Key AI Ethics¶
According to the Living Guidelines on the Responsible Use of Generative AI in Research, researchers should ensure responsible use of generative AI by:
1. Remain ultimately responsible for scientific output¶
- Researchers are responsible for the integrity of AI-generated or AI-supported content
- Researchers must recognize AI's limitations, including bias, hallucinations, and inaccuracies.
- AI systems are neither authors nor co-authors. Authorship implies agency and responsibility, so it lies with human researchers.
- No fabricated AI material: Researchers must not falsify, alter, or manipulate **original research data using AI.
2. Use generative AI transparently¶
- Transparency: Researchers disclose generative AI tools used (name, version, date, usage, impact) and, if relevant, share input and output per open science principles.
- Account for randomness: Recognize AI's stochastic nature and ensure reproducibility and robustness.
- Acknowledge limitations: Discuss AI tool biases, limitations, and mitigation measures.
3. Focus on Privacy, Confidentiality, and intellectual property¶
- Protect sensitive work: Avoid uploading unpublished or sensitive content into AI systems unless data reuse is assured.
- Respect personal data: Only input personal data with clear consent and a specific purpose, ensuring compliance with EU data protection rules.
- Understand implications: Be aware of privacy, confidentiality, and intellectual property concerns. Verify privacy settings, tool ownership, hosting environment, and data security measures.
Note: We recommended to use BlaBlaDor for you research in terms of its privacy focus (Blablador hosts both open-source models and those created by the Helmholtz Association).
4. Comply with Laws and Regulations¶
- Intellectual property and data sensitivity: Generative AI outputs may pose risks to intellectual property rights and personal data protection.
- Avoid plagiarism: Ensure proper citation and respect for others’ authorship; AI outputs may draw on existing works.
- Handle personal data responsibly: Address any personal data in AI outputs in line with EU data protection rules.
5. Learn and Train for Effective AI Use¶
- Stay informed: Researchers keep up with evolving generative AI tools and share best practices with colleagues and stakeholders.
6. Avoid AI in Sensitive Activities¶
- Mitigate risks: Avoiding generative AI prevents unfair treatment or assessment caused by hallucinations and biases.
- Protect unpublished work: It ensures the confidentiality of original research, shielding it from unintended exposure or AI model inclusion.