship between AI and human teachers in modern education. This framework can serve as a guide for educators to develop and implement AI-enhanced teaching strategies, fostering a more dynamic and responsive educational landscape.

KNOWLEDGE, ATTITUDES, AND PRACTICES REGARDING CHATBOTS AMONG HEALTHCARE PROFESSIONALS IN A TERTIARY CARE HOSPITAL

Objective: (1) Primary: Determine healthcare professionals’ knowledge, attitudes, and practices (KAP) related to AI Chatbots. (2) Secondary: Assess healthcare professionals’ perspectives on using Chatbots as teaching tools and implementing them in the Competency-Based Medical Education curriculum. Methods: An online questionnaire was distributed to 132 health professionals, including faculty and CRMIs, through Google Forms. Data on artificial intelligence (AI)-related KAP and barriers were collected following IHEC approval. The KAP scores, along with the relationship between categorical variables – such as population type and the use of AI Chatbots – were analyzed using Statistical Package for the Social Sciences software. Results: The study revealed that participants had a moderate level of knowledge about AI Chatbots. Attitudes were mixed, with some skepticism about AI replacing human teachers but also recognition of its benefits. Most participants infrequently used AI Chatbots in their daily activities. Barriers to usage included lack of knowledge, limited access, time constraints, and curriculum gaps. Conclusion: This study underscored the need to enhance medical education with AI topics and address existing barriers. It is crucial to better prepare health professionals for AI integration to leverage AI’s potential for improving patient care and training.

I Care That You Don’t Share: Confidentiality in Student-Robot Interactions

Enabled by technological advances, robot teachers have entered educational service frontlines. Scholars and policymakers suggest that during Human-Robot Interaction (HRI), human teachers should remain “in-the-loop” (i.e., oversee interactions between students and robots). Drawing on impression management theory, we challenge this belief to argue that robot teacher confidentiality (i.e., robot teachers not sharing student interactions with the human teacher) lets students make more use of the technology. To examine this effect and provide deeper insights into multiple mechanisms and boundary conditions, we conduct six field, laboratory and online experiments that use virtual and physical robot teachers (Total N = 2,012). We first show that students indeed make more use of a confidential (vs. nonconfidential) robot teacher (both physical and virtual). In a qualitative study (Study 2), we use structural topic modeling to inductively identify relevant mediators and moderators. Studies 3 through 5 provide support for these, showing two key mediators (i.e., social judgment concern and interaction anxiety) and two moderators (i.e., student prevention focus and teacher benevolence) for the effect of robot teacher confidentiality. Collectively, the present research introduces the concept of service robot confidentiality, illustrating why and how not sharing HRI with a third actor critically impacts educational service encounters.

Approaches to Digital Humanities Pedagogy: a systematic literature review within educational practice

Abstract Over the years, Digital Humanities Pedagogy has swiftly evolved, with a growing emphasis on the intersection between Computing and Humanities. However, in the associated literature, a lack of clarity has been found regarding the adoption of diverse approaches to DH pedagogy in different educational settings. This study tackles this research gap by analyzing the current state of knowledge in the field through a Systematic Literature Review. The aim of this review is to investigate the topic and provide an overview of the approaches currently being employed within educational practice, while also identifying areas that require additional investigation. The time frame of the reviewed studies spans from 2012 to 2022. After distinguishing between studies that deal with DH as a subject of study and those that deal with DH as a field of application, the analysis will focus on the latter category. Specifically, the review will examine the level of education and courses that employ DH educational practice, the types of technology used and their intended purpose, the literacies involved, and the aspects of digital resources and practices that are critically studied. By shedding light on the diverse practices employed in different educational settings, the findings reveal a notable gap in primary and secondary education and a lack of critical analysis toward digital resources and practices, suggesting important new directions for future research.

Student Perceptions of ChatGPT in Mandatory EFL Writing Classes at a Korean University : A Case Study

This study explored to what extent students at a Korean university perceive ChatGPT as useful in mandatory university EFL writing classes. In addition, preferences regarding methods of instruction and the outlook for future classes of a similar nature in the era of ChatGPT were also examined. 239 students enrolled at a university located on the outskirts of Seoul responded to a survey questionnaire consisting of a combination of multiple choice, 10-point Likert scale, and open-ended questions, which was used as the primary data collection method. Data analysis was conducted using general qualitative analysis for open ended questions and descriptive statistics for multiple choice and 10-point Likert scale questions. In addition, if consent was given, survey participants were contacted by e-mail to discuss responses in further detail. Despite concerns relating to ownership of learning and overreliance, the findings indicate generally positive perceptions of ChatGPT in terms of assisting with grammar, vocabulary, brainstorming, and proofreading. Conventional approaches (e.g., human teacher, in person lessons, and inter student / teacher communication) are still highly desired although the increasing influence of ChatGPT cannot be ignored. There is a belief that mandatory EFL writing classes currently play a vital role at Korean universities, and there is a strong preference for them to continue. Administrational and educational implications are also discussed.

From recorded to AI‐generated instructional videos: A comparison of learning performance and experience

AbstractGenerative AI (GAI) and AI‐generated content (AIGC) have been increasingly involved in our work and daily life, providing a new learning experience for students. This study examines whether AI‐generated instructional videos (AIIV) can facilitate learning as effectively as traditional recorded videos (RV). We propose an instructional video generation pipeline that includes customized GPT (Generative Pre‐trained Transformer), text‐to‐speech and lip synthesis techniques to generate videos from slides and a clip or a photo of a human instructor. Seventy‐six students were randomly assigned to learn English words using either AIIV or RV, with performance assessed by retention, transfer and subjective measures from cognitive, emotional, motivational and social perspectives. The findings indicate that the AIIV group performed as well as the RV group in facilitating learning, with AIIV showing higher retention but no significant differences in transfer. RV was found to offer a stronger sense of social presence. Although other subjective measures were similar between the two groups, AIIV was perceived as slightly less favourable. However, the AIIV was still found to be moderately to highly attractive, addressing concerns related to the uncanny valley effect. This research demonstrates that AIGC can be an effective tool for learning, offering valuable implications for the use of GAI in educational settings. Practitioner notesWhat is already known about this topic Instructional videos, especially those featuring a teacher's presence, have been widely used in second language learning to facilitate learning. Producing instructional videos is costly and burdensome. Generative AI has great potential for generating educational content. What this paper adds An AI‐generated instructional video (including generated lecture text, voice and appearance) demonstrated greater improvement in students' retention performance in English word learning than a traditional recorded video. Students perceived no significant differences between the AI‐generated instructional video and recorded video in satisfaction, motivation, trust, cognitive load, emotions and parasocial interaction dimensions, although the AI‐generated instructional video group reported slightly lower values. Despite AI‐generated instructional video eliciting a significantly lower value of social presence than recorded video, it led to a reduction in cognitive load and better performance. Implications for practice and/or policy We recommend using the AI‐generated instructional video in both physical and online classes for its positive effects on both learning achievement and learning experience. The findings indicate the equivalence principle in AI‐generated content, highlighting that the appearance, voice and lecture text generated by current AI technology have reached a certain level of quality.

Unmasking teachers’ proficiency in harnessing Artificial Intelligence (AI) for transformative education

This paper examines the awareness of artificial intelligence (AI) among school teachers in Nepal. We interviewed twenty teachers from ten urban schools in Nepal over four weeks using a semi-structured interview tool. The analysis of interview data reveals the teachers’ mixed perception of AI, including their understanding of innovative technologies in education, the potential challenges for new-generation teachers, and the opportunities for pedagogical transformations. The findings suggest that although the teachers have limited awareness of how AI can be used in teaching and learning, they possess some level of understanding that has opened up new opportunities for exploring innovative ways of teaching and learning. Notably, participants discussed tools, such as Education Management Information System (EMIS) for record keeping and automated grading and AI-driven chatbots for student tutoring. However, teachers also expressed concerns about the possibility of AI replacing human teachers. These insights provide valuable implications for policymakers and educators regarding the integration of AI into classroom instructions to promote innovative and equitable learning opportunities.

Enhancing instructional design learning: a comparative study of scaffolding by a 5E instructional model-informed artificial intelligence chatbot and a human teacher

ABSTRACT To address the limitations of general-purpose artificial intelligence (AI) tools, we developed a task-oriented AI chatbot based on the 5E (i.e. “engage”, “explore”, “explain”, “elaborate” and “evaluate”) model to scaffold students’ instructional design process. We examined the impact of integrating the 5E instructional model-informed AI chatbot on students’ learning performance and perceptions. The results indicated that the AI chatbot, when combined with human teacher scaffolding, significantly improved the students’ instructional design performance relative to receiving human teacher scaffolding only. The chatbot provided valuable suggestions on instructional design frameworks, class activities and teaching topics during the “explore” phase. In the “evaluate” phase, the chatbot offered immediate feedback on the students’ design plans and proposed alternative instructional frameworks regarding areas for improvement. However, the students expressed concerns about the chatbot’s evaluation quality, noting that it needed to be better aligned with the course assessment rubric. We recommend using AI chatbots for instructional design conceptualisation, although we emphasise the critical role of human teachers in evaluating final design work and providing timely support.

Human or Humanoid Animated Pedagogical Avatars in Video Lectures: The Impact of the Knowledge Type on Learning Outcomes

This study examines the impacts of educational videos with human or humanoid animated pedagogical avatars (APAs) on cognitive, affective, and social learning outcomes in declarative and procedural knowledge contexts. A 2 × 2 factorial design was used, considering instructor type (human vs. humanoid APA) and knowledge type (declarative vs. procedural). The study involved 139 university students with no prior knowledge of the video content and was conducted in a human-computer interaction lab using an eye-tracking device. There were no significant differences in learning achievement, perceived lecture engagement, emotion, satisfaction, and social presence between the human and humanoid APA. However, learners showed greater visual interest in the human instructor, which influenced their interest in the learning content. Viewers of declarative knowledge videos showed more visual interest and achieved higher learning scores, while procedural videos increased lecture engagement, emotion, satisfaction, and social presence. The study suggests that humanoid APAs can effectively replace human instructors in educational videos and emphasizes the importance of considering knowledge types in video design to impact learning outcomes.

AI-CiciBot as Conversational Partners in EFL Education, focusing on Intelligent Technology Adoption (ITA) to Mollify Speaking Anxiety

This study examines the role of AI-CiciBot as a conversational partner in mitigating speaking anxiety among Indonesian EFL (English as a Foreign Language) students. The aim is to explore the integration of intelligent technology adoption (ITA) into classrooms to enhance language learning and reduce anxiety during speaking activities. The primary gap addressed in this research is the lack of technology-based solutions for oral language practice in Indonesia, where traditional curricula emphasize reading and writing over speaking. This study aims to fill the gap by introducing AI-CiciBot, a novel AI-powered conversational tool that provides a supportive, low-pressure environment for students to practice speaking English. The study uses a mixed-methods approach, combining quantitative data from pre-test and post-test scores and qualitative insights from student interviews. The research tools include a questionnaire and an AI-CiciBot platform for speaking practice. IBM SPSS Statistics analyzed the data, focusing on students' speaking performance and anxiety levels before and after using AI-CiciBot. Findings in this study show a significant improvement in students' speaking skills and a reduction in anxiety. The mean score for pre-tests was 49.17, which increased to 90.47 in post-tests, indicating AI-CiciBot's effectiveness. Interviews revealed that students initially hesitant to engage in speaking activities due to fear of judgment became more confident and willing to participate after practicing with the AI bot. This study contributes to the field of EFL education by demonstrating the potential of AI-driven tools like AI-CiciBot to address speaking anxiety and improve oral proficiency. It also highlights the importance of ITA in encouraging students to embrace AI-based learning solutions. To maximize learning outcomes, future research should investigate the long-term impact of AI-CiciBot on students' speaking proficiency and its integration with human instructors.

ChatGPT in Teaching Linear Algebra: Strides Forward, Steps to Go

Abstract As soon as a new technology emerges, the education community explores its affordances and the possibilities to apply it in education. In this article, we analyze sessions with ChatGPT around topics in basic linear algebra. We reflect on the affordances and changes between two versions of ChatGPT since its worldwide publication in our area of interest, namely, linear algebra. In particular, the question of whether this software can be a teaching assistant or even somehow replace the human teacher is addressed. As of the time this article is written, the answer is generally negative. For the small part where the answer can be positive, some reflections about an original instrumental genesis are given.

A survey of communicating robot learning during human-robot interaction

For robots to seamlessly interact with humans, we first need to make sure that humans and robots understand one another. Diverse algorithms have been developed to enable robots to learn from humans (i.e., transferring information from humans to robots). In parallel, visual, haptic, and auditory communication interfaces have been designed to convey the robot’s internal state to the human (i.e., transferring information from robots to humans). Prior research often separates these two directions of information transfer, and focuses primarily on either learning algorithms or communication interfaces. By contrast, in this survey we take an interdisciplinary approach to identify common themes and emerging trends that close the loop between learning and communication. Specifically, we survey state-of-the-art methods and outcomes for communicating a robot’s learning back to the human teacher during human-robot interaction. This discussion connects human-in-the-loop learning methods and explainable robot learning with multimodal feedback systems and measures of human-robot interaction. We find that—when learning and communication are developed together—the resulting closed-loop system can lead to improved human teaching, increased human trust, and human-robot co-adaptation. The paper includes a perspective on several of the interdisciplinary research themes and open questions that could advance how future robots communicate their learning to everyday operators. Finally, we implement a selection of the reviewed methods in a case study where participants kinesthetically teach a robot arm. This case study documents and tests an integrated approach for learning in ways that can be communicated, conveying this learning across multimodal interfaces, and measuring the resulting changes in human and robot behavior.

Exploring feasibility of vision-based automated evaluation during laboratory courses in manufacturing

Manual assessment of laboratory exercises and assignments by human instructors is subjective and time-consuming, introducing errors, inconsistencies, and biases. A preliminary study developing an automated evaluation system utilizing computer vision has been proposed in this article to address these challenges. It utilizes computer vision to assess the accuracy and quality of components machined by students enrolled during the manufacturing laboratory course. The system includes image acquisition hardware, algorithms for objective decisions, and an interface for evaluating student performance. The system is implemented during one of the laboratory classes, and a comparative assessment is carried out with manual evaluations followed by student feedback. It has been shown that the system can address subjectivity concerns, reducing the workload on course instructors and teaching assistants. The research also broadens the utility of computer vision in manufacturing education and creates interest among enrolled students to appreciate the role of newer technologies in a core discipline.

AI Tutors vs. Human Instructors: Perceptions of Higher Education Students in Hungary and Spain

Integrating AI-powered tutoring systems in higher education represents a significant advancement in educational technology, offering personalized and adaptive learning experiences. This study investigates the perceptions and expectations of higher education students in Hungary and Spain regarding AI tutors. Despite extensive research on the technological efficacy of AI systems, there is limited understanding of student attitudes in these specific cultural contexts. This research aims to fill this gap by exploring student expectations, satisfaction levels, and perceived benefits of AI tutors compared to human instructors. To achieve this, a validated questionnaire was administered to 184 higher education students from Hungary and Spain, capturing data on various dimensions of their expectations. The study's findings indicate that students appreciate the adaptability and continuous guidance provided by AI tutors, with Hungarian students showing higher overall expectations compared to their Spanish counterparts. These insights suggest that AI tutoring systems can enhance the learning experience by addressing individual student needs more effectively. The implications of this study are significant for higher education institutions seeking to integrate AI technologies.

ChatGPT 사용 경험에 대한 대학생 인식: 영어 글쓰기 과제 수행 중심으로

This study explores the experiences and perceptions of Korean college students using ChatGPT as a tool for completing assignments in a general English course focused on academic writing skills. A total of 93 students participated in the study by completing a questionnaire that addressed the purpose and satisfaction of using ChatGPT, its advantages and disadvantages, and its overall effectiveness. Descriptive statistics and thematic analysis were employed to analyze the data. A Word Cloud was used as an analytical tool to visualize the keywords from the two open-ended questions regarding the advantages and disadvantages of ChatGPT. The results indicated that Korean college students generally had a positive perception of ChatGPT. In terms of the purpose of using ChatGPT, it was found that ChatGPT was most frequently used for academic purposes, such as completing assignments and understanding course content. In the context of studying English, students primarily used ChatGPT to revise grammar and sentence structure. For English writing, ChatGPT was most often utilized for proofreading and editing. The keywords associated with the advantages of ChatGPT were “easily,” “information,” and “desired,” while those related to its disadvantages were “errors” and “mistakes.” Interestingly, students reported relying more on ChatGPT’s feedback than on that of their peers or dictionaries, yet they still valued human instructors’ feedback more highly than ChatGPT's. These findings suggest that ChatGPT can be a valuable tool for English writing, given the students' positive perceptions and satisfaction. However, the discussion of the complementary roles of artificial intelligence and human instructors remains important for educators.

Exploring The Impact of Educational Leadership on Secondary School Teachers' Professional Growth: A Comprehensive Survey

The study titled "Educational Leadership: A Survey on Secondary School Teachers" was conducted in the Sadar block of Balasore, Odisha, with a focus on 121 teachers from 19 secondary schools. A sample of 93 teachers, selected using the Raosoft sample size calculator at a 95% confidence level through stratified random sampling, participated. The research aimed to explore leadership qualities among secondary school teachers, specifically comparing the perceptions of science and humanities teachers. Using percentage analysis, the study analyzed quantitative and qualitative data collected via questionnaires and standardized tools. The findings revealed that teachers generally have a positive perception of leadership traits such as interpersonal relations, intellectual operations, and emotional stability. However, many indicated a need for improvement in ethical and moral strength. While communication adequacy and operational citizenship were rated positively, a minority of teachers gave lower ratings in these areas. The study concluded with recommendations for schools to offer professional development programs, mentorship initiatives, and collaborative efforts to enhance leadership skills. Access to workshops, relevant literature, and conferences was suggested as essential for fostering continuous growth in educational leadership among teachers. Key words: Educational Leadership, Secondary School Teachers, Teacher Perceptions, Leadership Qualities

Comparing human-made and AI-generated teaching videos: An experimental study on learning effects

In the age of generative AI, can teaching videos be efficiently and effectively generated by large language models? In this study, the authors used generative AI tools to develop four short teaching videos for a management course and then compared them with human-generated videos on the same subjects in an online experiment. In an across-subject experimental design, 447 participants completed two treatment conditions presenting different mixes of AI-generated and human-made videos. The participants were asked to rate their learning experiences after each video and had their learning outcomes tested in a multiple-choice exam at the end of the session (N = 1788 video treatments). The findings show that human-generated videos provided a statistically significant but small advantage to participants in terms of learning experience, indicating that the participants still prefer to be taught by human teachers. However, a comparison of exam results between the experimental groups implies that the participants eventually acquired knowledge about the content to a similar degree. Given these findings and the ease with which AI-generated teaching videos can be created, this study concludes that AI-generated teaching videos will likely proliferate.

Examining Writing Feedback Dynamics from ChatGPT AI and Human Educators: a Comparative Study

As technology increasingly influences educational practices, it is crucial to comprehend the impact and effectiveness of AI in comparison to human-driven methodologies. In the field of language learning and writing instruction, one notable development is the emergence of AI-generated writing feedback tools. These tools, exemplified by ChatGPT in this study, promise efficiency, objectivity, and detailed insights for learners. This study aims to investigate the comparative effectiveness of an artificial intelligence (AI) automated writing feedback tool and human teacher feedback in enhancing the writing skills of English Language Learners. The research was conducted on fifteen third-year undergraduate students in the discipline of English Language Education, using a phenomenological design. The study collected qualitative data through assignments and feedback given to students at a higher education institution. The findings suggest that while human teachers were praised for their understanding, personalized guidance, and emotional intelligence in providing feedback, AI-generated feedback was considered more detailed and comprehensive. It highlights the importance of balancing efficiency with the human touch in the feedback process. This study makes a significant contribution to the ongoing discourse surrounding the integration of AI in education and its impact on the student-teacher dynamic.