Nov 15, 2024

Is Ethical Automation Possible?

Automation has been the watchword of our technology culture for the past so many years. The McKinsey Global Institute estimates that up to half of the total “work” that the economy now performs could be fully automated in the near future.

And despite the popular assumption that most automation will occur in jobs that involve manual labor, their research indicates that a large part of the “knowledge sector” is ripe for automation as well. This is particularly true for so-called “low skill” knowledge workers, such as receptionists, data entry specialists, cashiers, and personal assistants, among others.

Ethical Automation and Workers

It’s necessary now, and not later, to talk about the ethical implications of the disruption that this new wave of automation will create.

It’s clear to us that one of the challenges of dealing with social policy as a private company is that our immediate interests, as a business, often fail to align with the greater interests of society, the economy, or the individuals working in any one company.

Private companies are among the most impactful drivers of real, meaningful social policy and change, simply because they occupy so much of a modern worker’s productive life, and are the vessel in which so many people travel through life. In the United States, for example, private companies intermediate the worker’s access to basic needs, such as healthcare, ongoing education, social relationships, and more.

And yet it is in the United States particularly that businesses tend toward the least socially responsible policies. The more control over worker freedom, the more restrictive they tend to be. The more power over individuals’ lives companies have, the less responsibility they seem to be willing to accept.

Therefore it seems clear that an ethical approach to automation means more than simply creating safeguards for customer privacy, preventing systemic racial bias, and adopting the toughest safety standards that can be reasonably sustained. Those measures are important, and they’re things we’re going to discuss. But right now what concerns us as a small and growing company, is our responsibility to our own workforce, and by extension, the responsibility that every one of our business customers have to theirs.

Replacing Work, Not Jobs

In order to talk about responsibility, we need to have a good perspective on the current labor and automation situation.

The labor sector has been under pressure from automation for centuries. Once it was the steam engine replacing man and animal power, first in mining, and then across transportation and other industries. Today it is the Generative Ai that threatens massive displacement of jobs.

You might assume that the history of automation has always been the story of jobs “disappearing” as machines take over the work of human beings. That makes intuitive sense, particularly when our individual career experiences are partly defined by how much of our labor is being handed off to automated processes and machines. Few workers have gone their whole lives without significant aspects of their chosen work becoming automated.

To a great extent, that picture of automation was true a century ago, or even 50 years ago. At one time, now modern economies used slave, prisoner, and child labor to supply a great part of the labor necessary to industrialize themselves. Automation in that realm has had inarguably positive social implications.

Today, much of the corporate narrative of automation rests on the assumption that this kind of automation is as muc the future as it is the past.

That sense is only reinforced by the constant hyping of futuristic advances that promise to decrease the manual work of countless jobs in the workforce. It feels like our bosses are always trying to replace us with machines and automation, to varying levels of success.

“The Displacement Effect”

This is what economists Daren Acemoglu and Pascual Restrepo call the displacement effect, in their 2019 paper: Automation and New Tasks: How Technology Displaces and Reinstates Labor. The pair describe how business management tends to try to replace workers with machinery or software, hoping to lower costs and become more profitable.

Over time, as the demand for a certain type of labor increases, the need for workers to satiate that demand also increases, pushing companies to increase production and productivity. If the only way to increase short term productivity is to hire more workers, then pretty soon, more workers are harder to find, and will need to be paid more. Once labor becomes too hard to find, and thus too expensive, automation becomes more attractive as an investment.

A famous historical example is Ford Motors, which found that as demand for its Model T automobile increased, it was soon unable to find workers to satisfy that demand, owing to the tedious and difficult working conditions on its assembly line. Ford found it needed to hire up to four workers in order to retain one worker over a short period of time.

This led to the famous “$5 day,” which Henry Ford instituted in 1914, in order to retain workers. This move proved a historic inflection point, and worker rights and pay exploded over the next 70 years, leading to the longest period of middle class growth in history.

But vastly increasing worker pay is not always feasible on its own. That’s why this tendency is historically counterbalanced by what the authors call the productivity effect, which is the tendency for higher productivity to create new demands on workers. Essentially: a worker being paid $5 a day is expected to do more than a worker being paid $1.

Productivity Effect

The productivity effect states, in simple terms, that as demand increases, workers tend to become more productive. Because labor productivity is growing, therefore the market for that labor will grow also.

While one might expect that higher productivity would mean fewer workers, this is typically not the case. The productivity effect is counterintuitive, but it can make sense if considered in its historical context.

Consider the following real historical example: gardening. Stick with us because this is interesting!

In the 17th century, architects like Sir John Vanbrugh and Nicholas Hawksmoor originated the great English tradition of nature gardens, with their pleasing designs contributing to the popularity of public parks, and making a crisply manicured lawn and a set of perfect hedges the absolute bees-knees among the English aristocracy.

Gardening in the age of essentially zero automation was very hard work indeed. Every single job you could imagine had to be done by hand, with the greatest of care. Hedges had to be trimmed, lawns dug and then planted, then flattened and shaved precisely. Ponds had to be filled by hand, and endless mountains of sod and gravel had to be moved.

Not only all that, but the very tools and materials used in these pre-industrial gardens had to be made on-site, from the spades to the axes, and from the fertilizer to the cement and paving stones. If horses were needed to do the work, then a stable and workers to look after those horses were also needed. Every element of a garden had to be laboriously made and maintained. It took a virtual army of workers to maintain a stately grounds with a nature garden, and that was the point: it was a pursuit of the super-rich only.

Productivity Gains via Automation

That is, until industrialization made it so that the spades and axes could be bought in, along with the fertilizer and the stones. Pretty soon, horse-drawn carts were replaced with motor tractors, and hand-operated, self-propelled mowers. Today every job in gardening is heavily automated. We just don’t think of all those innovations as automation, even though they eliminate the vast majority of the labor involved.

That is the productivity effect in action. As demand increases, worker pay rises. But as pay rises, workers tend to become more able to do yet more work, via either automation or other technological advances.

Economists refer to this as a change in the “task content of production,” meaning that a particular task goes from requiring a large investment of labor to requiring almost none. Instead of one worker spending many hours crushing and mixing custom, homemade concrete, the worker buys the concrete in a bag and adds water. Instead of spending many days carrying water in buckets from a well, the gardener uses a pump, or turns on a tap.

400 years later, a stately garden can be tended to by a mere handful of workers, on a mostly part-time basis. The number of workers needed to do and maintain the work has reduced such that individual workers now spend only part of their time on one garden, spreading out among several clients to fill their days.

That is the productivity effect at work. The productivity effect decreases the need for labor, but does it eliminate jobs? Well, ask yourself: are there fewer gardeners in the world today than in 1600?

No, there are more.

Reinstatement Effect

Here the “reinstatement effect” comes into play.

Workers of previous generations rioted at the implementation of automation in manufacturing and other work, out of fear that it would permanently displace workers. We now know that what tends to happen is the opposite. The increased productivity in one area creates new work in other areas.

Today, one worker can work on many gardens, and so we have more gardens overall than we did before. It is no longer something reserved for the super-rich. It has become accessible to almost everyone.

The reinstatement effect balances both the productivity effect and the displacement effect. While workers are no longer needed for many mundane physical tasks, the increased individual productivity increases, and the accessibility of goods and services increases as well, while the overall relative cost of those goods and services goes down.

Today a gardener typically makes much more money than one who was working in a stately garden two centuries ago. And yet, the gardener’s job is easier, and the overall cost of having a garden is much lower. Truly, this is a scenario in which everyone wins (except perhaps for the weeds).

The Composition Effect

One things that changes drastically is the type of work people are expected to do. A gardener of 1800 would never have been expected to do their own bookkeeping, or telephone sales, or marketing, or many other elements of the job that are now common. Today, gardeners do all of those things themselves, or they contract with temporary and part-time agencies who manage many clients at once.

This is the “composition effect,” the tendency for the type of tasks in a worker’s job to change over time to reflect a wider range of activities. Overall, workers are in a better situation than they were centuries ago, even though they are expected to do many more tasks.

Breaking The Cycle

All this sounds great, but it’s not an ironclad law that automation and increased productivity always produce huge gains for workers or businesses. Sometimes automation makes things worse. a

Acemoglu and Restrepo point out several challenges that come with increased automation; particularly the kind using automated recognition systems, such as chatbots. Here, they point out that the gains from automation tend to be relatively small in comparison with the quality and productivity improvement that can be expected.

Worse still, if companies pursue automation strategies that decrease the responsibilities of their workers, such as by simplifying and eliminating tasks without assigning any new responsibilities, then the composition effect cannot take place, and the cycle of mutual improvement in productivity and work conditions and pay is broken.

Pay may stop rising, or even decrease. Workers may be eliminated without new jobs being created. Not only this but as automation continues to displace workers, the labor share of economic growth continues to shrink.

This means that over time, there is less upward pressure on wages, even as productivity continues to grow year after year. Particularly when automation allows businesses to actually lower the quality of their goods or services to save costs, such as by replacing a human being with a phone tree, or producing a lower-quality product using fully automated tooling, this has a negative impact both on workers, and the state of the wider economy.

Could we find ourselves in a future in which we can routinely expect the quality of everything to go down, while our wages continue to fall? Even worse, could this falling quality of production match up with ever-slowing wage increases to mean that we can only afford lower-quality products anyway, leading to even lower-quality products, and on and on forever?

Shoepocalypse?

Sci-fi writer Douglas Adams imagined such a scenario in his Hitchhiker’s Guide to the Galaxy, in which a civilization is torn asunder by the ever-worsening quality of its high-street shoe shops, leading eventually to the collapse of civilization, owing to the need for the majority of the labor force to be at work replacing the incredibly low-quality shoes that were necessary to make up for the lower pay they were receiving, owing to the reduction in the quality of shoes, etc.

It’s a comic scenario, but underlying it is a real-world frustration: some things really do seem to just get into a cycle like that.

We discussed this in a previous post on “Enshitification Theory.” Read it now.

But, it’s not an essential truth that automation always leads to this “race to the bottom” scenario. Cheaper goods can, for example, free up money that workers can then use to invest in higher-quality goods of a different kind.

Say for example that you don’t care whether your jeans are ripped or faded, but you are an avid hiker. If that’s the case, then you can put off buying new jeans, or buy cheaper machine-made ones, and use the money you saved to buy really good and expensive, hand-made hiking boots. In this way, the economy has produced a positive outcome for you, since you’re getting to choose between things that you find worthwhile and things that don’t matter to you.

The above leads us to the natural outcome that some of the share of labor in the economy will be lost to machine-made jeans, while new demand for labor will arise in the area of hand-made hiking boots. The displacement and reinstatement effects have run their course, and the composition effect takes over.

Ethical Automation and Labor

In a complex and modern economy, there is also a share of consumers who want handmade jeans and cheap hiking boots, and so in reality, all four industries co-exist: handmade jeans and hiking boots, and machine-made jeans and hiking boots. It falls to consumers to choose which things they’d rather spend their money on.

Such a happy medium is never perfectly maintained, but it is the situation that is most commonly imagined by those who defend both automation and increasing globalization. The theory is that choice will increase, and in this way, the “price” of lower-quality products is only one you will ever pay voluntarily. Good quality products will always be available, so in fact, things aren’t “getting worse,” at all. They’re just getting more diverse.

Ethical automation, we feel, has to be fundamentally based on an understanding of everything we’ve noted above. It also needs to take into account that even when the virtuous cycle of displacement, reinstatement, and composition change are occurring, there is always room for improvement. It’s also important to accept that, with apologies to Adam Smith, the unseen hand of the market might work this way, but there are hands that we can see, constantly getting in the way.

Free Trade and Financialization

Acemoglu notes that the utopian vision of globalism depends on the idea of fair competition and free markets. The problem is that the global marketplace is full of players who neither compete fairly nor participate entirely freely. There are always countries that have things that others want to take by force, rather than paying, and there are always governments that would like to act anti-competitively, by propping up their own favored industries at the cost of their trading partners and competitors.

Whether it’s by allowing or participating in (either passively or actively) an illicit human trafficking and slavery industry, or it’s by placing trade tarrifs on goods from competitor nations, governments and companies love to get their thumbs on the economic scales whenever possible. In a highly financialized world where short-term incentives seem to rule more and more, there is ever ever-greater temptation to contribute to exploitative practices and policies for quick wins.

As the finance industry has become such a large part of developed economies, the alacrity with which private companies can jump in and out of business arrangements with too-friendly governments and not-so-ethical business partners is a blessing and a curse.

The ability to change suppliers and reorient supply chains quickly can lead to greater potential accountability to consumers and regulators, since changes can be effected faster if ethical conflicts arise. But shady dealings can also be done with much more ease and deniability in a more complex and financialized world.

As companies continue to automate more labor over time, there will always be more opportunities for corner-cutting, when it comes to either a firm’s reputation or the letter of the law.

Next Time: Ethical Automation in Data, Services, and Advertising

In a follow-up post, we’ll dive deeper into other aspects of ethics in automation, such as the gathering and use of personal and demographic information, the use of automation in services such as customer service, and the role automation plays and will continue to play in advertising.

While these topics are absolutely no less important than the impact of automation on labor, they are effects that take place largely “outside” the perspective of a single organization or company. As such, they deserve to be considered deeply and separately from the discussion of the impact of automation on work and the workplace.

Our Proposals for “Ethical” Automation

Defining what ethical automation is, might be trickier than describing what we think it isn’t.

The ethics of any particular course of action depends on more than just economics, politics, morality, or technological feasibility. Most of all, it depends on the will of individuals and groups to attempt to live by the ethical and moral standards they have set for themselves. Its meaning will tend to vary depending on context.

Based on all we’ve talked about, it seems appropriate then, for us to name a handful of postulates that we think any ethical framework around the adoption of automation, either by a government or a private organization, should have:

Companies should not rely on automation that tends to decrease quality to customers

“Innovations” such as the phone tree or the dead-end chatbot are one kind of automation that can do this. Another is switching to inferior processes simply because they are easier and cheaper to automate, such as using high-fructose corn syrup instead of real sugar, because corn is simpler to work with. Many of the low-cost products now sold left and right in the developed world are made, sold, and supported using flimsy automated processes, such that the tendency is for everything we pay for to become disposable.

Wherever the main objective is to cut worker pay or worker numbers, and the result is an inferior product overall, this kind of automation is largely harmful. It hurts the reputation of the company, and brings down the bar of competition, allowing other firms to stop innovating or improving their products in kind.

Companies should not seek to cut their workforces while demand is rising

It seems obviously when you think about it, but how often have we seen companies cutting worker numbers while profits continue to roll in? Those savings then go to stock buybacks and other investor and C-suite goodies, handed out as kickbacks for a job well done (for the investors).

One of the most pernicious trends in automation is the so-called “streamlining” of booming businesses, that cut worker numbers while sales climb. Rather than seeking to increase individual productivity, which helps fulfill the labor displacement cycle, sometimes companies seek to increase production to meet demand while actually cutting their workforces.

This has several negative effects. For one thing, it reduces the institutional knowledge of a company at a time when that knowledge is in greater demand than before. If demand continues to rise, those workers will have to be brought back in, or new ones trained from scratch. This ends up costing a company as much or more than it ever saved, and customers often lose out as a result.

For another, it prevents a firm from increasing its efficiency. While the theory goes that cutting labor force allows a company to seek new efficiencies with the existing workers, the loss of both morale and organizational flexibility can tend to do the opposite. Workers who are under ever greater pressure make more mistakes, as well as find ways to cut corners in their work, leading to poorer results.

It might be desirable to increase production while retaining the same number of workers (although any growing firm ought to think about hiring more people), but it isn’t a good idea to cut your workforce while business is booming. Just ask Tesla CEO Elon Musk, who tried it in 2018, with terrible results.

Companies should look to continue to raise pay if demand is rising

Similar to the last point: a company facing rising demand should consider raising pay to reflect this. Henry Ford is the ur-example, but there are more who have discovered that increasing worker pay can have huge positive effects on the bottom line.

Better-paid workers tend to be more productive and more willing to seek new efficiencies. If workers feel that their employer is investing in them, then they will also be more comfortable investing in their careers and work. They’re also more likely to stay in their jobs for longer, when worker churn can be a huge expense.

Retained workers have better institutional knowledge, loyalty, and overall efficiency than fresh new workers in all but the most exploitative and automated workplaces. If your business model depends on constantly churning through fresh new workers, there may be something wrong with that business model.

As in the recent case of Amazon’s massive logistics operation, a business that continues to grow for too long without improving worker pay, conditions, or retention can lead to a situation where workers simply cannot be found. The well, so to speak, can become poisoned by that company, and workers will know to stay far away from it.

Today large employers like Amazon, Walmart, and Dollar Tree, among others, have issues attracting staff owing to their poor reputations. What your workers say about you matters very much to your future. If all the benefits of rising productivity go toward the shareholders, they notice.

Companies should always seek to retain workers whenever possible

Perhaps it goes without saying that our position is that worker retention is absolutely critical to a healthy firm. While it’s true that not all retention is good retention, and sometimes letting the least productive workers seek different positions is the best move, particularly when demand is slack or shrinking, for the most part, retention is the difference between an institutional brand that can continue to innovate over a long period of time, and one that gets stuck in a cycle of mismanagement and poor morale.

Think of companies like Apple, with retention rates across most of its divisions of an incredible 90%. This amazing retention rate allows firms like Apple to draw on a deep base of knowledge and skills that can prove essential to its future success.

In many cases, Apple products draw upon the experiences of employees who had been part of companies Apple had acquired years or even decades previously. Products and features such as multi-touch, handwriting recognition, Siri, and others were invented by team members who had been working on similar technologies for many years. As the marketplace changes, these dormant skills and experiences can suddenly become highly valuable. A company that has a history of retaining innovators tends to be able to innovate more quickly.