알고리즘을 통해 돌봄과 연민을 위한 시간을 증가시킬수 있을까?

By: Marina Durano (Adviser for Care Economy and Partnership Engagement, UNI Global Union; CTMS Advisory Board Member)

Healthcare professionals have gained significant experience with technological advances in their fields through medical devices, pharmaceutical products, health data management, among many uses. Nowadays, they have been trying to understand the implications of artificial intelligence and use of algorithms in various health and care settings. The American Nurses Association, for example, released a position statement in 2022 that provided its membership with “ethical guidance” on the use of artificial intelligence (AI) as they practice their profession. A year earlier, the World Health Organization, produced a guidance document on “Ethics and Governance of Artificial Intelligence for Health” that promoted six (6) ethical principles on the use of AI for health but also highlighted ten (10) sources of risks. While we continue to be far and away from a machine that approximates the general intelligence of a human being, the applications of machine learning–”concerned with building systems that improve their performance on a task when given examples of ideal performance on the task, or improve their performance with repeated experience on the task”–are now found in diagnosis and treatment, patient engagement and adherence, and administration and management.

I recognize that these guidelines are meant to be applied to the health care sector but I wonder if there are lessons from these debates about artificial intelligence that we can begin to take into account as we build a care economy. A group of researchers recently used a three (3) round Delphi survey of experts in artificial intelligence from the UK and Japan to forecast the automation of seventeen (17) domestic and care work tasks using typology from the UK TUS and the Japan STULA. The specific question requested the experts to predict the percentage share of total time that currently goes into performing the task that could be automated within five (5) years and ten (10) years.

When averaging across the seventeen tasks, AI experts from both countries forecasted a 27.5% and 39% reduction of time spent on domestic and care work due to automation in the next 5 and 10 years, respectively. The AI experts who responded to the survey think that domestic tasks are easier to automate than care tasks. Grocery shopping will see a 59% reduction of time, for example, while physical child care will only see a 21% reduction of time within 10 years due to automation, according to the experts. These results are consistent with an earlier study that estimated the probability that an occupation will be computerised leading to job loss. This study on computerisation and job loss identified barriers to computerisation that engineers continue to struggle over, namely: perception and manipulation tasks, creative intelligence tasks, and social intelligence tasks. Care entails many of these types of tasks.

Let us take for granted for now that the predictions of reductions in time will come to fruition within the next 5 to 10 years. We might then want to ask whose time will be freed up and how will that time be used. In the ANA statement, the nurses are quite clear. There is an expectation that their routinary tasks will be taken over, freeing up some of their time. With patient-centred care as a motivation, nurses are asked to divert efficiency gains into “caring for patients’ physical, emotional, and cognitive needs.”

AI is particularly useful in health care with respect to assisting or taking over mechanical tasks such as feeding, attending to patient hygiene, fetching, and dispensing or titrating medication, as well as conducting diagnostics. In such cases, nurses must safeguard patients so that efficiency gains are directed at activities that support or enhance caring for patients’ physical, emotional, and cognitive needs. Nurses should also consider how technology shapes the nurse-patient relationship with respect to patient expectations and perceptions of caring. (ANA Position Statement 2022).

ANA is referring very clearly to improving the quality of care when efficiency gains are achieved. In other words, when nurses have more time to spend with their patients, then they can better understand and attend to the totality of their needs. One of the classic studies that looked into the relationship between nurse staffing and preventing adverse patient outcomes highlighted the importance of staffing adequacy, which is also a factor in determining worker retention and job satisfaction. In another study, involving a longitudinal cohort in the general ward, the result showed that “[l]ower [registered nurse] RN staffing and higher levels of admissions per [registered nurse] RN are associated with increased risk of death during an admission to hospital.” These results were confirmed by this study that covered a variety of hospital departmental units, which concluded that “staffing of RNs below target levels was associated with increased mortality.”

These studies show the dangers when instead of increasing the amount of time spent with a patient, hopefully reflecting improved quality of care, efficiency gains will instead lead to an increase in the number of patients a nurse will be responsible for. Within a facility or a medical complex, it will be possible to negotiate with management how these efficiency gains could be used. In the US, a federal law to regulate staffing ratios is being introduced, while only sixteen (16) states have laws or regulations about it.

Indeed, there are interesting questions here about the elasticities of substitution, between low-skilled and high-skilled care workers as well as in the allocation of tasks between routine and non-routinary tasks. Another way to look at it is whether efficiency gains can be translated into improved wages and incomes for health care workers, which has the advantage of contributing to a decline in inequality across the skills ladder instead of widening the skills gap that has been associated with technological change. Will wage improvements as a result of the efficiency gains be an indicator of improved recognition of the value of care work?

The last point about the value of care work is certainly relevant in trying to understand what the impact of these new technologies might be in the home for two groups of people: the primary (probably unpaid) care caregiver and domestic workers and other home-based care workers. Increased female labor supply has been associated with lower child care responsibilities, among other reasons. Automation of child care is not likely to be a source of time use reduction here given the survey results above. We can, however, ask about the automation of domestic tasks. In the UK, for example, a decrease in the relative price of home appliances to the consumer price index was found to be a statistically significant contributor to increasing female labor supply. These figures will vary depending on the country in question but it is indicative of some potential. Unlike nurses in employment arrangements that can negotiate how efficiency improvements can be used, unpaid care workers at home will have to figure out (possibly with other members of the household) how they will use their newfound “free” time.

The AI expert survey also revealed an interesting aspect of AI innovation and that is the perspective of the AI experts themselves and how they view domestic and care tasks. Male AI experts from Japan tended to be pessimistic about innovations for domestic and care tasks compared to the female AI experts in Japan. The authors found that when it came to domestic and care tasks, experts tended to look beyond technical feasibility when forecasting, despite instructions to constrain themselves.

Underlying the country difference in our sample we also detected a strong gender dimension: our UK male experts were notably optimistic about the prospects of domestic automation, while our Japanese male experts were notably pessimistic, bucking usual male optimism about technology. As a possible interpretation for the Japanese male experts’ relative lack of enthusiasm, we pointed to the politics of the average Japanese household, where domestic work remains very much a woman’s occupation. Previous studies have treated all experts as interchangeable and have not disclosed the experts’ backgrounds. (Lehdonvirta, Shi, Hertog, Nagase, and Ohta 2023)

Technological developments in the care sector have been moving at a fast pace, despite multiple engineering challenges, but the engineers themselves may be unmotivated to develop new technologies that reduce time allocated to care tasks. Of course, the question of accessibility and affordability must come into play when forecasting potential impact of new technologies. The choice on how to use freed up time is important for understanding how much AI can contribute to well-being outcomes.