Charting Our Course- Information Professionals as AI Navigators
0 likes46 views
The document discusses the role of information professionals as AI navigators, emphasizing the need for ethical AI practices and risk management in libraries. It outlines key policies around AI usage, including issues of copyright, privacy, and the importance of transparency, fairness, and accountability. Furthermore, it highlights the necessity for community involvement and ongoing evaluation of AI tools to ensure they meet the needs of library services while safeguarding user trust.
Charting Our Course- Information Professionals as AI Navigators
- 1. Charting Our Course: Information Professionals as AI Navigators Brian Pichman
- 4. How will AI show up our world? Copyright and Privacy: • Evaluate the data available to the AI model, considering its sources and the impact of sourcing methods on model outputs. • Ensure your ‘text and data-mining addendum’ clearly outlines the terms and conditions for AI usage in your work. Advice and Expertise: • Identify the various AI tools available and their optimal applications. • Guide patrons toward services that incorporate robust risk management practices. Advocacy and Policy: • Address the dual nature of social media, which promotes open collaboration and spreads misinformation. Teach good epistemic practices to navigate this landscape. • Recognize the complexities in the Open Access movement, where information overload coexists with access issues, and strategize accordingly.
- 5. Use of AI in publishing • AI cannot be an author • Use of AI must be disclosed • No use of AI to create or edit images • No use in review or editing • AI cannot be an author • Use of AI must be disclosed • No use of AI to create or edit images • No use in review or editing • AI cannot be an author, nor can AI- authored publications be cited. • Use of AI must be disclosed, including the full prompts used • No use of AI to create images without permission • No use of AI in review • AI use must be disclosed. • AI can be used with a detailed description in the Methods section. https://www.linkedin.com/in/williamgunn/
- 6. Privacy concerns We also use data from versions of ChatGPT and DALL·E for individuals. Data from ChatGPT Team, ChatGPT Enterprise, and the API Platform (after March 1, 2023) isn't used for training our models. We will not train our models on any Materials that are not publicly available, except in two circumstances: If you provide Feedback to us and if your Materials are flagged for trust and safety review Gemini Apps use your past conversations, location, and related info to generate a response. Google uses conversations (as well as feedback and related data) from Gemini Apps users to improve Google products (such as the generative machine-learning models that power Gemini Apps). Human review is a necessary step of the model improvement process. Through their review, rating, and rewrites, humans help enable quality improvements of generative machine-learning models like the ones that power Gemini Apps. https://www.linkedin.com/in/williamgunn/
- 7. Risk management plans Amazon, Anthropic, Google, Inflection, Meta, Microsoft, and OpenAI agreed to test systems before release, collaborate with government and academia, invest in cybersecurity, build watermarking systems, publicly disclose capabilities, and research bias and privacy issues. • Safety Team for existing models • Preparedness Framework for frontier models • Assess and evaluate capabilities in persuasion, cybersecurity, CBRN threats, autonomous replication • Superalignment for AGI/ASI • Use AI to help align AGI Responsible Scaling Policy The plan outlines safety levels, ASL 1-5 and details plans to detect capabilities that have advanced to the next level and to decide whether and how the model should be deployed. Google mostly talks about cybersecurity and their research. Microsoft has a template for individual teams to design their own plans. Amazon has a set of tools to allow model builders to specify topics to be avoided and to understand how a dataset might lead to biased or unexpected outputs. https://www.linkedin.com/in/williamgunn/
- 9. Ethical AI • Definition: AI systems that adhere to agreed ethical principles ensuring fairness, transparency, and accountability. • Ethical AI preserves user trust and protects against harmful biases. • Core Principles: Transparency, justice & fairness, non-maleficence, responsibility, and privacy. • Common Ethical Issues: Bias in AI algorithms, privacy concerns, and decision-making transparency. • Relevance to Libraries: How can these issues affect library services, such as personalized recommendations and digital collections management?
- 10. Why does this matter • User Trust: Maintaining user trust is paramount for library services. • Social Responsibility: Libraries have a duty to promote inclusivity, prevent discrimination, and be sources of truth
- 11. Five Core Ethical Principles • Transparency: Making AI decisions understandable to users.
- 12. Five Core Ethical Principles • Justice & Fairness: Ensuring AI systems do not perpetuate inequalities.
- 13. Five Core Ethical Principles • Non-maleficence: Preventing harm to users by AI decisions.
- 14. Five Core Ethical Principles • Responsibility: Accountability for AI impacts.
- 15. Five Core Ethical Principles • Privacy: Safeguarding user data.
- 16. Developing Our Own AI For Library Use Opportunities Personalization / User Preferences Efficient Data Management Improved Accessibility Risks Privacy issues who owns the data? Bias Transparency of AI decisions
- 17. Mitigating Risks Audit how the AI is working Review responses, test the system, etc Diversified Data Sets diverse dataset training and involving community feedback Data Minimization Encryption Transparent Data Policies Explainable AI (XAI): A type of AI that provides insights into AI decision processes.
- 18. Case Studies – Anonymized Data • New York Public Library uses anonymized data to improve services without compromising individual privacy. • https://www.nypl.org/press/new-york-public-library-announces- participation-department-commerce-consortium-dedicated-ai
- 19. Case Study – AI Bias • A healthcare algorithm used by hospitals to prioritize patient care needs • This algorithm inaccurately concluded that Black patients were healthier than equally sick White patients because it used healthcare costs as a proxy for health needs • Black patients historically spend less on healthcare, and the algorithm discriminated against them, prioritizing White patients for further care • https://www.nature.com/articles/d41586-019-03228-6 • https://www.science.org/doi/full/10.1126/science.aax2342#:~:text=Bias%20occurs%20b ecause%20the%20algorithm,than%20equally%20sick%20White%20patients.
- 20. Case Study – Privacy Violation • Facebook and Cambridge Analytica incident serves as a potent example of AI-related privacy violations • Data from millions of Facebook users were harvested without consent and used for political advertising, highlighting significant privacy breaches
- 21. So…where do we go from here?
- 22. Selecting AI Tools Assess Needs Identify library services that could benefit from AI enhancement. Vendor Evaluation Choose AI vendors that adhere to ethical standards. Community Feedback Involve library users and staff in the selection process.
- 23. Data Collection and AI Models Data Ethics tells us to respect user privacy and offer consent • Help ensure diversity by representing all community segments Design (or choose) an AI System that is transparent where you can see how it responds to questions, how the model is trained, and what data sets its using
- 24. Grow and Monitor Conduct pilot tests to evaluate the AI performance before making live for everyone Use a feedback process to refine the responses Monitor the impact and effectiveness Does it cause an increase in program usage, circulation, and community members helped? Explain the role of AI in library services to users
- 25. Community Involvement Adapt strategies to identify ways to improve the AI tool Host sessions to education people about the use of AI If you are offering tools that use AI to the community, you will also want to teach them ethical use of AI (using it so it doesn’t cause harm to others or themselves)
- 26. Policies
- 27. Considerations for Policy Writing • Establish core ethical principles specific to library needs. • This could be safety and inclusion • Determine how you will review and update these policies • Use clear, accessible language to ensure all stakeholders understand the policies. • Define roles and responsibilities for enforcement and oversight. • Schedule regular policy reviews to adapt to new AI developments and community needs.
- 30. Training Provide comprehensive training on AI ethics and its importance. Launch campaigns to educate users on how AI is used in the library and its benefits. Promote transparency by making AI policies and practices accessible to the public. Actively seek input from the community on AI use in library services.
- 31. Recapping • It is important to be surgical in an approach to using AI – whether you’re developing it or purchasing a solution • The more communication you have around what is being done, the better the outcome and usage will be
- 32. AI Model Collapse / AI Degradation • Overtime, a model can get over saturated with “nonsense” data • Best solution – use an LLM and a RAG • Purge the LLM and refresh based on use • RAG Data doesn’t change, LLM data can with user input
- 33. Questions? • Brian Pichman • [email protected] https://links.evolveproject.org/sla2024