Ethics for Intelligent Machines.
Engineers design amazing hardware and software these days, from small connected devices and cloud services to increasingly autonomous vehicles, but the discussion of what ethics and values should be built into this technology is lagging behind technical advances.
That’s what technology thinker and writer John C. Havens tries to address. He is heading a new interdisciplinary group launched by the global engineering association IEEE that aims to close the gap with practical recommendations for engineers and programmers. Havens described the challenges and first results in an interview with NEXT at SXSW, the annual tech gathering in Austin, Texas.
Human Wellbeing is a subject too
The Global Initiative for Ethical Considerations in Artificial Intelligence and Autonomous Systems was launched in April 2016 and brings together 150 experts from various disciplines from around the world. They are broken up into 15 committees tackling topics from “Embedding Values into Autonomous Intelligent Systems” to economics and human wellbeing.
“We don’t formulate standards, but consider ourselves an innovation machine that tries to come up with recommendations for what can become new standards. The fundamental question is this: How can you design and build ethics into machines?” Havens explained. “It’s important that engineers and developers speak at eye level with sociologists, ethicists and philosophers. You can’t fault them if they build things without getting this input early on.”
The relatively uncharted territory of machine ethics is not only about risk and compliance, he said, but values. “It requires more thorough due diligence before any design goes to the engineers who are just trying to fulfill requests for features and functions.”
' It’s important that engineers and developers speak at eye level with sociologists, ethicists and philosophers.” '
John C. Havens, technology thinker and writer
A foundation for programming
Havens counts human wellbeing as one of those key guiding principles that should complement metrics such as economic growth or revenue numbers. In the case of autonomous vehicles, that approach would call for a debate on the repercussions automation will have on the labor force, such as truckers, or the trade-off between making mobility more accessible to wider parts of the population through ride sharing and its affordability.
“If you only look at GDP as a success metric, you leave things out that make a big difference on peoples’ happiness such as anxiety or depression from unemployment. Even the UN and the OECD recognize we need to look at those metrics.” The challenge, added, is to break the lofty goals down to details with which a programmer can work. The broad social issues Havens addresses here are directly related to the disruption that characterizes the digital world as a whole. The relevance of these questions is highlighted by the fact that these issues are increasingly being brought forward today.
The IEEE commission published a first report past December and opened it up for global comments. The 100-page document looked at eight specific topic areas and garnered more than 90 pages of input from around the world. The commission is currently working on the second, more comprehensive version of the report with a total of 13 sections, due out by October.
Values by Design
“Ideally you want the people building intelligent systems of tomorrow to use a principle I call Values by Design — similar to what companies already do with privacy by design,” explained Havens who has written two books on the subject. That process has to start with discussions like the one facilitated by IEEE. “We need to codify our own values first to best program how artificial assistants, companions, and algorithms will help us,” said Havens. “How will machines know what we value if we don’t know ourselves?”
' If you only look at GDP as a success metric, you leave things out that make a big difference on peoples’ happiness. '
John C. Havens