Multimodal large language models (MLLMs) are Generative AI models that take different modalities such as text, audio, and video as input and generate appropriate multimodal output.Since such models will be integrated into future educational tools, a humancentered design approach that takes students' perspectives into account is essential while designing such applications.This paper describes two co-design workshops which were conducted with 79 student groups to examine how they design and prototype future educational tools integrated with MLLMs.Through various activities in the workshops, students discussed relevant educational problems, created journey maps, storyboards and low fidelity prototypes for their applications, and evaluated their applications based on relevant design principles.We found that students' applications used MLLMs for important learning environment design features such as multimodal content creation, personalization, and feedback.Based on these findings, we discuss future research directions for the design of multimodality in generative AI educational applications.

The effectiveness of traces as a tool for understanding programs is well understood, however, research shows that [2, 3] students remain reluctant to use tracing, preferring to quickly write programs and use debuggers in an unsystematic, trial and error fashion.We report here some experiments that evaluate the effectiveness of a different approach called trace-first [4], where we ask students to create traces by analyzing problems, before they write code.The analysis derives intended behavior and then convert it to code.To investigate the impact of using traces as both an analysis and design tool, we conducted a pre/post study (n=29) with students solving programming problems.In the pre-test, students approached problems without explicit guidance on tracing.We then taught them to use traces to plan intended behavior, and then develop programs.We find that more students get programs correct, and surveys show evidence that students find traces a helpful tool for problem analysis and program design. CCS Concepts• Social and professional topics

Self-regulation refers to the ability to plan, monitor, control and reflect on one's problem-solving process. Prior research has shown that self-regulated learning (SRL) strategies help improve novice performance in solving programming problems. However, with the advent of LLM tools like ChatGPT, novices can generate fairly accurate code by just providing the problem prompt, and hence may forego applying essential self-regulation strategies such as planning and reflection to solve the problem.

In the age of rapid Artificial Intelligence (AI) advancement, universities worldwide have responded by offering specialized AI and Machine Learning (ML) courses to meet industry demands. However, amidst this surge in AI education, AI deployment's ethical and societal implications often need more attention. Unlike traditional algorithmic programming, AI involves intricate decision-making processes that are challenging to predict or explain, demanding a comprehensive understanding of its ethical dimensions. Various global initiatives have highlighted the ethical considerations surrounding AI, resulting in an increased emphasis on integrating ethics education into AI curricula. Despite this, there remains a gap in addressing these vital aspects across the broader AI education landscape, with ethics often relegated to the periphery of computer science courses. In this context, this paper explores the imperative for ethics courses in undergraduate AI education in India. We examine educators ' awareness and perceptions regarding ethics education in AI curricula through interactions with faculty members participating in professional development programs and workshops focused on AI pedagogy. Our analysis reveals that there is a pressing need to extend discussions on ethics beyond mere privacy concerns and traditional performance metrics, integrating real-life scenarios into the curriculum. This paper serves as a preliminary step towards a need for a general framework in structuring ethics education in AI in India, aiming to initiate a more comprehensive and standardized approach to AI ethics education, fostering responsible AI development in the future.

Recent advances in artificial intelligence have led to the development of large language models (LLMs), which are able to generate text, images, and source code based on prompts provided by humans. In this paper, we explore the capabilities of an LLM - OpenAI's GPT-3 model to provide feedback for student written code. Specifically, we examine the feasibility of GPT-3 to check, critique and suggest changes to code written by learners in an online programming exam of an undergraduate Python programming course.

In this paper, we investigate the relationship between student experiences and transactional distance in a 12 week online undergraduate Python programming course. Transactional distance is defined as the psychological and communication space between students and instructors due to geographical separation. Although several studies have examined learning from a transactional distance perspective, there has been a lack of research which describe computing education courses from this perspective.

In the past, unraveling learner interaction data in TELE was a challenge. However, the advent of LA has helped in uncovering latent information in log data to scaffold learning. This paper focuses on learner interaction in VeriSIM, a TELE, to teach software design diagrams. The learners’ performance in the system is used to categorize them into three groups, namely, "full scorers", "partial scorers", and "give uppers". Our analysis found that the full scorers spend a significantly higher duration per action than the give-uppers in an introductory challenge presented in the learning environment. Further analysis unravels the strategies used by consistent and inconsistent learners, and it was observed that the learner interaction strategies evolve with increasing difficulty levels as they navigate through the challenges.

Disciplinary practices are processes and skills applied for sensemaking, reasoning and problem solving. Studies have shown that experts are able to spontaneously apply these skills but novices have difficulty due to various reasons. The aim of this paper is to showcase learning environments that foster disciplinary practices in undergraduate computing students in three contexts - data structures, software design and computer networks. Findings from our studies show that students are able to apply these disciplinary practices to solve ill-structured problems. These findings provide motivation for exploring disciplinary practices in other CS courses as well.

An important skill graduating computing students require is to evaluate a given software design and ensure that it satisfies the intended requirements. Prior work has shown that while working with software designs, experts think deeply about the design and simulate scenarios where the design does not satisfy the requirements. In this paper, we examine how students evaluate a given set of software design diagrams (UML class and sequence diagrams) against the given requirements.

Computer Science undergraduates are expected to design software solutions and also verify that the design satisfies the intended requirements. The Ph.D. work discussed in this paper aims at designing and evaluating a learning environment to train computer science undergraduates to effectively verify properties of a software system design. Literature on expertise in software design has shown that experts create rich mental models of the software design on which they perform mental simulations. I propose a model-based learning strategy in order to foster the cognitive processes of mental modeling and mental simulation. I hypothesize that by triggering the cognitive processes of mental modeling and mental simulation, students will be able to perform design verification better. By the end of my doctoral research I expect the following contributions: 1) Understanding how the cognitive processes of mental modeling and mental simulation aid in software design verification. 2) Application of a model-based learning strategy and creation of a learning environment in order to foster these cognitive processes.

Novice researchers have difficulties in operationalization (breaking down of abstract concepts to measurables) and generalization (generalize the findings to make claims). We have designed a system called OPeD (Operationalizing using ProblEm Decomposition), for teaching learning of operationalization. OPeD trains novice researchers in operationalization through problem decomposition. OPeD is based on pedagogical theories of guided inquiry, problem visualization and adaptation. In order to visualize the operationalization process, OPeD guides the learner to create a decomposition tree and construct meaningful hypotheses based on the tree. The gradual and iterative construction of this tree can help learners develop their operationalization skills. In this paper we present the design of OPeD and provide an exemplar of a learner path in the context of educational research.

Computer Networks, an undergraduate computer science course, is taught through a variety of new technological tools but more attention on assessment strategies using those technological tools is desirable. Research studies have shown that assessment methods also influence students' learning. Current assessment techniques, even those which include technology, focus more on summative assessment rather than formative. Formative assessment is necessary for teaching and assessment of complex skills like troubleshooting or network design. Therefore, there exists a need for assessments to capture students' problem solving process and chosen approaches during the assessment. By tracking a student's behaviors in an immersive assessment environment, the path that the student takes towards the solution can be studied. Virtual Worlds are naturally immersive environments and can be effectively used to record interactions that students made in the worlds. We have developed NetWorld in Open Wonderland, an open source software for creating 3D virtual worlds. NetWorld is a 3D virtual world built for detailed assessment of computer network concepts for computer science undergraduate students. This detailed assessment happens at 3 levels -- conceptual understanding, diagnostic abilities and the ability to design a network. This paper describes the design principles and implementation of NetWorld and a plan to evaluate its usability.

In this paper we describe the architecture of virtual clot (vCLOT), a virtual world system designed to teach procedural knowledge of blood clotting process, in a game based learning environment. vCLOT utilizes virtual world to give learners an immersive learning experience while actively participating in tasks that require them to apply the procedural knowledge they have learned. Design of vCLOT combines the immersivity of virtual worlds with the power of knowledge structures. Immersivity provides learners with the opportunity to make decisions at every level of the game. This transfers control of interaction to the learner, enabling the learner to be actively engaged in knowledge construction process. Knowledge structures are a neat way to represent domain and learner data. The user interface of vCLOT is designed and implemented with Open Wonderland which is an open-source 3D toolkit. The learning goal which is to learn about blood clotting process, is aligned with the game goal, which is application of blood clot process steps to heal an injury. The game goal is presented as a quest in which the learner interacts with concepts by either dragging them or synthesizing them from other concepts. Learners complete the quest on successful formation of blood clot which in turn implies that they have learnt the blood clot process. We plan to do a usability study to improve the system before starting actual intervention.