COPENHAGEN SYMPOSIUM
The Copenhagen Manifesto
Generative AI in Software Engineering Must Be Human Centered
The advent of Generative Artificial Intelligence marks a pivotal moment for software engineering. Systems that can produce text, visual art, and source code are reshaping how software is built, yet the human aspects of developing software have historically been treated as secondary to technical innovations. The Copenhagen Manifesto reaffirms the primacy of human dignity, agency, and collective wellbeing in the face of rapid technological change.
Developed at the Copenhagen Symposium on Human-Centered AI in Software Engineering. Funded by the Alfred P. Sloan Foundation and the Carlsberg Foundation. Published April 2024.
Core Values
The Copenhagen Manifesto is built on five core values that guide the responsible development and use of Generative AI in software engineering.
1. Responsibility and Ethics. Emphasizing the duty to develop Generative AI in SE in a manner that does not bring physical, emotional, or financial harm to living beings.
2. Human-Centricity and Agency. Prioritizing human needs and autonomy when designing and using Generative AI in SE.
3. Transparency and Equity. Advocating for transparent, reproducible, and verifiable Generative AI in SE, including communications about research in this area, and equitable access and impact.
4. Inclusivity and Continuous Learning. Adopting, adapting, and empirically researching Generative AI in SE must inherently foster a culture of diversity, inclusivity, and continuous learning, embracing diverse, multidisciplinary perspectives for software engineering’s advancement.
5. Environmental Sustainability. Committing to the responsible selection and use of Generative AI technologies, with a focus on minimizing environmental impact through energy efficiency and sustainable practices.
Principles
Twelve principles translate the core values into actionable commitments for researchers, practitioners, and organizations working with Generative AI in software engineering.
1. Responsible Management of Generative AI in SE. Actively engage in responsibly managing and continuously evaluating Generative AI technologies, ensuring their alignment with ethical standards and societal needs.
2. Human Sovereignty Over AI. Ensure that Generative AI technologies enhance, rather than replace, human decision-making and creativity, prioritizing human oversight in Generative AI development and application.
3. Two Sides of Generative AI. Assess both the benefits and harms of Generative AI, promoting its use with caution and responsibility to avoid unintended consequences.
4. Sociotechnical Responsibility. Integrate social and technical considerations in the development of Generative AI-powered applications, aiming for solutions that are beneficial and respectful to all stakeholders.
5. Transparency and Fairness. Implement transparent processes that actively identify and mitigate Generative AI-related biases, ensuring fairness, accountability, and trustworthiness in Generative AI applications.
6. Sustainability and Environmental Impact. Select and advocate for Generative AI models and practices known for their lower environmental impact, emphasizing long-term sustainability.
7. Equitable Impact of Generative AI in Software. Champion Generative AI that fosters inclusivity and equity, working to eliminate disparities and support empowerment across diverse communities.
8. Ethical Engineering and Education. Uphold ethical standards in both engineering practices and education, ensuring future generations of engineers are prepared to make ethical decisions in Generative AI use.
9. Complex Adaptive Nature of Generative AI. Recognize and plan the evolving nature of Generative AI systems by committing to ongoing learning and adaptation in their design and application, ensuring they remain aligned with ethical and societal values.
10. Empirical Evaluation of Professional Practices. Support and conduct empirical research to base Generative AI adoption and practices on evidence, enhancing the positive impact of Generative AI on society and addressing any challenges.
11. Enhancement and Wellbeing through Adaptive Generative AI in SE. Develop Generative AI that amplifies the capabilities of software engineers, promoting professional growth and wellbeing.
12. Public Awareness and Discourse. Engage in and foster public discussions about Generative AI, aiming to educate and inform while combating misinformation, thereby shaping a balanced and factual narrative around Generative AI in SE.
Context and References
The Copenhagen Manifesto draws on foundational ethical frameworks for computing and engineering, including the ACM Code of Ethics, the IEEE Code of Ethics, and the EU AI Act. It extends these frameworks to address the specific challenges and opportunities that Generative AI presents for the practice of software engineering.
Download
The full Copenhagen Manifesto is available as a PDF for reference and citation.
The Copenhagen Manifesto was developed at the Copenhagen Symposium on Human-Centered AI in Software Engineering. To express interest in the next edition, get in touch.