“What on earth even IS humanity-centered AI?”… is the question we got asked a million times when we launched DEUS.
The short answer: artificial intelligence is traditionally considered from a technology-centered perspective (what technology is capable of and what data you have available), but we approach it from an understanding of humans, their needs and their environment.
The long answer requires a few more words, as there are many different interpretations of what it means, what its foundational principles are… and why it matters.
Not only are there different definitions of what human-centered and humanity-centered entail, there are also different ways in which artificial intelligence is defined. So when you combine these concepts all together, naturally you get… a lot of questions.
Here’s an attempt at answering those — through an exploration of the different concepts that underly human(ity)-centered AI.
Human-centered design and artificial intelligence have both gotten quite some hype over the past few years. If anything, human-centered artificial intelligence sounds like a great candidate for buzzword bingo. But, setting aside the buzz and over-usage of the terminologies, words have inherent meanings — even though those meanings can fade. So to start with, let’s take a look at the separate definitions of ‘human-centered’ and ‘artificial intelligence’:
Human-centered: used to describe computers, technology, systems, etc. that are designed to work in ways that people can easily understand and learn.
Artificial intelligence: the theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.
When you combine the two, you end up with a definition that’s somewhere in this direction:
To design and develop artificial intelligence applications in a way that people can easily understand them and learn.
But the combination of these two definitions doesn’t fully cover the extent of the field of human-centered artificial intelligence. There are a number of principles that should be embodied in this definition:
In short: human-centered AI is AI that is designed by and for-, has impact on-, and works in conjunction with humans.
Typically, when talking about human-centered AI, the implicit focus is on the micro level: the people directly impacted by the system, such as the end-users, employees or an organisation.
The humanity-centered aspect is about the macro level: the implications on the world when we create and introduce a certain system to it. This holds for any system or technology, not only artificial intelligence.
Humanity-centered means to focus our efforts in creating AI applications on both the possibilities of AI to create a positive impact on our world and society, as well as to thoughtfully consider and mitigate potentially harmful implications.
Focusing on humanity-centered artificial intelligence means that we pose the following questions:
So this brings us to our expanded definition: human(ity)-centered AI is designed by and for-, has impact on-, and works in conjunction with humans and their environment.
Combining human-centered artificial intelligence and humanity-centered artificial intelligence you get — easy guess — human(ity)-centered artificial intelligence.
Though this hybrid terminology is not widely adopted, we see it as a necessity when developing artificially intelligent systems: focusing on both the micro and macro levels of humans and humanity, and to be inclusive, responsible, sustainable and ethical on both levels, by considering both the direct and indirect consequences of what we create and put into the world.
In the end, we hope that at some point, the term human(ity)-centered will disappear altogether, simply because it has become standard practice.