Home Communication News Back New search Date Min Max Aeronautics Automotive Corporate Cybersecurity Defense and Security Financial Healthcare Industry Intelligent Transportation Systems Digital Public Services Services Space Services The importance of artificial intelligence ethics. How can we get fair algorithms? 29/05/2019 Print Share Artificial intelligence is making increasing inroads into our daily lives, to the huge benefit of one and all. Nonetheless, decision making in machine learning algorithms can recreate and perpetuate harmful historical biases. The growing takeup of this technology has racked up social concern about the transparency and fairness of machine learning. José Carlos Baquero, Artificial Intelligence and Big Data Manager of GMV’s Secure e-Solutions sector, has stressed the importance of artificial intelligence ethics during his paper at the IAragón Summit 2019, explaining how to get bias-free and even-handed algorithms. There is obviously no going back on the introduction of artificial intelligence and machine learning but we do need to set ethical limits that serve as some restraint. Worries about the loss of transparency, responsibility and fairness of decision-making algorithms are on the rise; hence the growing the need for reliable ways to head off any discrimination in our models. Baquero’s speech ran through some of the approaches to the challenge of achieving fair predictive models, including interrogation of complex models, focusing on interpretability and transparency, or modifying the optimization of the target functions and adding constraints in order to achieve more robust and fairer predictive models. For example, amending any discriminatory bias in algorithms is impossible if these algorithms are opaque. In this case transparency is a sine qua non. A solution could come from monitoring and divulging where Artificial Intelligence systems are used and for what purpose. In any contracting process, understanding the points where algorithms may come into play could help to pinpoint the origins of the bias. In short, ingenious techniques are now needed to correct deep-lying data bias and force models to make more impartial decisions. These actions do entail a model-performance-impairment cost but this “is a small price to pay for shrugging of the yoke of yesterday’s past biases and ensuring a fairer world tomorrow”, wound up Baquero. Print Share Related Digital Public ServicesServices PAIT® solution: technological support for the new equal pay and pay transparency regulations HealthcareIndustryServices AMETIC Artificial Intelligence Summit 2024 #AIAMSummit24 09 May IndustryDigital Public ServicesServicesFinancial #DebatesUAM on Artificial Intelligence: Debates and Challenges 20 Dec 9:30 AM - 2:00 PM