Technology is increasingly a crucial enabler for research. It is therefore important to understand how Generative Artificial Intelligence (GenAI) tools operate.
GenAI is a type of machine learning and artificial intelligence that can learn from and mimic large amounts of data to create content such as text, images, etc. based on user input in the form of prompts. GenAI tools can be used in many ways in your study or research - to synthesise, create, refine, test, inspire or modify.
Considering utilising AI in your research or study? Read this first:
- GenAI is based on Search Engine and Large Language Model technologies. It works by chaining words and sentences together, searching for patterns, and then generating responses. Whilst it is great in summarising larger amounts of text or answering questions based on those texts, it does not actually possess the knowledge or understand the subject. You should always validate GenAI responses.
- The quality of GenAI responses depends on the quality of both your prompts and your information-base. The more specific your information-base and prompts are, the higher the quality of your response.
Protect Confidential Data
Anything you provide to GenAI technology such as prompts, and information is processed and could be retained by that technology. Do not enter data that would be classified as “In-Confidence”, “Sensitive” or “Special” under the UC data classification table in public GenAI tools. This information is not private and could be collected by unauthorized parties.
Please consider the following constraints and considerations prior to entering data into public GenAI tools:
- Privacy Protection: Does the dataset contain sensitive information that requires protection to uphold the privacy rights of individuals involved? Ensure that any personal or confidential information is handled in accordance with applicable data protection regulations.
- Confidentiality Obligations: Are there any Confidentiality Agreements that may be associated with the data? Avoid disclosing any information that could compromise confidentiality agreements with third parties or affiliated institutions.
- Third-Party Data: Acknowledge and respect any third-party data included in the dataset. Comply with data sharing / licensing agreements and attribution requirements associated with third-party data sources.
- Commercialisation: Consider whether the data is associated with a project which may have a commercial outcome which could be prejudiced by the release of the data.
- Ethical Considerations: Ensure that the use of the data aligns with ethical standards and guidelines within the academic community. Seek approval from relevant ethical review boards if the research involves human subjects or sensitive topics.
- Compliance with UC’s Policies: Verify whether the release of the data complies with institutional rules and policies.
- Security Measures: Take appropriate security measures to prevent unauthorised access, modification, or misuse of the dataset. Clearly state any restrictions or conditions for accessing and using the data to maintain its integrity.
- Attribution and Citation: Provide proper attribution to the original data source and cite the dataset appropriately in any publications or derivative works.
UC has procured secure versions ChatGPT and Microsoft Copilot for use at UC. For tools endorsed by UC Digital Services, such information is generally protected from misuse. For guidance on using GenAI tools, please reach out to your eResearch Consultant.
Review Content Before Publication
A primary function of GenAI technologies is to "make things up" - hence "generative". It generates a response based on applying your prompt against an information-base. There is every possibility that there is inaccuracy in a GenAI response. Again, you should always validate GenAI responses (e.g. fact-checking, verifying references and information sources, applying your own understanding of the subject, etc).
AI-generated content can be inaccurate, misleading, or entirely fabricated (hallucinations) or may contain copyrighted material. The responsibility for published content rests with the person that publishes it, so be careful when publishing AI-generated content.
Be Alert for Phishing Attempts
Generative AI has made it easier for malicious actors to create sophisticated phishing emails and video or audio (called deepfakes) that can convincingly mimic a person’s voice and physical appearance without their consent. [Maybe link to cybersecurity guidance]
Connect with an eResearch Consultant
The University of Canterbury is working to ensure that GenAI tools procured have the appropriate and security protections for its staff and students. If you have any questions about other GenAI tools available that is not endorsed by UC Digital Services, please contact your eResearch Consultant for more information.
