Should Robots Pay Taxes?

In June 2021, we started considering the provocatively titled podcast transcript, “Can a Robot Be Arrested? Hold a Patent? Pay Income Taxes?”, posted on the IEEE Spectrum site. Steven Cherry interviewed Ryan Abbott, physician, lawyer, and professor, about these topics and referencing his 2019 book, The Reasonable Robot: Artificial Intelligence and the Law. We’ve discussed whether artificial intelligence (AI) systems could be charged with crimes or can hold a patent. Whether “robots should pay taxes” turns out to be the scariest question yet.

Touching upon the subject only lightly in the podcast, Abbott details the problem of taxing AI in Reasonable Robot, following this thought process:

  1. Automation using AI threatens to increase human unemployment.
  2. Current U.S. tax law encourages automation through favorable treatment of business investment in equipment (best known are the tax deductions for depreciation of capital equipment).
  3. Businesses pay less in total taxes by having automation technology replace human workers doing the same work.
  4. The U.S. Internal Revenue Service (IRS) currently collects over 55% of its revenue from individual human taxpayers, and over 32% from employment taxes (e.g., Social Security and Medicare).
  5. When automation replaces human workers and increases unemployment, the amount of tax revenue decreases – because machines (AI or not) don’t “pay taxes,” and the businesses don’t pay as much in employment taxes either.
  6. AI-based automation will likely create unemployment, and the government will have to financially support and retrain people who lost their jobs.
  7. AI-based automation will create greater “income inequality” that will lead to social unrest, crime, and self-destructive behaviors.
  8. The solution is “AI neutrality,” i.e., have government make sure that there is no net tax advantage for businesses to use AI automation, and to make sure the tax revenues are equal or increase as AI automation proliferates.

Abbott contends that government tax and fiscal policies should “address the legal imbalance between AI and people and create a more neutral system in which actors will make decisions based on nontax reasons” (p. 49).

Abbott views favorably the notion of guaranteed universal income, the idea that every living human receives a minimum amount of money from taxes paid by others, presumably aiming to share the benefits of AI-generated wealth and compensating for AI-caused unemployment (p. 41).

To help ensure government gets all the tax money it wants, Abbott suggests several plans: (1) shift away from individual income taxation to increase sales taxes and property taxes; (2) devise a system to target tax increases upon businesses that have caused “technological unemployment”; (3) increase taxes upon businesses that generate more profit using fewer human workers; and (4) increase taxes upon capital equipment generally (pp. 45-48). 

Taxing AI Means Bigger Government

Readers might now expect a deep dive into how AI will affect manufacturing, distribution, the labor market, and education options – and then mixing and matching the merits of one or another taxation scheme. Not this time, because Abbott’s presentation promotes thinking about the economy as a cross between a chess game and a complicated Lego project. Abbott follows the intelligentsia pack that presents economic theory questions as puzzles to be solved by experts. The overarching concern in Reasonable Robot is: How will government integrate AI into the economy it is managing?

That line of thinking fits into the vision that expert humans are really smart, and expert AI robots even smarter. The sci-fi picture of human society becomes one of interactions among the “powers that be”: the multinational corporations, the governments, the tech innovators, the individual power brokers, the opinion influencers, and the robots. Just before this future world, however, is the present world, and the current problem is: How will governments deal with the robots?

This whole focus stands upon an unspoken presumption, indeed a fatal conceit: the idea that government entities could actually manage the economy to achieve general prosperity and get even close to the efficiency of a free market system. AI and government in fact do not yield prosperity – free people do.

Government Couldn’t Even Make a Pencil

Cover illustration by Tim Webster, excerpt from I, Pencil, by Leonard Read, published by the Foundation for Economic Education (fee.org), used under Creative Commons License 4.0.

The free market economic system thrives in an environment where every human evaluates his or her situation at any given moment and takes action toward desired outcomes, free to interact and exchange goods and services with other humans, and secure against physical force, theft, and fraud. Human satisfaction is increased with every voluntary transaction in the market. Prosperity rests upon millions and billions of such transactions, the vast majority at a very local level. And the individual choices and transactions occur non-stop, 24/7.

What coordinates this market economy? Not governments. Not computers. Humans – at liberty to envision their future states of being, to choose their actions, and to seek their desired results by peaceful voluntary interactions with other humans – are all coordinating the economy. 

As the main character in Leonard Read’s story, I, Pencil, pointed out:

I, Pencil, simple though I appear to be, merit your wonder and awe … [because] not a single person on the face of this earth knows how to make me.

The Pencil proves its point by canvassing the unchartable ocean of factors: raw materials from around the world, transportation of everything involved, building of capital equipment, designing and specifying components and final product, manufacturing of components, assembly of final products, ongoing communications and dealings among every person and entity at every level – all made possible by humans imagining their personal goals, taking action to obtain them, and interacting peacefully and voluntarily. As the Pencil remarks:

[M]illions of human beings have had a hand in my creation, no one of whom even knows more than a very few of the others. … There isn’t a single person in all these millions, including the president of the pencil company, who contributes more than a tiny, infinitesimal bit of know-how. … Each of the millions sees that he can exchange his tiny know-how for the goods and services he needs or wants … [each person acting in the] absence of a master mind, of anyone dictating or forcibly directing these countless actions[.]

Trillions of Choices and Actions

The primary data item in the market economy is the price, whether for consumer goods or for land and raw materials; equipment for making, repairing and transporting goods; or wages for labor. As with the pencil, each price is determined by myriad interconnections and dealings among individuals and businesses, each seeking to minimize costs to buy while maximizing prices to sell the goods and services. In nearly every transaction the price of the good or service plays a key role. And information about prices affects how people make not only today’s decisions but future decisions. 

Profits and losses concretely show whether each person has effectively combined materials with services to produce something people will buy at a price higher than the total costs of production. Profits encourage future production, sometimes expanded production, while losses inspire cost cutting and possibly discontinuing the enterprise. Every person in the market must account for personal “profits and losses” to guide decisions concerning a particular job, the home, purchases, investments, or outside ventures.        

Contrary to Abbott’s unspoken assumption, government rulers and agencies simply cannot try to run an economy as though it were like assembling pre-fabricated pieces in a Lego project. Long ago it was shown that socialism is fatally flawed because the central planners cannot know enough facts and make enough predictions about the future. Friederich Hayek’s famous analysis, The Use of Knowledge in Society (1945), spotlighted why economic decision-making is best decentralized to the actual human beings involved:

[T]he economic problem of society is mainly one of rapid adaptation to changes in the particular circumstances of time and place, [so] the ultimate decisions must be left to the people who are familiar with these circumstances, who know directly of the relevant changes and of the resources immediately available to meet them. 

Government agencies and taxing authorities are underpowered to understand the intricacies and respond:

We cannot expect that this problem will be solved by first communicating all this knowledge to a central board which, after integrating all knowledge, issues its orders. We must solve it by some form of decentralization. But this answers only part of our problem. We need decentralization because only thus can we insure that the knowledge of the particular circumstances of time and place will be promptly used. But the “man on the spot” cannot decide solely on the basis of his limited but intimate knowledge of the facts of his immediate surroundings. There still remains the problem of communicating to him such further information as he needs to fit his decisions into the whole pattern of changes of the larger economic system.

That communication occurs via the prices of real estate, materials, machines, labor, services, and products at every level. Hayek and other market economists have thus realized: “We must look at the price system as such a mechanism for communicating information if we want to understand its real function.” Prices tell people what they need to know about scarcity, difficulty of providing or delivering, and relative availability of items needed in their business, academic, and private lives. And such full-meaning prices can never come from governments, only from the economy-wide web of planning, working, dealing, investing, buying, and selling.

Realizing the exponential complexity in the modern technology-enhanced economy inextricably linked with people worldwide, government tax decisions cannot be treated like chess moves – Knight to King Bishop 5 – thinking the responding party will always do what you expect. Even though IBM’s Deep Blue computer chess system could usually beat the reigning world champion Garry Kasparov, the computer could not predict Kasparov’s every move – Kasparov won some games and tied others. Central government cannot direct the economy with any precision, and certainly not better than the free market network of intelligent humans who comprise the economy.

Central government cannot direct the economy with any precision, and certainly not better than the free market network of intelligent humans who comprise the economy.

AI Taxation Centralizes Power

Back to taxation of AI automation: Abbott’s policy suggestions read like Lego project instructions, giving the impression that “all you have to do is…” If some workers’ jobs are replaced by AI, then that means less tax money for the government. Solution? Impose a new tax upon the business that uses the AI system, and spend the tax receipts on government employee salaries, unemployment benefits, universal basic income payments, or training programs for displaced workers. Problem solved!

Ignored in the government “policy A gives result B” approach are several facts of economics. Abbott advocates AI neutrality and thus “tax neutrality” between hiring workers and deploying AI systems in place of workers, but the plan of imposing new taxes on businesses that use AI is nowhere near the precision tool it sounds like. 

First, taxes are costs of production but in a competitive market they cannot simply be “passed on” to the buyers. Buyers don’t simply pay higher prices for a product or service just because the taxes are higher. They use less or shift to substitutes, and sometimes discontinue a now-unprofitable line of business altogether. 

Second, a tax policy today will reverberate through the business community, with entrepreneurs and managers looking for ways to minimize the impacts of the taxes. The effect of a new tax on Day 1 will not stay constant through Day 100. Market forces and human decisions will change the situation over time. Because the numbers of ongoing and future decisions are so huge, and each decision arises from a human’s analysis of each affected transaction, the government agencies will have no way to forecast the effects of the tax once and for all

Third, how the government spends the new tax money will itself impact the economy broadly. Unemployment payments may delay workers re-entry into the labor force. Training programs favored by government bureaucrats will see new opportunities and profits, while others will lose out. The effectiveness of the training is not predictable, and the needs for training will themselves be a moving target. Increased government workers with increasing salaries will receive funds for doing nothing more than redirecting money from people who earned it by producing something to people who didn’t.

Fourth, there is no such thing as an unbiased government taxing or spending policy. Even governments in the self-professed free U.S. economy quite regularly implement tax policies – exemptions, deductions, phase-outs, and credits – that favor one industry, region, or group of people over another. The notion that politicians and bureaucrats will treat all AI automation situations identically is not credible. There will be favored and disfavored objects of tax law. The same is and will always be true with government spending. This means that the more government attempts to rule AI in the economy, the more that stakeholders will fight to control that discretionary power. 

In AI discussions a mystique encircles the subjects of what AI can do and whether it can be limited to doing only “good” things. The mystique seems to overshadow seriously thinking about how much money the government should coerce from its citizens and how much power it should exert over human choices, options, and decision-making. The conversation must face a key point: The bigger the government, the smaller the citizen. 

When humans and robots become just “factors” in government decision making, we get Abbott’s AI-neutrality incarnate. The idea that big government tax and spend policies are the way to “equalize” humans and robots ought to trigger fears of dystopia as never before. 


You may also wish to read the first two posts in this series:

Can a Robot be Arrested and Prosecuted? An Uber driver is held liable if he runs over someone. But what if a driverless taxi ran over someone? A legal culture should be developed now to hold people accountable for AI software code designed to steal, cause damage, or even kill. (Richard W. Stevens)

Can a Robot Hold a Patent? The boring answer is no, but the question raises intriguing thoughts about AI and intellectual property law. As spectacular as the AI systems can be, the results of AI computing still owe their existence to humans at all key stages. (Richard W. Stevens)

Leave a Comment