TheMarketForIdeas

There is no such thing as a closed science – the problem is not the (un)expected novelty, but the old, still unresolved epistemic conflicts. It is also the story of economics. From the classics and the revelation of price as a social compass, to the marginalists and the subjective revolution of utility; from the Keynesians, with their emphasis on aggregate demand and public policies, to the monetarists’ reply and then to the rigorous micro-foundations of the neoclassical “synthesis”; from the new macro- to the behavioural economics to be incorporated into fancy impact assessments (which no longer add apples to oranges, but make juice out of the both). Each stage wanted to have the “last word”, but the quarrel of the wise (among themselves and then with the world) continues unabated. The contemporary “synthesis” is not a stable temple, but a construction site. With plenty of scaffolding and sequelae.

And in the era of concubinage with Artificial Intelligence, at the heart of the dispute remains a blunt quiz: do we study human action, oriented by goals and meaning (teleologically), or impersonal aggregates, governed by stable relationships (physically-mathematically), Machiavellianly-manipulated, further on, in social laboratories? The first perspective sees the agent as an author, capable of learning and innovating, of making mistakes and rewriting his preferences; the second translates society into vectors, systems of equations and statistical averages. Teleology takes seriously the longings, rules and narratives that power decisions; determinism demands regularities and distributions that are smooth enough to be calibrated. We all live in the “customs office” between the formalism that extracts us from the anecdotal and the anthropological that separates us from the zootechnical.

AIconomics forces us to constantly negotiate this epistemological frontier, but full of ethological consequences: for example, regarding the relationship between the freedom of agents, through markets, versus the corrective imperatives of states/governments. Markets decipher dispersed information through prices and incentives, ingeniously scaling anonymous collaboration into spontaneous orders; but they are not flawless, being burdened by (negative) externalities and (insufficient) public goods, by information asymmetries escalated into abuses of power. Public policy exists precisely to repair these failures, but it comes with its own dangers: latencies and errors, instability and corruptibility. Mature economies seek to combine institutional architectures that maximize freedom compatible with the dignity and safety of others, minimizing social regret in this bitterly-uncertain world.

How does AI enter the profoundly human story of economic science? Operatively and tactically, it does “more, faster”: it sanitizes data, extracts patterns, runs scenarios, finishes forecasts and automates tasks (from nowcasting to anomaly detection, etc.). Yet, couldn’t the impact be even more strategic: not just fuelling theories, but… re-interrogating them from scratch? AI can generate hypotheses, simulate artificial ecosystems with a multitude of agents, test the robustness of mechanisms in counterfactual worlds, and capture nonlinearities that are alien to standard models. But what if AI might convince us that our ultimate trait – “free will” – is nude in front of “Laplace’s (deterministic) demon” and that we are neither too free, nor so surprising, nor able of who-knows-what revolutions, but poor beasts, calculable and passive-reactive? What’ll remain of us and our theories about ourselves?

Without going that far, can we rewrite, with the help of AI, the micro and macro foundations? Shall we see economic person and society “differently”? Our digital traces already make previously invisible dynamics observable: networks of influence, emerging norms, social learning in real time. Models with heterogeneous agents, trained by machine learning, can explore how, “out of the blue”, rules are born, expectations are formed and crises occur. Micro becomes less about “representative” and more about distributions and interactions; and macro, an emerging phenomenon of connections. They can also be fused, after the Keynesian fission. And instead of equilibrium as an analytical obsession, adaptive processes, fragility and resilience prevail. However, an ethics-bias caveat: let us not confuse simulation with society, nor the frigidness of “its” agents with “our” human()kind(ness).

The irony is that the same AI capable of illuminating the mechanisms of freedom can build economic architectures that rob people of choice: it starts modestly with hyper-personalized recommendations that capture you, opaque dynamic pricing, irreversible smart contracts, programmable payments, auctions in which algorithms bid on your behalf. If the mechanisms optimize proxies of well-being, they may maximize compliance, not autonomy; and the promised efficiency may come at the cost of losing the “right to change your mind.” AIconomics must therefore include a new chapter: the principles of designing systems in which AI remains the servant of human decision, not its surrogate. The solution is to provide pause buttons and preserve the superpower of… the plug. “More and faster” has an opportunity cost – the “old economics” does warn us: the nostalgia of a quiet life.