By AI Trends Staff
Efforts to further AI transparency and fairness got a boost recently with the naming of Prof. Judea Pearl of UCLA as the World Leader of 2020 by the AI World Society, a joint effort with the Boston Global Forum that calls for AI to be developed and deployed in ways that benefit all mankind.
Judea Pearl, Professor of Computer Science and Director, Cognition Systems Lab at UCLA
In presenting the award to Prof. Pearl, former Gov. Michael Dukakis, chairman of the institute bearing his name, stated, “I am inspired by your watershed work in establishing cause-and-effect relationships as a statistical and mathematical concept, most especially as we strive to more completely understand the rapidly-evolving impact of AI and machine learning on society.”
An offshoot of the Boston Global Forum, the Michael Dukakis Institute for Leadership and Innovation was born in 2015 with the mission of generating ideas, creating solutions and deploying initiatives to solve global issues, especially focused on cybersecurity and AI.
Prof. Pearl is the author of the recent, “The Book of Why: The New Science of Cause and Effect,” published in 2018, a study of cause and effect that helps answer difficult questions such as whether a drug cured an illness. Dukakis stated that the book “provides us with the new tools needed to navigate the uncharted waters of causality for students of statistics, economics, social sciences, mathematics and most urgently today, epidemiology.”
Prof. Pearl will serve as a mentor in the AIWS Innovation Network programs in support of those goals.
Causal Reasoning Can Guide AI Algorithms Towards Fairness
How causal reasoning applies to learning algorithms is the subject of a recent paper, “Using Causal Reasoning to Guide Algorithms Toward a Fairer World,” by Ilya Shpitser, Associate Professor of Computer Science at Johns Hopkins University, and Daniel Malinsky, Researcher, Johns Hopkins University. The paper is available at The Ethical Machine.
Learning algorithms find patterns in data they are given. However, in the processes by which the data is collected, relevant variables are defined and hypotheses are formulated that may depend on structural unfairness found in society, the paper suggests.
“Algorithms based on such data could introduce or perpetuate a variety of discriminatory biases, thereby maintaining a cycle of injustice,” the authors state. “The community within statistics and machine learning that works on issues of fairness in data analysis have taken a variety of approaches to defining fairness formally, with the aim of ultimately ensuring that learning algorithms are fair.”
The paper poses some tough questions. For instance, “Since, unsurprisingly, learning algorithms that use unfair data can lead to biased or unfair conclusions, two questions immediately suggest themselves. First, what does it mean for a world and data that comes from this world to be fair? And second, if data is indeed unfair, what adjustments must be made to learning algorithms that use this data as input to produce fairer outputs?”
Cause and effect is a challenging area of statistics; correlation does not imply causation, the experts say. Teasing out causality often involved obtaining data in a carefully controlled way. An early example is the work done by James Lindt for the Royal Navy, when scurvy among sailors was a health crisis. Lindt organized what later came to be viewed as one of the first instances of a clinical trial. He arranged 12 sailors into six pairs, and gave each pair one of six scurvy treatments thought at the time to be effective. Of the treatments, only citrus was effective. That led to citrus products being issued on all Royal Navy ships.
Whether fairness can be defined by computer scientists and engineers is an open question. “Issues of fairness and justice have occupied the ethical, legal, and political literature for centuries. While many general principles are known, such as fairness-as-proportionality, just compensation, and social equality, general definitions have proven elusive,” the paper states.
Moreover, “Indeed, a general definition may not be possible since notions of fairness are ultimately rooted in either ethical principle or ethical intuition, and both principles and intuitions may conflict.”
Mediation analysis is one approach to making algorithms more fair. Needless to say, the work is continuing.
When Big Tech Power Crashes into Fairness
Issues of the power of US big tech companies over personal, private data versus interest in ensuring AI systems are fair, can crash into each other. This is happening in the context of the Global Partnership on AI announced by Canada and France in late 2018, with efforts to extend to the Group of Seven western economies happening since then.
The US has been a holdout. The White House has characterized the effort as unnecessary bureaucracy that threatens to dampen AI development by being overly cautious, according to an account in Wired.
Cédric O, the digital affairs minister of France, raised the question of the Global Partnership in Washington late last year with US chief technology officer Michael Kratsios. Later in an interview, O stated, “There is common consensus but for one country.”
Cédric O, digital minister of France
O fears that without international coordination, unethical uses of AI could flourish. He uses as an example how China has used facial recognition and other technologies to strengthen its authoritarian security apparatus. “If you don’t want a Chinese model in western countries, for instance to use AI to control your population, then you need to set up some rules that must be common,” O stated.
Lynne Parker, the US deputy chief technology officers, says the US worries the group would be too restrictive. “Our concerns are that the group could be too heavy-handed,” she stated. “We believe it’s unethical to hamper and squash down the development of AI technology to the point where you don’t want to use it.”
She noted the US has joined with 40 countries in an effort by the Organization for Economic Cooperation and Development to advise on policy and endorse a set of AI principles.