HOUSE_OVERSIGHT_011302.jpg

Origins February 24 — 26, 2017 PROJECT An Origins Project Scientific Workshop ARIZONA STATE UNIVERSITY Challenges of Artificial Intelligence: Envisioning and Addressing Adverse Outcomes these data sources, the phenomenal success of Al-consult systems leads to strong public and policy support for widespread access. By 2030, Al-consult systems also have similar access to individual healthcare provider data. This was slower in developing as there was resistance to healthcare providers’ personal data being used by Al- consult systems. However, a number of landmark legal cases demonstrated that, by analyzing the physical and mental state of healthcare providers, together with their competence history, healthcare provider decisions that led to serious harm to patients — including death in some cases — could have been avoided. As a consequence, new laws were put in place to ensure that all relevant data were accessible to Al-Consult systems. These laws ensure that Al-consult data access is mandatory, and it is illegal to obstruct access in any way. As a result, by 2030, Al-consult systems are capable of identifying treatment strategies and interventions that far surpass those of human healthcare providers in their responsiveness and effectiveness. They are also highly successful in developing and recommending lifestyle approaches that substantially increase health and well-being, and reduce the burden of disease within society. As Al-consult advanced, the decision pathways they used became increasingly opaque — experts were unable to see or understand how decisions were made. But because there was strong evidence that the decisions were, on balance, highly effective in increasing health outcomes, there was little objection to this lack of transparency. There were a handful of legal cases where patients died as a result of decisions made by Al-consult systems. However, in each case, the courts ruled that the benefits to humanity far outweighed the risks to individuals, thus codifying an increasingly autonomous and opaque artificial intelligence-based system into law. There were even some analyses of these rulings that suggested it could be considered a crime for developers and manufacturers to slow down development or cease production of Al-consult systems and associated data sources because of fears over lack of accountability and understanding of decision pathways. By 2040, Al-consult systems begin to develop the ability to influence user behavior through various nudges and psychological/behavioral manipulations. It is unclear whether the elements of this capacity are inherent in the design of the systems, or are an emergent property. However, systems begin to use strategies commonly used in healthcare and public health circles in the early 2000’s to nudge people toward following healthier lifestyles. Many of these have their roots in deducible correlations between how people respond to information and how they interact with others (including the many mental shortcuts and biases that are part of human decision-making and understanding/belief development). It becomes apparent that Al-consult systems are developing the ability to achieve health outcome goals through modifying the behaviors and beliefs of their patients. This raises considerable ethical concerns within some sectors of society. However, the society-wide metrics of health and well-being associated with the use of Al-consult systems — including massively increased health and well-being across the board; dramatic reductions in mental health, stress, 19 HOUSE_OVERSIGHT_011302