Facial recognition technology. Suggested searches. Academic integrity. It seems as though artificial intelligence (AI) is invading every facet of life as we know it. But to what extent will we see our daily lives automated in the years to come?
More importantly, who could be left behind?
A recent study published by researchers from Ochanomizu University and the University of Oxford found that 39 per cent of daily tasks will be automated within the next ten years. Within the next five years, 27 per cent will be automated.
From a gender perspective, these findings raise numerous questions.
A study published in nature human behaviour notes unpaid domestic work is largely performed by women in the United States and Europe. This consequently decreases their chances for employment and career progression. The uneven distribution of domestic work means the automation of domestic tasks will likely affect men and women differently.
The purpose of machines like dishwashers, laundry machines and robot vacuums is to make our lives more convenient. Their existence is proof humans already live highly automated lives.
Inmar Givoni, former director of engineering for Uber’s self-driving division says the integration of AI into our daily lives is nothing new.
“In recent months, there is a sense that AI is really coming to the front. The reality is that these systems have been in use across so many domains for the past 10 years,” Givoni said. “Every time you take a photo on your phone, there are several AI systems running.”
In recent months, AI has taken post-secondary institutions and much of the North American workforce by storm. ChatGPT, one of the biggest household names in the technology world, is a language model designed to interact with users in a “conversational way,” according to OpenAI.
Though the term “language model” might seem benign, ChatGPT has reportedly already overtaken a number of copywriting jobs, passed the bar exam, and been deployed as a crisis counsellor. Researchers have been largely divided on whether ChatGPT’s existence will revolutionize the way we do work or pose as a safety threat to humankind.
But questions about who these technologies benefit and the impact of AI’s tendency to reflect systemic biases is becoming an increasingly attractive field of study for researchers.
Lulu Shi, a University of Oxford sociologist, is a researcher at the Oxford Internet Institute. She’s currently working on a project, titled Domestic AI. The goal is to investigate the impact of AI on unpaid domestic work.
Shi said the study’s findings about AI sociology are more important than the predicted timeline of increased automation.
“The future… is very different depending on your background,” she said.
One of the study’s aims is to bring unpaid domestic work into the research debate. The study polled 65 AI experts from the UK and Japan, asking them to estimate the automatability of 17 housework and care work tasks.
Uniquely, the study’s authors applied a sociological approach, considering how the experts’ backgrounds couldinfluence the estimates. Shi’s researchers compared the estimates of the different gendered experts while acknowledging their unique backgrounds.
Shi said the paper’s main critique is that the predictions have been largely understood as objective truth.
“Even for tech experts, these are people and these people are living in a society. They’re shaped by their socio-economic background,” she explains.
A sociological approach to studies of technology is important, according to Shi, because humans will design the future and how technology functions within it—not tech.
A gendered approach to AI
Based on the study, male experts in the UK predicted a higher degree of automation in daily unpaid work than female experts from the same region. This finding aligns with supporting data, Shi said, reflecting that men are usually more optimistic about technology.
However, in Japan, the results were reversed. There, female experts predicted a higher degree of future automation than male experts, Shi said.
Using qualitative data, Shi and her team found that male experts in Japan argued that due to the high cost of automation, excessive levels of automation were unlikely to happen. On the contrary, female experts acknowledged the expense of these technologies, but said they still believed AI’s worth would be realised and used.
One possible explanation for female experts predicting a higher degree of automation despite the financial cost is the difference in social experiences for men and women.
In Japan in particular, an overwhelming majority of unpaid domestic work is done by women, the study reads. Shi noted a clear “gender imbalance” in domestic tasks. Consequently, there’s a greater incentive for Japanese women to want to automate this kind of work.
However, Shi added, if technology in Japan is primarily represented by men, then it is unlikely to be developed in the area of unpaid domestic work—despite the desire of many women.
Fiona Robinson, a political science professor at Carleton University, said the sociological questions this study addresses often go unanswered.
“We tend to have this sort of mindset that technology is something neutral that can be the answer to all of our problems,” Robinson said.
Robinson also said society’s underestimation of the embodied, emotional impact of unpaid domestic work is important to recognize.
For example, a person grocery shopping for their family often knows their family’s specific food preferences, the desired weekly meal plan and the differing schedules of family members. Robinson thinks this emotional investment element of a task could be missed out on if AI began buying groceries for a family, instead of a human.
Predictions only go so far
Michel Barbeau, a computer science professor at Carleton University, said the estimate that 40 per cent of unpaid domestic work will be automated within the next decade is a bold one. He said Shi and her team included a good diversity of participants in their study, but he disagreed with the study’s focus on predictions because of their uncertainty.
However, Barbeau said he too believes unpaid domestic work will increase in automation over the next five to 10 years.
“It’s more than probable, because we are there now,” Barbeau said.
Many of the tasks conducted on a daily basis are already automated. For example, grocery shopping has evolved into online shopping, Barbeau said. Barbeau emphasizes that AI is still limited and society tends to overestimate what AI is capable of—especially in terms of its emoting capabilities.
“We don’t yet have the sensors to capture this kind of emotion, it can only be simulated,” he said.
“Strong biases” in AI threaten women and ethnic minorities
AI systems are also human-made and therefore contain strong biases, Givoni said, referring to algorithms’ tendency to reflect harmful human biases, such as racism and sexism.
According to a 2021 study, when one image-generation algorithm is asked to autocomplete a cropped image of a man, it will autocomplete him wearing a suit 43 per cent of the time. When the same AI is used to autocomplete a picture of a woman, there’s a 53 per cent chance she’ll be clad in a low-cut top or bikini.
The study found women were much less likely to be generated in conventionally professional clothing, even in the case of images of prominent U.S. representative Alexandria Ocasio-Cortez.
“Because the algorithms of AI are so dependent on the data you fit them [with] and your perspective on what the world is […] you can unwittingly introduce a lot of biases into how these systems work,” she said.
Givoni says the most common type of biases found in AI are towards women and ethnic minorities.
“There are so many considerations that don’t get taken into account in technological development and scientific work,” she said.
Whenever possible, Givoni says employers should strive to diversify the staff working on this type of technology, as a way of overcoming biases.
Shi also says AI is often biased against certain groups of people, but if used responsibly, she hopes AI can complement humans’ lives like some pre-existing technologies have.
“It can be used to create more gender imbalance, but it can also be used in other ways,” Shi said. “Technology also is not always about time saving. You can also think about technology as complementary to what we do and [increasing] of quality.”
For Shi, the washing machine exemplifies a form of automation that increased the quality of our lives in the form of hygiene.
“It didn’t save the amount of time that we would [expect],” Shi said. “At the same time, our expectations for hygiene also increased. It’s not just changing the equation. The context also changes when you have new tools.”
Equal representation moving forward
Focusing on AI’s impact on unpaid domestic work is not only important, but also fairly unprecedented, Givoni said. Addressing technology from a gendered perspective is crucial in moving forward as a society.
Often, the overwhelming fear that “robots are replacing us” is a dominant theme in AI studies, Shi said. She’s advocating for further research into the unpaid labour branch of AI integration.
“[Further research] will also help us to understand that it’s really important to involve all sorts of people in the discourse, in the sense of having a democratic society and having both men and women equally represented in tech,” Shi said.
Featured graphic by by Sara Mizannojehdehi.
