Recent work has explored the interests that draw learners to Machine Learning (ML), aiming to support their success and broaden participation in the field. However, whether strategies used in textbooks align with these interests is unexplored. We perform a thematic analysis of the introductions from ten openly available ML textbooks to identify their motivational strategies and compare them with student interests documented in prior research. We find that textbooks frequently motivate learners in their introductions by setting learning goals, previewing core ML topics to be covered, showcasing applications and current successes, and, less often, by using learner-centered strategies such as reassurance or curiosity prompts. We group these motivations into three overarching themes: theoretical, practical, and learner-centered. These motivations largely align with student interests, particularly in theory and applications, even in textbooks published before the recent surge of ML and Artificial Intelligence. These findings reveal how textbooks frame ML’s value and offer evidence-based guidance for developing future materials that better engage and support diverse learners.

As machine learning (ML) becomes integral in more disciplines, introductory courses in the field are attracting increasingly diverse audiences. Design of these introductory ML courses needs to be theoretically sound, but also intuitive, engaging, and accessible to a range of students. Effective teaching of ML must go beyond teaching the theoretical or practical mechanics of algorithms. In this paper, we synthesize effective teaching strategies from 6 experienced ML instructors across 5 institutions to help students define appropriate ML problems, build intuition, develop reasoning skills, and apply models responsibly. We organize these strategies into eight thematic areas: preparing students for success, motivating learners through real-world relevance, integrating ethics and societal impact, avoiding common methodological pitfalls in model evaluation, guiding students on design decisions, adapting effective classroom practices, assessing student learning, and preparing for the future. Each section offers practical examples of classroom-tested activities (or references to existing resources), and in many cases, reflections on our experiences with the strategies. Our aim is for this paper to be a starting point for instructors aiming to improve learning in introductory ML courses. We hope this is a resource-rich guide for teaching ML to diverse learners, grounded in both pedagogy and practice.

Previous work on prerequisites in computing found that GPAs correlated more strongly with course grades than direct measures of prerequisite knowledge. This suggests that course grades encompass more than subject knowledge. This work was conducted in an advanced data structures (ADS) course at a single institution. To investigate the generalizability of these findings, we replicate the study in an upper-year machine learning course. We find that the pattern holds. These results underscore the need to distinguish course performance from demonstrated skill.

This paper describes the instructor and student experience of a ''just-in-time'', blended approach to prerequisite review, implemented in a machine learning course but applicable elsewhere. Although pre-requisites are commonly used to structure university curricula, both the literature and our experience show that students sometimes forget prerequisite knowledge when it is needed in subsequent courses. This challenge is especially pronounced in courses where diverse prerequisite concepts are applied throughout the semester, with different concepts required for different units. Our approach consisted of short prerequisite review quizzes due before each lecture, where the quiz questions assessed mastery of key prerequisite concepts needed for that lecture. Moreover, each quiz was accompanied by a brief instructional video that provided a targeted review of the content. We evaluated this approach across two course implementations, based on perspectives from 2 instructors and 353 students. Both instructors and students felt positively: a reduction in prerequisite related questions during lectures was observed, and students reported that the approach helped bridge gaps in preparedness, improved self-efficacy, and was efficient. More interestingly, the responses showed key tradeoffs regarding the timing, modality, and level of support in a prerequisite review intervention. While we believe this approach to be applicable for other courses with diverse requirements, our results lead us to believe that there is no one-size-fits-all for prerequisite review, and that it is highly context-dependent.

The Model AI Assignments session seeks to gather and disseminate the best assignment designs of the Artificial Intelligence (AI) Education community. Recognizing that assignments form the core of student learning experience, we here present abstracts of thirteen AI assignments from the 2025 session that are easily adoptable, playfully engaging, and flexible for a variety of instructor needs. Assignment specifications and supporting resources may be found at http://modelai.gettysburg.edu

Artificial Intelligence (AI) has impacted the world tremendously in the last decade, causing an increased demand for accessible AI education globally. Students benefit from studying AI earlier in the curriculum; however, AI courses can require a range of prerequisites, which can be structured differently in various educational contexts. In this paper, we study the curriculum structure of AI, Machine Learning (ML), and Data Science (DS) courses in Canadian Universities and compare it with that of US Research-1 institutions. There are many similarities between AI, ML, and DS courses in Canada and the US. For example, DS courses tend to be more accessible earlier in the CS curriculum compared to AI and ML. However, there are key differences between the two countries, with Canadian AI, ML, and DS courses generally being a part of a longer prerequisites chain, and Canadian CS departments offering fewer DS courses. Still, both Canadian and US institutions find innovative ways to introduce AI earlier in the curriculum, including via interdisciplinary courses and specialized courses with few prerequisites. This study corroborates earlier work in recognizing diversity in curricular frameworks in North America and recommends curricular revisions and early academic advising to ensure access to AI courses.

Feedback is a cornerstone of effective learning, offering students insight into their progress, while also providing instructors with information to refine their teaching. This paper presents a practical tool leveraging large language models (LLMs) to cluster and summarize teaching assistant (TA) feedback. Designed for educators, the tool streamlines the identification of common student issues, provides course-level insights, and generates actionable summaries intended to be modified by educators and shared with students. To validate the tool, we conducted an instructor survey assessing its perceived usefulness and accuracy, compared TA and LLM-generated feedback on a shared assignment, and presented a case study where the tool informed improvements to the assignment handout. Our findings suggest that this practitioner-focused tool can enhance feedback workflows, promote consistency in instruction, and support scalable improvements in teaching and learning.

Machine learning (ML) has become increasingly important for students, yet university-level ML courses are often perceived as challenging and time-intensive. This study explores the perceived challenges and motivations of students in a university ML course to inform curricular and teaching strategies. Through 5 surveys conducted in two instances of a 12-week introductory ML course, we examined students' engagement with both theoretical and practical aspects of ML. Results indicate that while students initially express strong interest in applying ML concepts, their reported interests can shift toward theoretical foundations. Challenges in both theory and practice are reported, including difficulties in mathematical notation and vectorization of gradient components, as well as model implementation. Students also discuss the time commitment required in a course with both theoretical and practical content. We recommend aligning course content with student motivations, providing targeted support for mathematical notation and vectorization, and balancing theoretical depth with practical application.

This Research Full paper focuses on the challenges in artificial intelligence, machine learning, and data science education—referred to as “artificial intelligence” courses here-after-often characterized by extensive prerequisites that limit student access. We analyze the course structures and prerequisites of these courses in computing departments at 50 Research-1 institutions in the United States, recognized for their “Very High Research Activity.” Our methodology involves analyzing course syllabi to examine the structure and prerequisites of these courses, using open coding to develop a unified codebook to identify prerequisites and determine the earliest exposure levels for students. A clustering analysis was also conducted to identify common and differing curriculum approaches among institutions. Results show that data science courses require less initial exposure, while artificial intelligence and machine learning courses require more prerequisites. Standard requirements for artificial intelligence courses include basic data structure (Computer Science 2) and algorithms, with machine learning courses requiring more mathematics preparation. Moreover, public institutions offer advanced courses with more prerequisites compared to private institutions. Overall, this study recognizes considerable diversity in curricular frameworks across Research-1 institutions and encourages institutions to revise curricula to broaden access to artificial intelligence education and increase participation in research.

Faculty and industry practitioners recognize written communication to be important in computer science, but it can be challenging to convince students of the same. As student perceptions are molded early in a program of study, we focus on early-year CS students to understand their perceptions towards the importance of writing in CS, with the goal of framing discipline-specific writing pedagogy. We qualitatively analyze responses from first and second-year CS students in a survey about the role of writing in their field. The responses reveal that a majority view writing as an indispensable skill. Specifically, students recognize it as a fundamental skill, applicable across diverse contexts, and uniquely relevant in CS compared to other fields. We identified 4 perceptions that they hold which are helpful to their development as writers: that writing is a useful fundamental skill, which is useful for achieving various goals in a variety of contexts, and that writing in CS is different than in other fields. However, 20% of responses include reasons why writing is not important in CS, and we identify 4 perceptions harmful to students' development as writers: that writing skills can be avoided, are defined narrowly, do not need to be developed beyond a baseline, and come at the cost of computing skills. We believe that there is an opportunity to align discipline-specific writing instruction with these useful and harmful perceptions.

The panel comprises a diverse set of Computing educators working on AI in education. The panelists will address four areas of AI in Computing education: 1) AI for introductory CS classrooms, 2) Investigating opportunities presented by LLMs, 3) LLM-based tool development, and 4) Ethics and inclusion in AI curriculum. The panel will share experiences and discuss opportunities and challenges in AI education with the community.

Recent work has confirmed that computing students experience the Imposter Phenomenon (IP) at higher rates than reported in other disciplines. However, no work has examined what aspects of the university computing experience might lead to a higher rate of IP experiences. We aim to illustrate the IP experiences students have, identify common sources of these experiences, and document the effects of these experiences and how students respond to them. We asked undergraduate students to share recent experiences that illustrate their experiences with the IP. We conducted an inductive thematic analysis on these open-ended responses, resulting in a set of inter-connected themes. A significant fraction of students related stories about making comparisons with peers or observing peer behaviour that made them question their abilities. Students also spoke about holding unrealistic expectations learned from their peers or imposed by the environment. These experiences may be particularly acute for minority-affiliated students who may come to feel they do not belong. Ultimately, these IP experiences can lead to a loss of motivation or a cycle of failure that leads students to leave computing. The central role social comparisons play in IP experiences suggests that it is particularly important to foster communities where opportunities for comparison are reduced and where realistic expectations are explicitly set.

Computing science education has experienced low attendance and historic declines in registration from different minority groups. The past decade of enrollment surge in computer science undergraduate programs has increased the number of women and minorities in the field, but the improvements are inconsistent and less than expected. An increase in the use of computing science and in the demand of technology workforce is expected in the upcoming years. Thus, computing science is set to shape the future of technology for a diverse set of technology users. Therefore, it is important to analyze how undergraduate program admission procedures are affecting Equity, Diversity, and Inclusion of historically marginalized groups in computing science.

Large language models (LLMs) can be used to answer student questions on course discussion boards, but there is a risk of LLMs answering questions they are unable to address. We propose and evaluate an LLM-based system that classifies student questions into one of four types: conceptual, homework, logistics, and not answerable. We then prompt an LLM using a type-specific prompt. Using GPT-3, we achieve 81% classification accuracy across the four categories. Furthermore, we achieve 93% accuracy on classifying not answerable questions. This indicates that our system effectively ignores questions that it cannot address.

This panel will highlight the experiences of Black, Asian, and Latinx women and non-binary instructors as junior faculty. Despite efforts to broaden participation, CS departments are often quite homogeneous, forcing faculty of marginalized identities to attenuate certain aspects of, or entire identities, to better "fit in" with their colleagues. We will discuss the impacts on faculty and engage with the audience to identify avenues of individual (i.e., colleagues) and systemic (i.e., department culture and policies) allyship.

Writing skills are often considered unimportant by computer science students and were under-emphasized in our curriculum. We describe our experience embedding CS-specific writing instruction at scale in most of our large, core, first- and second-year Computer Science courses, each with 300-800+ students. Our approach is to collaborate with a writing specialist and a community of course instructors, centralize the management of writing teaching assistants, and introduce a variety of relevant genres and contexts to help students develop and apply writing skills. We outline the institutional support and organization crucial to a project of this scale. In addition, we report on a survey collecting student perception of the writing instruction/assessment. We reflect on quantitative and qualitative evidence of success, as well as the challenges that we faced. We believe that many of these challenges will be common across institutions, particularly those with large courses.

Motivation Bots can alleviate the workload of instructors supporting students in large course Q&A platforms, but it's not clear whether students will be receptive to the use of automated assistants in this setting. Objectives We aim to observe student reactions when they encounter bot-generated follow-ups to Q&A board posts. We investigate the effect of revealing that a bot, rather than a human, is suggesting that the current post is a duplicate. Methods Our bot revealed or hid its bot identity when suggesting duplicate posts, with the condition selected randomly. We observed students' reactions in both conditions. A post-course survey was distributed to collect students' demographic data, previous experiences with bots, and attitudes toward our bot. Results We observed a slight increase in students' response rate when the bot hid its identity. We compared the positive response rate in both conditions and did not find evidence suggesting that students had less trust in bot-generated answers. From the survey, we only saw minimal direct evidence that students might mistrust the bot: 7 of 59 students reported worries about receiving an inaccurate bot-generated answer. Other students were concerned that they would not receive attention from an instructor. Discussion We did not find evidence that revealing the bot's identity has a negative impact on student reactions. However, future bot design should consider the emotional impact of deploying a bot as there may be negative emotional effects to receiving a bot-generated response.

Motivation Despite the widespread belief that computing practitioners frequently experience the Imposter Phenomenon (IP), little formal work has measured the prevalence of IP in the computing community despite its negative effect on achievement.

We explore RNN and CodeBERT deep learning models that highlight errors in student submissions to Python coding problems. We find that a standard automatic metric like AUC does not correspond well to human evaluation, and that the scale of the benefits of transfer learning and pre-training are only seen when using human evaluation.

This study analyzes common issues in the writing of our upper-year, undergraduate computer science students in timed (e.g. tests) and untimed (e.g. longer assignment reports) scenarios. Our goal is to identify writing issues that should be addressed earlier in the CS curriculum. In collaboration with a writing specialist, we develop and fine-tune a rubric with Grammar, Conciseness, Clarity, Organization, Structure, and Formality as the main categories.

The Model AI Assignments session seeks to gather and disseminate the best assignment designs of the Artificial Intelligence (AI) Education community. Recognizing that assignments form the core of student learning experience, we here present abstracts of six AI assignments from the 2021 session that are easily adoptable, playfully engaging, and flexible for a variety of instructor needs. Assignment specifications and supporting resources may be found at http://modelai.gettysburg.edu.

The Model AI Assignments session seeks to gather and disseminate the best assignment designs of the Artificial Intelligence (AI) Education community. Recognizing that assignments form the core of student learning experience, we here present abstracts of nine AI assignments from the 2020 session that are easily adoptable, playfully engaging, and flexible for a variety of instructor needs. Assignment specifications and supporting resources may be found at http://modelai.gettysburg.edu.

We present a machine learning model to obtain vector representations of student code submissions for a CS1 programming problem. These vectorembeddings can be used to compare code, cluster code submissions, and identify errors. We hope to use these embeddings to identify conceptual misunderstandings in student code.

The Model AI Assignments session seeks to gather and disseminate the best assignment designs of the Artificial Intelligence (AI) Education community. Recognizing that assignments form the core of student learning experience, we here present abstracts of ten AI assignments from the 2019 session that are easily adoptable, playfully engaging, and flexible for a variety of instructor needs. Assignment specifications and supporting resources may be found at http: //modelai.gettysburg.edu.

The world is covered with millions of buildings, and precisely knowing each instance's position and extents is vital to a multitude of applications. Recently, automated building footprint segmentation models have shown superior detection accuracy thanks to the usage of Convolutional Neural Networks (CNN). However, even the latest evolutions struggle to precisely delineating borders, which often leads to geometric distortions and inadvertent fusion of adjacent building instances. We propose to overcome this issue by exploiting the distinct geometric properties of buildings. To this end, we present Deep Structured Active Contours (DSAC), a novel framework that integrates priors and constraints into the segmentation process, such as continuous boundaries, smooth edges, and sharp corners. To do so, DSAC employs Active Contour Models (ACM), a family of constraint- and prior-based polygonal models. We learn ACM parameterizations per instance using a CNN, and show how to incorporate all components in a structured output model, making DSAC trainable end-to-end. We evaluate DSAC on three challenging building instance segmentation datasets, where it compares favorably against state-of-the-art. Code will be made available on https://github.com/dmarcosg/DSAC.

In this paper, we revisit the recurrent back-propagation (RBP) algorithm, discuss the conditions under which it applies as well as how to satisfy them in deep neural networks. We show that RBP can be unstable and propose two variants based on conjugate gradient on the normal equations (CG-RBP) and Neumann series (Neumann-RBP). We further investigate the relationship between Neumann-RBP and back propagation through time (BPTT) and its truncated version (TBPTT). Our Neumann-RBP has the same time complexity as TBPTT but only requires constant memory, whereas TBPTT’s memory cost scales linearly with the number of truncation steps. We examine all RBP variants along with BPTT and TBPTT in three different application domains: associative memory with continuous Hopfield networks, document classification in citation networks using graph neural networks and hyperparameter optimization for fully connected networks. All experiments demonstrate that RBPs, especially the Neumann-RBP variant, are efficient and effective for optimizing convergent recurrent neural networks.

Synthesizing programs using example input/outputs is a classic problem in artificial intelligence. We present a method for solving Programming By Example (PBE) problems by using a neural model to guide the search of a constraint logic programming system called miniKanren. Crucially, the neural model uses miniKanren's internal representation as input; miniKanren represents a PBE problem as recursive constraints imposed by the provided examples. We explore Recurrent Neural Network and Graph Neural Network models. We contribute a modified miniKanren, drivable by an external agent, available at https://github.com/xuexue/neuralkanren. We show that our neural-guided approach using constraints can synthesize programs faster in many cases, and importantly, can generalize to larger problems.