Researchers Develop New Principle to Make AI Take Ethical Decisions

AI is likely to go for unethical decisions as it is profitable. Researchers have developed 'Unethical optimization principle' in collaboration with many universities

Artificial Intelligence (AI), in its present form, cannot distinguish between what's ethical and what's not. There's no sense of ethics in AI. But, researchers from Warwick University and The Imperial College, London, EPFL research institute in Switzerland's Lausanne, and Sciteb Ltd have found a mathematical method to manage the unethical biases in AI systems when dealing with regulators and businesses.

Unethical decisions can cost a lot of companies while damaging their commerce. Researchers call it the "unethical optimization principle." In rethinking the way AI operates, the principle is developed so that unethical outcomes are rejected by the optimization process, according to Professor Robert MacKay of the Mathematics Institute of the University of Warwick. The research paper is published in The Royal Society Open Science Journal.

AI Opts for Unethical Decisions for Profitability

Artificial Intelligence
Representational Picture Pixabay

"This is a really important paper," says Professor Wendy Hall of the University of Southampton, famed for her work on the potential benefits and problems of AI. She further says relying on AI systems to act ethically is not possible, because even though their objectives seem ethically neutral. Under some conditions, the "AI system will disproportionately find unethical solutions unless it is carefully designed to avoid them."

In some cases, AI can choose from many potential strategies, among which some may be discriminatory or may misuse the customer data leading at penalties for the company.

The Principle

Dr Heather Battey, the paper's co-author from the Department of Mathematics at Imperial College, said, "Our work shows that certain types of commercial artificial intelligence systems can significantly amplify the risk of choosing unethical strategies relative to a less sophisticated system that would pick a strategy arbitrarily."

The researchers discovered that even in very few unethical strategies among a pool of possibilities, AI systems are likely to choose unethical ones, simply because they are profitable. The New 'Unethical Optimization Principle' provides a simple formula for this. It helps regulators and companies in finding problematic strategies that lay hidden among the pool of potential strategies and suggest how the AI search algorithm shall be modified in avoiding them.

Related topics : Artificial intelligence