Not So Fast — Humans Need To Work Together Before We Can Expect Them To Work With Machines

Tech giants Amazon
AMZN
and IBM
IBM
recently announced they were hitting pause on the development of facial recognition technology, once hailed as an important tool for law enforcement. The companies decided to halt their programs after recognizing — rightly — that facial recognition technology could further exacerbate racial disparities.

A 2019 study, conducted by the National Institute of Standards and Technology, found that facial recognition technology misidentifies Black faces up to 100 times more than white faces (Amazon and IBM’s technology were not included in the study). Current facial recognition technology may be operating to the best of its programming in assessing photographs and data to match patterns in people’s faces, but it fails to properly account for the cultural and societal biases present in the data.

This oversight created a fundamental flaw in facial recognition technology’s output. Other technologies, including self-driving cars and telemedicine, have struggled with similar issues. That’s what makes the field of study of human factors so crucial for engineers and users alike — especially today.

Human factors incorporates disciplines that aren’t typically studied by engineers — such as cognitive and social psychology, anthropology and sociology — by bringing them to the forefront of engineering design. This includes considerations about humans, such as our characteristics, abilities and propensities. As we increasingly collide with automated or autonomous systems — whether through self-checkout at the grocery store or with food delivery robots — it is abundantly clear that despite their benefits, there is room for improvement. Through cross-disciplinary collaboration with broader areas of engineering, human factors can play a major role in bridging the gap that exists between technology and how it actually interfaces with humans. To minimize the oversights mentioned earlier, and make sure humans are being served in the best possible way, emphasizing this collaboration will be key.

As a researcher who focuses on human trust and distrust in autonomous systems, I have seen how vital the human factors perspective is to improve these technologies. For example, colleagues in human factors and psychology have taught me how to design experiments involving humans, a skill that never came up in my mechanical engineering courses. I, in turn, have incorporated experimental techniques from my engineering background to ensure that the data we collect can be translated to designing autonomous systems that are responsive to the human. By blending our expertise, we are able to get closer to devising and designing systems and products that benefit the eventual users.

In academia, where we are each generally trained in individual disciplines, breaking out of our “lanes” is not easy, particularly when we are eager to get to work and discover a solution. But institutions like mine are tasked with creating a pipeline of engineers that understand the human side of their work and products, and how to engage with that side of the equation. As a professor preparing students for their careers in the private sector, interdisciplinary training and familiarity is of the utmost importance if companies are to develop technology that works for everyone. Students are not on campus solely to develop technical skills in their field of study, but to gain cross-cutting experience and collaborate, which requires learning how to communicate their discipline to people outside of it.

This is increasingly necessary and critical as there is a growing consensus that the benefits of incorporating human factors, and interdisciplinary collaboration in general, are too big to ignore. The National Science Foundation has recently created new programs, such as the Future of Work at the Human-Technology Frontier, to support research that encourages interdisciplinary teams and more explicitly considers humans in technology development. Similarly, the Defense Advanced Research Projects Agency, or DARPA — which has been driving innovation in the service of national security for six decades, most notably by supporting the development of the internet — also supports research that draws on several different disciplines to design technology that is appropriate for humans.

Many new frontiers of discovery have occurred at the intersection of disciplines; collaborations between biologists and engineers have given rise to innovation in everything from human prostheses to bio-inspired robots. Advancing the state of human interaction with automation, and algorithms that rely heavily on human data, requires engineers and computer scientists to take a similar interdisciplinary approach with social scientists. By marrying these perspectives and expertise, we not only get closer to eliminating troubling outcomes, as seen in facial recognition technology, but will produce more effective, useful — and fair — tools and technology for humans and our society as a whole.