There are ethical and regulatory considerations for the use of generative Artificial Intelligence (AI) in research involving human subjects. Researchers and the IRB must carefully evaluate risks related to privacy, consent, and secondary use of data when considering the use of AI in human subjects research. Please consider the following:
Deidentification and Privacy: It is best practice to de-identify human subjects data before using AI whenever possible. In your IRB application, please describe how data will be de-identified prior to AI use.
Even when identifiers are removed, AI tools may infer personal information from patterns in the de-identified data. Additionally, metadata can be identifiable. File names, geolocation, timestamps, or other metadata can inadvertently disclose sensitive information. All metadata should be removed prior to AI use. Please describe these procedures in your IRB materials.
Bias & Hallucinations: AI may yield biased and inaccurate information. In the IRB materials, researchers must verify how conclusions derived from AI will be verified. In the IRB materials, please also describe how research methods using AI can be replicated.
Informed Consent and Disclosure: Use of AI to analyze or process data must be disclosed to participants as part of the consent process. The consent language must describe which AI tools will be used during or after the study, and whether AI tools will be used to process identifiable or de-identified data.
Use of AI must also be disclosed to the scientific community in publications and presentations.
Conducting research with non-English speakers: AI may not be used as a substitute for an interpreter or a fluent/certified translator when conducting research with participants who are not comfortable speaking or understanding English. See Guidance: International Research and Research with Non-English Speaking Populations.
If AI is used for translations of study materials, the AI translation must be verified via a human translator who is fluent or certificated (required for greater than minimal risk research). See Guidance: International Research and Research with Non-English Speaking Populations.
Approved AI Services: Lehigh offers a selection of contracted AI services. Services that are acceptable for Class II data may be used in human subjects research upon approval from the IRB. In order to avoid security breaches, you must access these AI services via your Lehigh account. If you wish to use a service that has not been approved for human subjects research, please consult with LTS to determine whether the AI service meets data security requirements.