How can libraries ensure transparent and accountable use of AI?

1 Define your goals and values

Before adopting or implementing any AI or ML solution, you should clarify your goals and values as a library. What are you trying to achieve with AI? How does it align with your mission, vision, and values? How does it benefit your users and stakeholders? How does it respect their rights and preferences? By defining your goals and values, you can set clear and consistent criteria for evaluating and selecting AI or ML tools and vendors, as well as for monitoring and measuring their impact and performance.

2 Assess the risks and benefits

AI and ML are not inherently good or bad, but they can have positive or negative consequences depending on how they are designed, developed, deployed, and used. Therefore, you should conduct a thorough and systematic assessment of the risks and benefits of any AI or ML solution before, during, and after its implementation. Some of the questions you should ask are: What are the potential harms and benefits of the solution for your users, staff, and community? How likely and severe are they? How can they be prevented, mitigated, or remediated? How can they be balanced and weighed against each other? How can they be communicated and reported?

3 Adopt ethical principles and standards

To ensure transparent and accountable use of AI, you should adopt ethical principles and standards that guide your decision-making and actions. These principles and standards should reflect your goals and values, as well as the best practices and norms of your profession and society. For example, you can refer to the ALA Code of Ethics, the IFLA Statement on Artificial Intelligence and Data Mining, or the OECD Principles on AI. You should also align your principles and standards with the relevant laws and regulations that apply to your jurisdiction and sector.

4 Involve your users and stakeholders

AI and ML affect not only your library, but also your users and stakeholders, such as patrons, staff, partners, funders, regulators, and suppliers. Therefore, you should involve them in the planning, implementation, evaluation, and governance of AI or ML solutions. You can do this by soliciting their feedback, input, and consent; informing them of their rights and responsibilities; educating them about the benefits and risks of AI or ML; and empowering them to participate in the oversight and review of AI or ML outcomes and processes.

5 Document and disclose your practices and policies

Transparency and accountability require that you document and disclose your practices and policies regarding AI or ML. This means that you should keep records of the data, methods, models, algorithms, and results of your AI or ML solutions; explain the rationale, purpose, and limitations of your AI or ML solutions; and report the performance, impact, and issues of your AI or ML solutions. You should also make your documentation and disclosure accessible, understandable, and verifiable for your users and stakeholders, as well as for external auditors and regulators.

6 Review and update your practices and policies

AI and ML are dynamic and evolving technologies that require constant review and update of your practices and policies. You should regularly monitor and measure the effectiveness, efficiency, and equity of your AI or ML solutions; identify and address any gaps, errors, or biases in your AI or ML solutions; and adapt and improve your AI or ML solutions to changing needs, expectations, and standards. You should also seek feedback and input from your users and stakeholders, as well as from peers and experts, to learn from their experiences and insights.

7 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?