Rapid advancements in Artificial Intelligence (AI), and generative AI (genAI) specifically, are revolutionizing the research landscape, ushering in unprecedented possibilities and challenges. This page aims to provide UCSB researchers and scholars with basic guidance, information on available resources, and contacts that can help you ensure your use of these tools is productive, safe, and compliant. The resources on this page will be regularly updated to keep pace with the rapidly evolving technology field.
AI@UCSB
Center for Responsible Machine Learning
Creating new software and methods that benefit society. The Center works on making AI and ML fair, unbiased, private, transparent, and accountable. It also studies the ethical, policy, legal, and energy-efficiency aspects of machine learning.
REAL AI
REAL AI is pioneering an approach to reliable, ethical, accessible and long-shot AI.
Geometric Intelligence Lab
Using tools from geometry, topology, computer vision, machine learning and deep learning to reveal the structures of intelligence.
AI Institute for Agent-based Cyber Threat Intelligence and Operation (ACTION)
UCSB leads this NSF funded Institute as a collaboration among more than 20 research institutions focused on how AI can anticipate and interrupt cybercrime.
Mellichamp Initiative in Mind & Machine Intelligence (MMI)
Brings together faculty from Psychological & Brain Sciences, Computer Science, and Economics to study how the human mind and intelligent machines compare and interact
Frequently Asked Questions
Various academic publishers have different rules regarding the use of AI in research papers. Some may even ban the use of AI for certain parts of the paper creation process. It's important to check the specific policies of the intended publisher to understand what is allowed.
Here are some links to specific publisher policies:
Funding agencies have also begun to provide guidance on the use of AI in developing research proposals:
Scientific societies have also developed policies regarding the use of AI tools in research and publication preparation:
- AGU [TBD: link to be added once policy is approved]
AI systems are rapidly evolving, and guidance is likely to change. However, some leading style guides have offered recommendations for appropriately citing the use of AI-generated content:
Yes. Some of the main risks and safety concerns include:
- Bias and Discrimination: AI systems have the potential to exhibit unfair or discriminatory behavior.
- Misinformation, Impersonation, and Manipulation: There is a risk of AI systems spreading false or misleading information, or being used to deceive or manipulate individuals.
- Research and IP Compliance: AI systems must adhere to legal and ethical guidelines when using proprietary information or conducting research.
- Security Vulnerabilities: AI systems are susceptible to hacking or unauthorized access.
- Unpredictability: It can be difficult to predict the behavior or outcomes of AI systems.
- Overreliance: There is a risk of relying too heavily on AI systems without considering their limitations or potential errors.
See the Recommendations to Guide the University of California's Artificial Intelligence Strategy for more information.
Yes. You should not be providing any University protected data to AI systems that may incorporate that information into their training systems. Please use our Research Data Security guide [TBD: link] to better understand this issue.
In addition, funding agencies and publishers have begun to develop disclosure policies regarding the use of generative AI in the authorship of research products and proposals. Finally, the use of generative AI in peer review has been prohibited by both the NIH and the NSF.
General Guidelines for the Use of AI in Research
Before using AI in research, it is essential to understand the ethical and responsible practices associated with it. These include adhering to a duty of discretion and a duty of verification, and avoiding unintentional AI plagiarism.
Your use of GenAI and other AI tools should be based on a judicious selection of appropriate AI tools and data. Approaches to ensuring the thoughtful and responsible use of AI in research should include:
- Assessing the suitability of AI tools for specific research tasks. How well do you understand the AI tools' underlying algorithms and data sets and their capabilities, limitations, and context?
- Researchers should exercise discretion in choosing appropriate data for AI-related tasks. You should ensure that data used in training and applying AI tools is ethically sourced and free from biases that could skew research outcomes. You must also safeguard sensitive or proprietary data to prevent misuse or breaches.
- You must adhere to ethical guidelines and institutional policies that govern the use of AI. This includes considerations for data privacy, intellectual property, copyright, and avoiding potential conflicts of interest.
You are responsible for verifying the accuracy and consistency of AI-generated content to ensure the integrity and reliability of your research outcomes. Approaches to ensuring AI-assisted research should include:
- Conduct rigorous checks to confirm the accuracy and validity of results generated by AI tools. This may include cross-validation with other data sources, statistical testing, and replicating results using different methods or datasets.
- You should thoroughly understand an AI model's mechanisms, including its limitations and the assumptions on which it is built. Test the model under different scenarios to understand how input changes affect outputs.
- You should be transparent about your research's AI tools and methodologies. This includes detailed documentation of the AI model's configuration, training, deployment, and any modifications made during your research.
You should avoid unintentional plagiarism by authoring original text and then only using GenAI tools for refinement. This approach follows the distinctions identified by van Dis, Bollen, Zuidema, et al., 2023, who highlighted the potentially beneficial use of AI-assisted text generation in contrast to the much riskier use of AI-written text.
Additional Resources
- Recommendations to Guide the University of California's Artificial Intelligence Strategy
- UCSB Office of the CIO Guidelines for AI Use and FAQs
Generative AI Tools Available to the UCSB Community
UCSB offers several generative AI tools under campus agreements that keep your data out of vendor training. Always sign in with your UCSB account: personal accounts lack these protections. For current eligibility and approved data protection levels, see the CIO's GenAI Tools at UCSB page.
- UCSB AI Commons: A campus-supported environment for working with a range of generative AI models, including custom bots built with your own instructions and reference documents. It is centrally funded, and each user gets a monthly usage allowance.
- Google Gemini and NotebookLM: Included with your UCSB Google account. Gemini is a general-purpose assistant, and NotebookLM summarizes and answers questions about documents you upload. Under UCSB's agreement, people do not review your chats and files, and Google does not use them to train its models. Departments can buy Google AI Pro for Education licenses for more features and higher limits.
- Zoom AI Companion: Available to faculty and staff through UCSB Zoom accounts, for features such as meeting summaries.
AI Commons, Gemini, and NotebookLM are approved for protected data, but P4 data requires a consult with the Office of Information Security first. Zoom AI Companion is approved up to P3. For research data, upload only de-identified, non-confidential data (P1 or P2), per the CIO AI FAQ. Protection levels follow the UC Protection Levels.
Cloud AI Platforms for Research Computing
Researchers building their own AI applications can request access to cloud AI platforms, such as Azure OpenAI, Google Vertex AI, AWS Bedrock, and AWS SageMaker. Contact Research Computing and Data or ITS to discuss options and data requirements.