Source-linked AI summary
Embedded EthiCS: Integrating Ethics Broadly Across Computer Science Education
Barbara J. Grosz, David Gray Grant, Kate Vredenburgh, Jeff Behrends, Lily Hu, Alison Simmons, Jim Waldo
TL;DR
Computing technologies can produce unintended harmful consequences, creating a need for computer science students to reason about what technologies should be built. Embedded EthiCS embeds philosophy-led ethics mini-modules in existing computer science courses, and the pilot received strong student and faculty support, though long-term effectiveness and institutional implementation remain challenges.
Problem
Students may lose sight of ethical concerns during technical design work, motivating curriculum-wide ethical reasoning about computing’s social impact.
Method
Embedded EthiCS pairs philosophy teaching staff with computer science faculty to design and deliver ethics mini-modules within existing core courses.
Results
Over 80% of students in every pilot course found the sessions interesting, and faculty reported modest burden alongside learning from the modules.
Takeaways & Limitations
The approach gives students repeated practice identifying, analyzing, and communicating about ethical issues across computing contexts.
Takeaways & Limitations
The pilot did not yet measure whether recurring exposure changes ethical reasoning over years or compare the approach with alternatives.
Abstract
from arXiv · showhide
Computing technologies have become pervasive in daily life, sometimes bringing unintended but harmful consequences. For students to learn to think not only about what technology they could create, but also about what technology they should create, computer science curricula must expand to include ethical reasoning about the societal value and impact of these technologies. This paper presents Embedded EthiCS, a novel approach to integrating ethics into computer science education that incorporates ethical reasoning throughout courses in the standard computer science curriculum. It thus changes existing courses rather than requiring wholly new courses. The paper describes a pilot Embedded EthiCS program that embeds philosophers teaching ethical reasoning directly into computer science courses. It discusses lessons learned and challenges to implementing such a program across different types of academic institutions.
Why Embed Ethics and Philosophers in the Teaching of Computer Science?
Embedded EthiCS was designed to habituate students to ethical reasoning by integrating it throughout existing computer science courses rather than isolating it in a one-off course. Its interdisciplinary model connects technical knowledge of what can be built with normative reasoning about what should be built.
- Why Embed Ethics and Philosophers in the Teaching of Computer Science?: Students can lose sight of ethical concerns during technical design tasks, motivating repeated ethics instruction within computer science courses.The program was created after instructors observed this difficulty even among students committed to ethical reasoning.
- Why Embed Ethics and Philosophers in the Teaching of Computer Science?: Embedded EthiCS adds ethics mini-modules across the core curriculum to show that ethical issues permeate computing and to provide repeated practice in reasoning and communication.The approach also exposes students to concrete ethical problems across computational contexts.
- Why Embed Ethics and Philosophers in the Teaching of Computer Science?: The authors expect curriculum-wide repetition to reinforce ethical reasoning as an expected part of computer scientists’ work, while acknowledging that one module alone is insufficient.This expected compounding effect is presented as the rationale for embedding modules broadly.
- Why Embed Ethics and Philosophers in the Teaching of Computer Science?: The approach is interdisciplinary because computer science addresses what technology can do, while ethics helps students evaluate and defend what technology should do.The paper presents both forms of understanding as necessary for learning how to build systems responsibly.
How Does Embedded EthiCS Work?
Embedded EthiCS pairs computer science courses with philosophy teaching staff to design and deliver focused ethics mini-modules. These modules develop students’ abilities to identify ethical problems, reason about them, and communicate their approaches.
- How Does Embedded EthiCS Work?: A philosophy Ph.D. student or postdoctoral fellow works with the computer science course head to create an ethics mini-module tied to course content.They select an ethically significant topic, develop case studies, and design an assignment and assessment.
- How Does Embedded EthiCS Work?: The fellow leads one or two class sessions and designs the associated assignment and assessment, with grading handled by course staff, the fellow, or peers.The grading arrangement depends on class size.
- How Does Embedded EthiCS Work?: The modules target three skills: identifying ethical problems, reasoning about them and possible solutions using moral philosophy, and communicating conclusions effectively.These skills apply to the development and use of computing technologies.
Embedded EthiCS Pilot
The Embedded EthiCS pilot ran across multiple computer science courses and connected course-specific technical topics with ethical questions. Four highlighted examples addressed censorship, privacy, inclusive design, and discrimination in machine learning.
- Embedded EthiCS Pilot: Three-semester pilot participation included 14 separate courses, with three courses participating twice.The pilot covered Spring 2017, Fall 2017, and Spring 2018.
- Embedded EthiCS Pilot: A networks course examined whether social media companies are morally obligated to suppress so-called fake news.The module used ethical reasoning about censorship by social media companies.
- Embedded EthiCS Pilot: A big data systems course considered how large datasets and analytical tools complicate privacy protection through powerful inference capabilities.The course served graduate students and advanced undergraduates.
- Embedded EthiCS Pilot: A human-computer interaction design course focused on inclusive design for people with disabilities and developers’ possible moral obligations to design for inclusion.The class began by examining different meanings of inclusive design.
- Embedded EthiCS Pilot: An introductory machine learning course addressed theories of wrongful discrimination, accuracy’s limits, and formalized fairness criteria.The module also introduced students to implications of an impossibility result.
Embedded EthiCS: Assessment of the Pilot-Program Class Modules
The pilot assessment found that ethics could be integrated into core computer science teaching and was viewed positively by students and faculty. Surveys showed strong interest in the sessions and in further ethics education.
- Embedded EthiCS: Assessment of the Pilot-Program Class Modules: The pilot showed that integrating ethical reasoning with core computer science methods was possible and rewarding for students and faculty.Faculty also reported that the modules imposed only a modest burden and that they learned from them.
- Embedded EthiCS: Assessment of the Pilot-Program Class Modules: Over 80% of students in every course agreed that the Embedded EthiCS sessions were interesting.More than 90% agreed in five classes.
- Embedded EthiCS: Assessment of the Pilot-Program Class Modules: In all but two classes, more than 80% of students reported interest in learning more about ethics in future computer science courses.Student comments were overwhelmingly positive, with many requesting more ethics content and reasoning practice.
From the Pilot to a Sustainable Program
The pilot expanded Embedded EthiCS across computer science courses while revealing implementation, cross-disciplinary, staffing, assessment, and institutional challenges. Emerging practices emphasize active learning, faculty involvement, reusable materials, and sustained evaluation.
- Program growth: Four pilot classes in spring 2017 expanded to 10 additional courses in AY 2017–18, while three original courses continued modules.One Ph.D. student developed the first modules; two Ph.D. students supported the later expansion.
- Student engagement: Embedded EthiCS works best when ethical issues connect to salient technical material, sessions use short participatory activities, and assignments provide practice.These design choices were identified as supporting student engagement with ethical reasoning.
- Faculty roles: Faculty support modules by co-designing them, assigning graded work, attending sessions, and revisiting ethical issues elsewhere in the course.Graded assignments and faculty presence signal that ethics is core course content rather than an optional supplement.
- Teaching assistant experience: Philosophy teaching assistants can embed modules in three to four courses per term, with the workload roughly comparable to teaching two discussion sections.Participating assistants also reported substantial benefits from exposure to computer science concepts and methods.
- Cross-disciplinary barriers: Cross-disciplinary vocabulary differences can confuse students, making computer science faculty and teaching assistants important throughout module design and implementation.Their involvement helps align ethics material with course vocabulary and identify philosophical material needing more explanation.
- Sustainability challenges: Recruiting philosophers, securing university funding and administration, and measuring long-term effectiveness remain challenges for broader implementation.The program seeks evaluation beyond post-module surveys, including comparisons with other approaches over students’ degrees and careers.
Looking Forward
Embedded EthiCS integrates ethical reasoning with computational techniques so students practice addressing ethical questions throughout computer science work. The authors call for broader adoption supported by shared repositories and institutional learning.
- Educational rationale: Integrating ethical reasoning into existing coursework gives students practice identifying, confronting, and working through ethical questions alongside computational techniques.The approach begins ethical problem-solving at the design phase, alongside developing efficient algorithms and programs.
- Long-term consequence: The approach communicates that ethical questions should be addressed throughout computer science work, including after graduation.The authors connect these skills with socially and ethically responsible technology work.
- Broader adoption: The authors invite other institutions to integrate ethics across their curricula and expand open repositories of modules, activities, assignments, and readings.They also propose sharing lessons about institutional support and philosopher training.
Appendix A
The appendix presents two broadly multidisciplinary courses that integrate substantial ethics and computer science material, illustrating a more extensive model that Embedded EthiCS distributes across the curriculum.
- Course model: “Privacy and Technology” and “Intelligent Systems: Design and Ethical Challenges” combine significant ethical and computer science material for students with varied backgrounds and interests.Embedded EthiCS is presented as pursuing similar extensiveness in a distributed manner across existing courses.
1. Privacy and Technology
“Privacy and Technology” examines how technological advances challenge privacy concepts, laws, and policies. Students assess technologies’ capabilities, privacy threats, and possible technological or legal responses through applied group assignments.
- Course focus: The course examines technological advances that challenge intuitive privacy notions and the laws and policies intended to protect privacy.It serves students pursuing law, policy, or technology careers.
- Analytical approach: Students evaluate what technologies can do, whether they pose genuine privacy threats, and whether responses should be technological, legal, or policy-based.Assignments include position, policy, or briefing papers and technology-assessment exercises, often completed in groups.
- Final projects: Final projects included re-identifying purportedly de-identified datasets, mapping drug-related photographs from Silk Road advertisements, and analyzing privacy laws and India’s Aadhar system.These projects apply the course’s technology and policy analysis to concrete cases.
2. Intelligent Systems: Design and Ethical Challenges
This course interweaves artificial-intelligence methods with ethical analysis, so students examine both how systems work and the design challenges they raise. Its projects require technical or ethical work to attend to both dimensions.
- The course combines an introduction to current and potential AI uses with strategies for addressing the design and ethical challenges they raise.Readings, discussions, and assignments cover AI methods and help students identify technologies’ strengths and weaknesses.
- Four course sections interweave technical topics and ethical issues through readings, discussions, and assignments.One example connects planning and decision-theoretic reasoning, including Markov decision processes and reinforcement learning, with autonomous-vehicle design challenges.
- Students learn broad ethical theories alongside applied-ethics readings rather than treating ethical reasoning as theory application alone.The course uses ethical theories to attune students to features of particular cases.
- Final projects integrate technical and normative thinking by focusing primarily on either a technical or ethical problem while addressing both.Examples include a Naive Bayes natural-language-processing system and an ethical analysis of news recommenders proposing design alternatives for fake news.
Appendix B: Embedded EthiCS Pilot Evalution Summary
Figure 2 reports the percentage of responding students in each course who agreed with each student-evaluation statement. It uses Likert responses and distinguishes repeated offerings for CS134 and CS179.
- Figure 2 shows the percentage of responding students who agreed with each evaluation statement for each course.Agreement is based on Likert responses from 1 to 7, where 7 means strongly agree and 1 means strongly disagree.
- CS134 includes results from both offerings in chronological order, while CS179 shows only its initial survey results.The later CS179 survey had higher percentages in every category.