We need to stop developing new artificial intelligence technologies just for the sake of innovation, which forces us to adapt our practices, habits and laws to accommodate the technology, according to a global team of experts. Instead, they advocate following human-centered AI design principles to create AI that precisely meets our needs.
Fifty experts from around the world have written research papers for a new book on how to make artificial intelligence more "human-centered", exploring the risks and missed opportunities of not using this approach, as well as practical ways to implement it.
These experts come from 12 countries including Canada, France, Italy, Japan, New Zealand and the United Kingdom, as well as 12 disciplines including computer science, education, law, management, political science and sociology.
Human-Centered Artificial Intelligence examines AI technologies in a variety of contexts, including agriculture, workplace settings, healthcare, criminal justice and higher education, and proposes more "human-centred" applicable measures, including approaches to regulatory sandboxes and interdisciplinary working frameworks.
What is human-centered artificial intelligence?
Artificial intelligence (AI) is permeating our lives at an ever-increasing rate, and some experts believe that relying solely on technology companies to develop and deploy this technology to truly enhance the human experience will be detrimental to humanity in the long run. This is where human-centered artificial intelligence comes in.
Shannon Vallor of the University of Edinburgh in Scotland, one of the world's foremost experts on human-centered AI, explains that human-centered AI refers to technologies that help humans thrive.
"Human-centered technologies are about integrating the entire technology ecosystem with human health and well-being. This contrasts with technologies that are designed to replace, compete with, or devalue humans, versus technologies that are designed to support, enhance, enrich, and enhance humans," she said.
She pointed to generative artificial intelligence, which has become popular in recent years, as an example of technology that is not human-centered - she believes that the creators of this technology just want to see how powerful they can make the system become, rather than to meet human needs.
"What we get is something we have to deal with, rather than something designed by us, to serve us, and to benefit us. It's not the technology we need," she explains. "Instead of adapting the technology to our needs, we adapt ourselves to the needs of the technology."
What’s wrong with artificial intelligence?
The contributors to Human-Centered Artificial Intelligence lay out their hopes for AI while also raising many concerns about its current trajectory.
Malwina Anna Wójcik from the University of Bologna and the University of Luxembourg in Italy pointed out the systemic bias in the current development of artificial intelligence. She noted that historically marginalized groups have not played a meaningful role in the design and development of AI technologies, leading to "the solidification of dominant power narratives."
She believes that a lack of data on minority groups, or that existing data is inaccurate, leads to discrimination. Furthermore, unequal availability of AI systems results in a widening power gap that prevents marginalized groups from entering the AI data loop and benefiting from these technologies.
Her solution is diversity in research and interdisciplinary collaborative projects at the intersection of computer science, ethics, law and social sciences. At the policy level, she suggested that international initiatives need to involve intercultural dialogue with non-Western traditions.
Meanwhile, Matt Malone of Canada's Thompson Rivers University explained how AI poses privacy challenges because few people truly understand how their data is being collected or used.
He explains: "These gaps in consent and knowledge result in permanent intrusions into the realm of privacy that might otherwise seek control. Privacy determines the extent to which we allow technology to enter the realm of human life and consciousness. But as these onslaughts fade, privacy will soon be redefined and rethought, and as AI captures more of our time, attention, and trust, it will continue to play a decisive role in drawing the line between humans and technology."
Malone believes that "privacy will change as people accept or reject artificial intelligence technology" and that even if technology brings greater equality, individuality is likely to be threatened.
Artificial Intelligence and Human Behavior
In addition to exploring the social impact, the contributors also investigate the behavioral impacts of current forms of AI use.
Oshri Bar-Gil of the Israel Institute of Behavioral Sciences conducted a research project to explore how using Google services changes the self and self-concept. He explains that when we use a platform, we create a data "self" and the platform then gets more data from how we use it and then uses the data and preferences we provide to improve its own performance.
"These efficient and beneficial recommendation engines have a hidden cost - their impact on us humans," he said. "They alter our thought processes, altering core aspects of our humanity such as intentionality, rationality and memory in both the digital realm and the real world, reducing our agency and autonomy."
Alistair Knott of Victoria University of Wellington in New Zealand, Tapabrata Chakraborti of the Alan Turing Institute of University College London, and Dino Pedreschi of the University of Pisa in Italy have also conducted research on the widespread application of artificial intelligence in social media. "While the AI systems used by social media platforms are in some senses human-centered, several aspects of their operation deserve closer examination," they explain.
The root of the problem is that artificial intelligence continuously learns from user behavior and improves its user model as users continue to use the platform. But users tend to click on items recommended to them by the recommendation system, which means that over time, the AI system is likely to narrow the user's interests. If users interact with biased content, they are more likely to be recommended that content, and if they continue to interact with that content, they may find themselves seeing more of it: "In short, there are reasons to worry that recommendation systems may push users toward extremist positions."
They propose a number of solutions to these problems, including increasing transparency from the companies that hold data on recommendation systems, allowing for more in-depth research and reporting on the impact these systems have on user attitudes toward harmful content.
How does human-centered artificial intelligence work in reality?
Pierre Larouche of the University of Montreal in Canada believes that treating artificial intelligence as an "independent legal and regulatory object" and assuming that "there is currently no law applicable to artificial intelligence" makes some policymakers feel that this is an unachievable task.
He explained: "Because artificial intelligence is considered a new technological development, it can be presumed that no laws related to it currently exist. Likewise, although there are few or even no specific rules about artificial intelligence, there is no shortage of laws that can apply to artificial intelligence due to its inherent connection with social and economic relations."
Larouche believes that the challenge is not to create new laws, but to determine how to extend existing laws and apply them to artificial intelligence: "Allowing the debate to be framed as an open-ended ethical discussion on a blank legal page can be counterproductive to policymaking, because it opens the door to a variety of delaying tactics designed to prolong discussions indefinitely, while technology continues to advance rapidly."
Benjamin Prud'homme, deputy director for policy, social and global affairs at the Quebec Institute for Artificial Intelligence (Mila), one of the largest academic groups dedicated to artificial intelligence, also called on policymakers to have confidence.
He explained: "My first suggestion, or my first hope, is that we start moving away from the binary opposition between innovation and regulation - that we acknowledge that it's okay to stifle innovation if it's irresponsible. What I would tell policymakers is to be more careful about their ability to regulate artificial intelligence. Confidence; yes, the technology is new, but it would be inaccurate to say that they have not dealt with the challenges associated with innovation (successfully) in the past. Many people in the AI governance community are afraid of not getting it right from the beginning. You know, my experience in policy circles tells me that we probably won't get it right from the beginning. Get it right from the start. No one has a magic wand. So here's what I say to policymakers: Do your best to get all parties involved - including marginalized communities and end users - in the discussion. Paralyzed by voices who argue that governments cannot regulate AI without stifling innovation, the EU can set an example in this regard as the ambitious Artificial Intelligence Act, the first systemic law on artificial intelligence, is due to receive final approval in the coming months.
Reference: "Human-centered artificial intelligence - a multidisciplinary perspective for policymakers, auditors and users", March 21, 2024.
doi:10.1201/9781003320791
Compiled from:ScitechDaily