A common reference for practical work
The Principles for Responsible AI are the first step in Ethical AI Macedonia’s work. They bring together established ideas about responsible technology and give us a common reference for discussing how those ideas apply in North Macedonia. Their practical usefulness will become clearer as organizations examine them against actual tasks, constraints and experiences. We want those discussions to influence both future revisions of the principles and the subjects we investigate.
A business may agree with a principle while remaining uncertain about what it requires in a particular workflow. Human oversight, for example, raises further questions about who should review an output, what expertise that person needs and what information they must be able to access. Similar questions arise when an organization considers confidentiality, fairness or accountability. Exploring those details is an important part of moving from broad agreement towards practices people can use.
Discussions shaped by business experience
Over the coming year, we intend to organize further discussions involving businesses, technical practitioners, researchers and others with relevant experience. The November event is an initial opportunity to bring these perspectives together. We want participants to explain the applications they are considering, the difficulties they encounter and the safeguards they have found useful. Disagreement can be valuable when it helps identify different assumptions about how a tool works or what a particular risk requires.
These discussions should also give attention to successful applications. There is much to learn from businesses that have introduced AI thoughtfully and can explain how it improved their work. Understanding their choices, the limitations they encountered and the changes they made may be more useful to another organization than a general account of AI’s potential. We want to recognize those experiences and examine what can reasonably be learned from them.
Research and use cases
Company research will help us understand adoption beyond the experiences of the founders and individual participants. Our initial questionnaire provides a starting point, and we intend to explore more specific challenges in particular industries. Working groups can help formulate useful questions, reach relevant businesses and interpret what participants tell us. Their composition should reflect the subject being examined, bringing together practical knowledge and suitable research expertise.
We also plan to publish use cases that describe how AI is applied in a particular setting. A useful case should explain the task, the information involved, the benefits sought, the difficulties encountered and the way responsibility is managed. It should give readers enough context to judge whether the experience is relevant to their own organization. Over time, these cases can help us identify recurring problems and examples of practices worth discussing more widely.
Building the capacity to assist organizations
A longer-term ambition is to develop the expertise and team needed to help organizations establish their own internal approaches to responsible AI. Businesses need to consider how general principles fit their operations, who should make decisions and how safeguards will work in practice. That may involve internal procedures, clear responsibilities, staff training and records of significant decisions.
Our current discussions and research are intended to build the foundations for that assistance. We want to understand the range of organizational circumstances before presenting detailed approaches as generally useful. The work should help us develop guidance that is proportionate, informed by experience and capable of improvement as businesses and technologies change.