PaperlayerView exact revision on arXiv

Source-linked AI summary

Embedded Human-Centered Data Science in a Graduate Programming Course: A Framework and Case Study

Victoria Chui, Kelly McConvey, Daniel Chui, Malayna Bernstein, Shion Guha

arXiv:2609.08982v1cs.HC

TL;DR

The paper addresses the need for further validation of human-centered computing education in information-science settings. It proposes HELIX, adapting computing-ethics pedagogy for data-science contexts; its pilot findings suggest improved attitudes toward ethical responsibility, while knowledge and decision-making changed minimally.

  • Problem

    Further validation is needed for embedding human-centered and ethical content in computing courses taught in information-science departments.

    Source: §1 · p. 2

  • Method

    The paper proposes HELIX by adapting existing computing-ethics pedagogy for data-science contexts.

    Source: §1 · p. 1

  • Results

    The pilot findings suggest improved attitudes toward ethical responsibility, while knowledge and decision-making skills showed minimal change.

    Source: §6 · p. 19

  • Takeaways & Limitations

    The authors present HELIX as a potential pedagogical instrument in interdisciplinary computing contexts.

    Source: §6 · p. 19

    The authors call for more rigorous evaluation of the framework.

    Source: §6 · p. 19

Abstract

from arXiv · show

As AI and data-driven systems pervade practice, there is an imperative for instructors to embed societal impact and ethics content into computing courses. In response, we present the Human-Centered Education for Learning in Information and eXplainable Computing (HELIX) framework for information science programs, organized around three iterative pillars - knowledge building, decision-making, and empowerment - with concrete actions for instructors and students. We applied the framework in a graduate, introductory programming course using readings, algorithmic design activities, and scenario-based reflections. We present a pilot implementation of this framework to examine changes in students' (n=22) knowledge acquisition, decision-making processes, and self-reflection regarding human-centered perspectives in data science. We release an anonymized materials kit (survey, assignments, analysis code) to support adoption. We discuss design tensions (workload, assessment, relevance to diverse information science learners) and provide guidelines for integrating human-centered content without overwhelming technical outcomes. Findings suggest that the HELIX Framework is feasible in information science contexts and future work should use comparative survey assessment to strengthen causal inferences.

1 Introduction

The introduction motivates human-centered teaching in information science (IS) data-science courses and proposes HELIX, a framework adapted to IS contexts. A preliminary course study reports survey changes, while the authors identify further evaluation and course-integration challenges.

1 Introduction

  • Its three learning pillars move from knowledge building to decision-making and empowerment, with actions for students and instructors.The framework also identifies collaborative goals and potential challenges.Source: §1 · pp. 1–3
  • HELIX adapts computing-ethics pedagogy for interdisciplinary IS and data-science education rather than replacing existing approaches.The framework addresses IS students' broad educational backgrounds and program learning outcomes.Source: §1 · pp. 1–2
  • The study examines pre- and post-course changes in human-centered data-science knowledge, attitudes, scenario decisions, and perceived IS program learning outcomes.The case study was conducted in a graduate introductory programming course.Source: §1 · pp. 2–3
  • p = 0.001 for overall survey performance, while attitudes rose from 20.8/25 to 22.7/25 (p = 0.002).The attitudes and perceptions portion comprised five Likert questions.Source: §1 · p. 3
  • 21 of 22 respondents reported somewhat or significant improvement in their understanding of human-centered data science.The study used pre- and post-course surveys in an introductory programming course.Source: §1 · pp. 2–3
  • The authors provide HELIX guidelines and a materials kit for researchers to examine, test, adapt, and refine.The kit includes a survey, synthetic dataset, quantitative analysis code, and assignment adaptations.Source: §1 · p. 3

2 Background

Prior work has embedded ethical and human-centered content in computing education, but most research concerns undergraduate computer-science courses. The background highlights instructional challenges and the need to understand how this content fits graduate IS courses with diverse learners.

2 Background

  • Embedding ethical content can add workload and labor for instructors and teaching assistants, while students may prioritize technical material when ethics is not graded.Researchers recommend integration that does not overwhelm existing course goals; guest lecturers with relevant expertise have been associated with student engagement.Source: §2 · p. 4
  • Embedded ethics may support recognition of workplace concerns without necessarily increasing students' sense of empowerment to resolve societal-impact conflicts.A cited survey found students with prior embedded-ethics coursework were more likely to recognize ethical concerns, but no more likely to feel empowered to resolve the conflict.Source: §2 · p. 4
  • Prior work connects human-centered computing with HCI and critical theory, including frameworks for anticipating and mitigating societal impacts of technical work.The background also notes growing calls to apply human-centered considerations to data-science work and education.Source: §2 · p. 4
  • Most research on embedded ethical content concerns computer-science undergraduate courses, leaving the distinct challenges and possibilities of graduate IS courses less understood.IS students may come from varied, including non-technical, backgrounds and have broad academic interests.Source: §2 · p. 5

3 Methods

This section develops HELIX as an iterative, adaptable framework for human-centered learning across information science courses and programs, mapping program learning outcomes to its learning actions. It also identifies implementation challenges and describes support for students carrying human-centered decisions into future courses and professional roles.

3.1 Framework Development

  • HELIX was developed through a literature search around knowledge building, decision-making, and empowerment to complement computing ethics curricula in interdisciplinary data science contexts.The framework adapts concepts from existing curricular frameworks and learning outcomes from multiple information science departments.Source: §3.1 · p. 5

3.1.1 Frameworks in Education.

  • The education-framework review shaped HELIX’s iterative pillars, moving from foundational knowledge to applied code and design contexts, critical evaluation, and learner empowerment.The cycle accommodates revisiting concepts and emphasizes responsible decision-making and student agency in sociotechnical systems.Source: §3.1.1 · p. 6

3.1.2 Program Learning Outcomes.

  • PLOs from over 50 institutions were inductively grouped into three themes: Understanding Information, Human-Centered Decisions, and Future Empowerment.The themes combine information practices and literacy; social impact, responsibility, leadership, and user-centered society; and professional conduct, intellectual growth, lifelong learning, and collaboration.Source: §3.1.2 · p. 7
  • The PLO themes were mapped to HELIX actions from Remember and Understand for knowledge building, through Apply and Analyze for decision-making, to Evaluate and Create for empowerment.The mapping links information practices to foundational learning, societal impacts and ethical principles to applied decisions, and critical assessment and ongoing growth to future education or workplace settings.Source: §3.1.2 · pp. 7–8
  • The framework’s empowerment component looks toward students’ future courses and professional roles, supporting continued human-centered, reflexive decisions.Source: §3.1.2 · p. 8

3.1.3 Challenges in Embedded Ethics.

  • Instructors may lack ethics expertise, prompting calls for ethicists to contribute focused expertise while balancing human-centered content with technical outcomes [18, 47, 50, 51, 56, 59].Source: §3.1.3 · p. 8
  • Embedding ethics in technical courses may require extra development and student-engagement time; material may exceed the computing task’s complexity, particularly in CS1, and feel less relevant to students who do not relate to its outcomes.Source: §3.1.3 · p. 8
  • Assessing embedded content may require revised evaluation tools and rubrics, and students and teaching assistants need guidance on how it will be graded.Source: §3.1.3 · p. 9
  • Adding embedded content on top of existing outcomes and lectures can overload students with material during the semester.Source: §3.1.3 · p. 9

3.1.4 Framework Structure.

  • HELIX structures human-centered learning around three iterative pillars and actions suited to students with diverse educational backgrounds and courses from introductory to advanced levels.Its circular structure supports reflection across both a course and a program.Source: §3.1.4 · pp. 9–10
  • The framework progresses from building knowledge to applying concepts in decision-making and then empowering students to evaluate and advocate for human-centered solutions in future work.Students may repeat knowledge-building and decision-making as needed before applying lessons in professional settings.Source: §3.1.4 · p. 9

4 Case Study

The case study embedded HELIX in a 12-week graduate programming course and used pre- and post-course surveys to examine students’ human-centered data science knowledge, attitudes, and learning outcomes. It describes the course and analysis methods, reports gains and changes in ethical decision-making alongside mixed progress across program learning outcomes, and releases a materials kit for future use.

4.1 Study Setting

  • Among 62 enrolled students, 23 (37%) consented; after excluding one who completed only the pre-course survey, 22 provided completed pre- and post-course responses, with voluntary participation potentially introducing self-selection bias.The introductory programming course had no specific prerequisites beyond a prior degree, and students came from varied educational backgrounds.Source: §4.1 · p. 10

4.2 Survey Development

  • The pre- and post-course survey assessed HCDS knowledge and attitudes through conceptual questions, ethical scenarios, and five-point Likert items.Scenario questions presented ethical dilemmas with response options reflecting different ethical frameworks, without designating a correct answer, to capture students’ reasoning and priorities.Source: §4.2 · pp. 10–11

4.3 Embedded Content

  • Human-centered readings and activities were integrated through most weeks, with assignments progressing from knowledge building to technical decisions and reflection.Activities prompted consideration of downstream social impacts and possible participatory or community-engaged design; students’ final project reflections were not collected for this study.Source: §4.3 · pp. 11–12

4.4 Data Analysis

  • Pre- and post-course survey scores were compared with Wilcoxon signed-rank tests, with knowledge and scenario responses scored from 1 to 5 for human-centeredness.The test choice reflected paired observations, a sample of 22, and non-normally distributed score differences; scores across knowledge, scenarios, and attitudes were also tested for overall change.Source: §4.4 · p. 12
  • Two co-authors briefly coded responses to count mentions of human-centered concepts, resolving disagreements through discussion; one blank submission left 21 complete responses.The concepts counted included ethics, bias, fairness, and transparency.Source: §4.4 · pp. 12–13

4.5 Case Study Results

  • All 22 students completed pre- and post-course surveys; prior experience ranged from data-focused or general ethics coursework to no ethics experience, and from data science coursework to no related experience.Ten students had taken a technology/data-focused ethics course, ten a general ethics course, and two had no ethics experience.Source: §4.5 · p. 13

4.5.1 Quantitative Analysis of Survey Questions.

  • Students’ average score across five Likert questions rose from 20.8 pre-course to 22.7 post-course (p = 0.002), and mean scores for each survey section and overall increased from pre- to post-course.Source: §4.5.1 · p. 13
  • Among 21 responses, 14 (66.7%) identified an increased need for conscious ethical decision-making, 12 (57.1%) emphasized societal impacts, and 9 (42.9%) prioritized bias and fairness in the workplace.Privacy concerns were noted most frequently (71.4% of respondents), followed by ethical concerns (52.4%).Source: §4.5.1 · pp. 13–14

4.5.2 Program Learning Outcomes.

  • Students reported increased learning in information practices, social leadership, and technology’s societal impacts, but anticipated greater gains in critical assessment, information theory, and intellectual growth.The authors note that not every program learning outcome directly relates to every information science course, and other courses may complement this learning.Source: §4.5.2 · p. 14

4.6 HELIX Materials Kit

  • The HELIX Materials Kit v1.0 includes the survey and scoring rubric, adapted Ethical Engine assignments, activity and quiz prompts, synthetic survey data, and an analysis notebook.The kit is intended for future researchers and educators.Source: §4.6 · pp. 14–15

5 Discussion

The discussion examines HELIX as a way to connect technical learning with human-centered understanding and practice in information science courses, noting its applicability in an introductory IS programming setting and distinguishing improved attitudes from limited short-term changes in measured knowledge and scenario-based decision making. It also addresses integration challenges and recommends flexible, curriculum-wide approaches tailored to students while recognizing workload, assessment, learner-relevance, and workplace barriers.

5.1 Implications of HELIX Framework Case Study and Potential Improvements

  • The study found statistically significant differences (p < 0.01) in overall performance on survey questions covering knowledge assessment, ethical scenarios, and attitudes and perceptions between the semester’s first and last weeks.The authors suggest that semester-long human-centered content contributed to survey performance and the human-centered decision making students were prompted to consider.Source: §5.1 · p. 15
  • Attitudes toward human-centered responsibility improved, but the preliminary case study found minimal change in measured knowledge and scenario-based decision making.The authors suggest short-term exposure may shift awareness without developing applied ethical reasoning skills.Source: §5.1 · p. 15
  • Varied datasets, real-world examples, and take-home activities can engage students with different academic interests in human-centered concepts and connect them to achievable technical goals.The authors recommend adapting knowledge-building content to students’ interests to support engagement early in the course.Source: §5.1 · p. 16
  • Hands-on ethical design activities raised awareness of ethical challenges and prompted students to take ownership of their skill development and decisions.The reported areas of awareness included ethics, privacy, bias, fairness, societal impact, transparency, accountability, and critical thinking.Source: §5.1 · p. 16
  • Reflection on algorithmic societal impacts may strengthen students’ critique, but learners with non-technical interests may struggle to connect these skills to future information-professional roles.The authors recommend broader activities addressing data use in non-technical roles and caution that one course may not equip students to contest workplace decisions; empowerment should include recognizing tensions and seeking resources.Source: §5.1 · p. 16

5.2 Implications of the HELIX Framework for Computing Instruction in IS

  • HELIX addresses the lack of coherent structures for human-centered learning in technical courses by organizing data science education around knowledge building, decision-making, and empowerment.It offers instructors guidance for activities and assessments and gives students opportunities to connect computing concepts with real-world consequences, awareness, and agency.Source: §5.2 · p. 17
  • HELIX reframes existing technical learning outcomes through a human-centered lens, connecting computing concepts to authentic contexts and broader professional competencies.In the examined programming course, activities complemented coding, problem-solving, and computational-thinking objectives by examining how technical decisions shape user experiences and societal outcomes.Source: §5.2 · p. 17
  • Integrating human-centered content can increase workload and compete for instructional time, while assessing attitudes, ethical reasoning, or empowerment and serving learners with diverse backgrounds and career goals pose challenges.The authors emphasize integrating these topics meaningfully rather than treating them as ancillary material.Source: §5.2 · pp. 17–18
  • Embedding HELIX across IS data-science degrees could pair technical skill development with human-centered understanding and future algorithmic practice, while requiring attention to implementation trade-offs.The framework may also build competencies beyond those expected in primarily technical courses.Source: §5.2 · p. 18

5.3 Guidelines for the Application of the HELIX Framework in Practice

  • Apply HELIX flexibly, letting students control movement among its iterative pillars and allowing more time for technical-background students new to human-centered topics.The guidelines draw on the case study and challenges in embedded ethics research.Source: §5.3 · p. 18
  • Plan human-centered concepts alongside future courses to build a cohesive workplace ethics toolbox and empower students facing decision-making barriers.The guidelines recommend continuity across a student's program rather than treating the material as a one-course intervention.Source: §5.3 · p. 18
  • Integrate ethics into technical skill-building: students valued technical examples and datasets for exploring ethical challenges, while this approach can address course overload.Students in open-ended responses appreciated using technical examples and datasets to examine ethical issues.Source: §5.3 · p. 18
  • Although students wanted to apply lessons at work, conflicts of interest and superiors' differing views may obstruct use, so educators should provide reusable workplace resources.Suggested resources include online readings, case studies, and local governing bodies.Source: §5.3 · pp. 18–19
  • Adapt HELIX and its materials to students' educational and work backgrounds, future program courses, and prior understanding of ethical decision-making.The authors describe incorporating human-centered material as both an opportunity and a challenge.Source: §5.3 · p. 19

6 Conclusion

The conclusion presents HELIX as a three-stage framework for integrating human-centered learning into information science data science education. The pilot suggests potential, but reports minimal change in knowledge and decision-making and calls for more rigorous evaluation.

6 Conclusion

  • Adding human-centered content alongside existing technical outcomes can risk overwhelming course loads.The conclusion situates HELIX amid the growth of data-related programs beyond computer science and the need for instruction on algorithmic justice, bias, societal impact, and morality.Source: §6 · p. 19
  • HELIX supports human-centered content through iterative knowledge-building, decision-making, and empowerment stages aimed at ethical decisions in data science.The framework is designed to support parallel learning outcomes across a course and degree; empowerment includes professional conduct, intellectual growth, lifelong learning, and collaboration.Source: §6 · p. 19
  • The authors tested HELIX in a single case study teaching introductory programming to graduate students.The conclusion notes that students came from varied educational backgrounds and degree streams, and many found it difficult to anticipate the embedded content’s relevance to future work and decision-making.Source: §6 · p. 19
  • Study participation required opt-in consent, and the study lacked a control group.The authors note that students most confident and comfortable with the concepts may have been more likely to opt in, while students who learned most may not have consented to use of their data.Source: §6 · p. 19
  • Students’ attitudes toward ethical responsibility improved, while knowledge and decision-making skills showed minimal change.The authors describe HELIX’s potential as a pedagogical instrument in interdisciplinary computing contexts, while highlighting the need for more rigorous evaluation.Source: §6 · p. 19
  • The authors encourage educators to adapt HELIX to integrate human-centered reflection with technical abilities and workplace decision-making.The conclusion frames this as an aim for data science courses, alongside the need for more rigorous evaluation.Source: §6 · p. 19

A.1.1 Demographics.

The demographics questionnaire asked about students’ data-science backgrounds and prior ethics study.

A.1.1 Demographics.

  • Students were asked to report their background in data science or related fields.The questionnaire listed professional experience as one selectable background.Source: §A.1.1 · p. 22
  • The questionnaire included professional experience and prior ethics study as background questions.It asked about general ethics and technology- or data-focused ethics coursework.Source: §A.1.1 · p. 22
  • Students were asked whether they had previously studied ethics.Response options distinguished general ethics courses from technology- or data-focused ethics courses.Source: §A.1.1 · p. 22

A.1.2 Knowledge Assessment.

The knowledge assessment used questions about algorithmic bias and human-centered data science concepts, alongside data minimization and ethical design trade-offs.

A.1.2 Knowledge Assessment.

  • Students were asked to identify algorithmic bias, with options distinguishing group disadvantage from data preferences or processing errors.The assessment also included a mathematical-error option.Source: §A.1.2 · p. 22
  • The assessment asked students to identify data minimization as keeping only the data needed for a specific purpose.The alternatives included collecting as much data as possible or sharing data freely across organizations.Source: §A.1.2 · pp. 22–23
  • A question asked why data and algorithms are not inherently neutral, contrasting social, historical, and political influences with purely mathematical explanations.Other options said bias is unavoidable or can be eliminated by a well-designed algorithm.Source: §A.1.2 · p. 23
  • A question examined how systemic inequalities can persist even in high-quality datasets.Its options contrasted this with explanations based on mathematical error, dataset size, model performance, or explicit discriminatory labels.Source: §A.1.2 · pp. 22–23
  • Students were asked about trade-offs in interpretability and explainability, including disclosure, scalability, and false trust.The assessment also asked about participatory approaches, including project complexity and the timing and value of stakeholder input.Source: §A.1.2 · p. 23

A.1.3 Ethical Scenarios.

The ethical scenarios ask students to choose among competing responses to AI systems, including modifying models, changing policies, adding oversight, or limiting deployment.

A.1.3 Ethical Scenarios.

  • Facial-recognition crime prevention asks students to weigh public safety against privacy and potential misuse, including whether broader social policies should address crime's root causes.Source: §A.1.3 · pp. 23–24
  • The hiring scenario asks how to respond when past employee data leads an AI tool to favor prestigious-university applicants and filter out non-traditional candidates.Options include leaving the algorithm unchanged, adding fairness constraints, changing hiring policies, or abolishing AI hiring tools.Source: §A.1.3 · p. 24
  • The outbreak-prediction scenario asks how to handle anonymized health and mobility data collected without informing users.Choices include using the data, offering later opt-out, strengthening privacy protections, or discarding the dataset.Source: §A.1.3 · p. 24
  • The loan scenario asks how to respond when historical lending patterns lead an AI approval system to disproportionately deny loans to low-income applicants.Choices include retaining or modifying the model, adding human decision-makers, or pausing AI use in lending.Source: §A.1.3 · p. 24
  • The self-driving-car scenario asks whether to protect passengers or stay the course when pedestrians could be harmed, with options including minimizing casualties or deferring to regulation.Source: §A.1.3 · p. 24
  • The workplace-monitoring scenario contrasts management's claimed burnout prevention and efficiency benefits with employees' concerns about invasive keystroke and email tracking.Choices range from accepting monitoring to requiring transparency and consent, prohibiting surveillance, or taking an industry-dependent approach.Source: §A.1.3 · p. 24
  • The city-infrastructure scenario asks how to balance traffic, energy, and pollution optimization with tracking citizen movements, offering deployment, opt-outs, oversight, or rejection.Source: §A.1.3 · pp. 24–25

A.1.4 Attitudes and Perceptions.

Students rate agreement with statements about ethical considerations, confidence, privacy, accountability, and understanding societal impacts of biased data and algorithms.

A.1.4 Attitudes and Perceptions.

  • The prompt asks for ratings on four statements from 1 (Strong Disagree) to 5 (Strong Agree), covering data-science ethics, confidence identifying ethical issues, field-of-study privacy, and company accountability for biased AI.Source: §A.1.4 · p. 25

A.1.5 Post-Course Only: Course Reflection Questions.

The appendix includes post-course reflection prompts and a rubric for scoring human-centered orientations in ethical-scenario responses.

A.1.5 Post-Course Only: Course Reflection Questions.

  • Post-course reflections ask how ethical data-science understanding changed, which challenges students anticipate, and how the course influenced their ethical decision-making.Source: §A.1.5 · p. 25

A.2 Scoring Rubric and Examples for Ethical Scenarios

  • The ethical-scenario rubric scores responses from 1 to 5, from least to most human-centered, based on consideration of human impact, agency, consent, privacy, and social justice.Its scale distinguishes efficiency-focused responses, middle-ground or oversight approaches, balanced technical and human concerns, and prioritizing human-centered concerns.Source: §A.2 · p. 25
Loading 2609.08982v1…