Acton Institute Powerblog

The futility of artificial intelligence economics

Salesforce, an American cloud-based software company, earlier this year announced an initiative to develop an artificial intelligence economist. Stephan Zheng, the lead research scientist at Salesforce Research, describes the moonshot goal of this project as to build a reinforcement learning framework that will recommend economic policies that drive social outcomes in the real world, such as improving sustainability, productivity, and equality.” One of the major requirements he outlines as necessary to achieve such a goal is to “challenge conventional economic thinking.” This initiative requires more than just a “challenge to conventional economic thinking” but a fundamental abandonment of economics as a science.

The science of economics is a social science. Its subject is human persons who are by nature acting persons. Acting persons are always economizing, constantly choosing among different possibilities to realize diverse goals with different degrees of success. The process of exchange produces information relevant to human action in the form of prices. It is for this reason that the German economist Wilhelm Röpke argues in his essay “The Place of Economics among the Sciences”:

Only a market economy makes it possible for economic science to go beyond those general and platitudinous truths and to discover relationships that have the objective definitiveness and validity which a market economy actually establishes by means of the mechanism of price. Only a market economy makes of economic science an analytical social science rather than a science which is merely a descriptive-understanding one having a logical structure like that of historiography.

Beyond simple axioms and truisms such as “incentives matter”:

The particular intellectual effort required of us economists consists in recognizing that economic science deals essentially not with constants but with functions, with relations, with interdependent forces. The logic peculiar to economic science is the logic of relationships.

The complex web of human relationships cannot be reduced to lines of code – even code which can learn – as they are not abstract static phenomena but emergent phenomena within the real world:

As Alfred Marshall once observed, all simple statements in economics are erroneous. But when we modify them and make them conform to pertinent relationships, we soon arrive at a point where the process gets out of control and where it would be possible to reason out economic justification for any abuse that assumes the name of economic policy.

The efforts of Salesforce, while mistaken and ultimately futile, are not without analogues in the dead ends of economic history. Röpke observed with dismay the tendency to regard the whole economic process as something objective and mechanical:

Hence purely mathematical and statistical methods, it seems, can be applied and the whole economic process can therefore be quantitatively determined and even pre-determined. Under those circumstances an economic system readily takes on the appearance of a sort of huge waterworks, and the science which treats of that economic system quite logically assumes the appearance of a kind of engineering science, which teems with equations in ever-increasing profusion. And so oblivion threatens to engulf what, as I see it, is the actual fruit of a century and a half of intellectual effort in the field of economics, namely, the doctrine of the movement of individual prices.

The greatest achievement of economics, price theory, explains how order, cooperation, and coordination can emerge in the real world of risk and uncertainty. In his final book, The Fatal Conciet: The Errors of Socialism, the late Nobel Laureate Friedrich von Hayek elegantly explains how the market process of the emergence of prices generates more real-world information than even the most sophisticated natural or artificial intelligence:

The curious task of economics is to demonstrate to men how little they really know about what they imagine they can design. To the naive mind that can conceive of order only as the product of deliberate arrangement, it may seem absurd that in complex conditions order, and adaptation to the unknown, can be achieved more effectively by decentralizing decisions and that a division of authority will actually extend the possibility of overall order. Yet that decentralization actually leads to more information being taken into account.

Almost two millennia ago, Jesus of Nazareth posed the provocative question: “Suppose one of you wants to build a tower. Won’t you first sit down and estimate the cost to see if you have enough money to complete it?” (Luke 14:28). The mainline tradition in economics, from Adam Smith to Vernon Smith, sees this as a necessary and perennial question which we all must answer. In so doing, we each contribute not only to the realization of our own ends, but we also provide information which aids our neighbor in carrying out his or her duties. Our choices are our own to make not only as a personal right but as a social responsibility. The outsourcing of that right and responsibility to any other intelligence, natural or artificial, cannot lead to true human flourishing.

(Photo credit: Public domain.)

Dan Hugger

Dan Hugger is Librarian and Research Associate at the Acton Institute.