class: center, middle, inverse, title-slide # EU Ethics Guidelines for Trustworthy AI --- We hear frequently of the challenges with AI and even simple predictive algorithms share the same issues. Recently, The EU came up with a set of guidelines to help diminish the chance of ethical issues in such algorithms. --- ## 1. Human agency and oversight Users should be able to make informed autonomous decisions regarding AI systems. They should be given the knowledge and tools to comprehend and interact with AI systems to a satisfactory degree and, where possible, be enabled to reasonably self-assess or challenge the system. Humans are involved in the design and governance of the system including when and where the system is used. --- ## 2. Technical robustness and safety Including resilience to attack and security, fall back plan and general safety, accuracy, reliability with a range of inputs and reproducibility of results with same inputs. The system is well-tested during creation and reviewed during use. --- ## 3. Privacy and data governance Including respect for privacy, quality and integrity of data, and access to data. Data must be ethically cleaned before it is used. --- ## 4. Transparency The data sets on which the decision making is based must be well-documented with regard to origin and alteration (such as in cleaning). Decisions made by AI must be understandable and able to be traced. Users should have access to the decision-making process and be aware that the decision was made, at least in part, but a computer. --- ## 5. Diversity, non-discrimination and fairness Including the avoidance of unfair bias, accessibility and universal design, and stakeholder participation. Diversity should be encouraged and designers should consider carefully whether past positive results (success of people with certain gender, ethnicity, personality etc.) were related or unrelated to the success. Avoid propagation of past bias. Systems should be user-centric and accessible to the widest range of users. People who will be affected by the system should be involved throughout the design of the system. --- ## 6. Societal and environmental wellbeing Including sustainability and environmental friendliness, social impact, society and democracy. Supply chain for the system should be sustainable. Its use should impact social relationships in a negative way. It should enhance, rather than detract from, a democratic society. --- ## 7. Accountability This necessitates that mechanisms be put in place to ensure responsibility and accountability for AI systems and their outcomes, both before and after their development, deployment and use. Potential impacts should be minimized and when encountered, they should be addressed and there should be a policy for making amends to the injured party. The system should be adjusted to eliminate the possibility of this happening in the future. When a conflict of interest arises, the tradeoffs made should be acknowledged.